The dot-com bubble didn’t prove the internet was a fad — it proved the internet was inevitable, but the valuations assumed adoption would happen in 2 years instead of 15–20. To me it feels like the AI inevitability will be much quicker.
> To me it feels like the AI inevitability will be much quicker.
Based on what? We're only seeing linear improvements for increasing spending. There's no new algorithm ideas on the horizon, just more and more hardware, in the hopes that if we throw enough RAM and CPU at the problem, it will suddenly become "AGI."
No one has their eye on power budgets or sustainability or durability of the system. The human brain has such a high degree of energy efficiency that I don't think people understand the realities of competing with it digitally.
The main problem "AI" seems to solve is that humans get bored with certain tasks. The language models obviously don't, but they do hallucinate, and checking for hallucinations is an exceedingly boring task. It's a coffin corner of bad ideas.
The LLM algorithms seem pretty clunky to me, like a hack designed for text translation that surprised everyone by getting quite smart. The reason the human brain is so much more energy efficient is quite likely better design/algorithms. I was watching some video comparing brain and LLM function and almost tempted to try building something myself (https://youtu.be/3SUqBUGlyh8). I'm sure there are many more competent people looking at similar things.
Everyone says "those autoregressive transformer LLMs are obviously flawed", and then fails to come up with anything that outperforms them.
I'm not too bullish on architectural gains. There are efficiencies to be had, but far closer to "+5% a year" than "+5000% in a single breakthrough".
You can try to build a novel AI architecture, at a small scale. Just be ready. This field will kick your teeth in. ML doesn't like grand ideas and isn't kind to high aspirations.
Physics is obviously incomplete and yet nobody can solve quantum gravity. Being obviously flawed doesn't mean the solution is obvious. That's the whole problem.
I think in this case, people tend to underrate just how capable and flexible the basic LLM architecture is. And, also, underrate how many gains are there in better training vs better architecture.
Most people are not ML researchers. Most of the AI industry is not AI researchers. Most of the AI spending is not going to AI researchers.
AI researchers came up with an architectural improvement that made a lot of previously impossible stuff barely possible. Then, industry ran with it. Scaling that particular trick to the limits by throwing as much raw compute and data at it as humanly possible.
You don't need to be an AI expert to know that there are probably more advances to be had and that funding foundational research is the way to get them.
Internet did not have enough devices to reach people. At the height of 2002 only a fraction of people worldwide had an already expensive computer and an internet to go with it.
I ran a e-commerce startup from 2005-2010. Having access to demand is a thing.
Today everyone has access in their pockets. Go to small city in Africa, India and China, and observe how they use AI. See how perplexity has put AI answers in hundreds of millions of people's hands before Google in a matter of months.
Forgive me for saying — but "Based on what" for comparing accelerated adoption between 2005 and 2025 — is discarding many huge elephants in the room, starting with that small thing you're reading this in your hand with, and the invisible thing that's sending you this comment.
The current LLM models are too fallible and inefficient for everyday use. Energy requirements have become a major concern, bucking the trend of banking exponential computing gains through the efficiency improvements. Until recently, year-on-year global energy demands of data-centers were increasing linearly despite the exponential increase in computing power.
This has changed since the cost of equipment and infrastructure has been close to "free" for the large corporations running a cycle of funny-money investment in a bubble. This has allowed them sell access to the computing and models for just above operational energy costs (to the extent of increasing global energy prices), whilst offering free accounts to harvest data. No small competitor could possibly compete with that model.
The calculation for profitability (useful output for humans in a cost-benefit analysis) of the current setup is broken, and trades on our dreams of the future. Scaling computing forever simply does not work. It will never be profitable without further leaps forward in the technology--either more efficient models or hardware. By this time, the extent to the excessive investment in new data-centers will be clear.
The internet predated dot-coms by two or three decades and wasn't very bubbly - mostly government funded links between academic and military institutions. It was only when commerce got in there in the 90s that things started getting busy and then bubbly.
A lot of those early .com companies could have been profitable. They chose to go for rapid growth instead. Some people here probably remember discussion of users mattering more than revenue.
You see that same pattern with AI now. Products are being provided for free or nearly free, and plenty is being spent on marketing.
Technology for all the above existed in rudimentary form, faster Internet, faster machines and adoption was missing. But Current bets are assuming AGI. No one knows how soon. To predict that would be foolish.
If technology had been the same for the past 20 years, basically none of these would have existed, or would be even close as large as today. We needed way faster cable and mobile internet, and smartphones. Probably even smaller laptops. It was possible to predict these more or less, however, it was impossible to predict when or whether people start to really utilize the internet. Even now, we needed COVID to have another shift regarding this. The general acceptance of “internet first” kind of worldview maybe would have never happened without forcing us to have.
How does this add to the discussion? Is the goal to make HN as toxic as everywhere else online? If you have something to say, say it. Otherwise this performative negativity and cynicism is boring honestly.
On the RHS, post hype, the second movers could work on the boring, unsexy problems in those domains nobody wanted to solve. And solve them extremely well. Then build a moat around that.
There is also a customer adoption curve of technology that lags far behind the technologist adoption curve. For example video on the Web failed a long time, until it didn't, when Youtube began to succeed. The problem became "boring" to technologists in some ways, but consumers gradually caught up.
Blockchain the the internet seem to have stabilized. The internet as fast video capable links between computers, blockchain as a tech for speculation, gambling and some criminal stuff. AI has not and is still on the exponential bit of the S curve.
Aside from a belief that the AI adoption will happen very quickly, which maybe that’s your main point, you’re not really disagreeing with the article:
> All this means two things to us: 1)The AI revolution will indeed be one of the biggest technology shifts in history. It will spark a generation of innovations that we can’t yet even imagine. 2) It’s going to take way longer to see those changes than we think it’s going to take right now.
The internet was monumental and valuable, offering instant conveyance of media and data. AI is monumental, allowing instant access to near infinite data; but whether instant access is as valuable as instant conveyance yet alone five times the value, appears to be the question
Honestly anyone who thinks AI has intrinsic value to rival the GDP of nations is a bagholder in denial and I'll be happy to buy your puts.
"instant access to near infinite data" is Google search.
AI if it gets better will be thinking / intellectual work which is a different category.
I remember in the 90s people would say the internet will bring instant conveyance of media including video and people were like yeah right I can just about get one page of crappy text in a minute over the unreliable dial up modem. It's like that now with AI thinking.
To say whether it's a bubble, we need to know the value of the technology.
The value of modern AI seems very high. That nobody knows how high, that they still haven't figured out applications, and that the technology and its tools are still far from refined, is normal for any new technology.
If you add the value of the potential political power gained by controlling AI, then the value to the owners and investors is astronomical. Many of the investors have demonstrated a strong motivation to sacrifice money for political power, for example by supporting nationalism that undermines a global economy that they benefit enormously from. Somewhere, I read someone explaining their investment by saying 'it's the greatest transfer of power in (modern) history'. Also see: https://news.ycombinator.com/item?id=45983700
The implied promise was that these things were going to “revolutionize the workplace” i.e. massively automate middle class office jobs
A couple of years down the road, their useful applications still are summarizing text, transferring style to text, generating code under strict supervision, and generating images that need retouching.
That’s a lot to get out of a tool, but it’s dubious that investors were pouring trillions of dollars into these things thinking of automating away junior software engineers and low end design work.
Edit: I forgot their other niche, that of generating homework and school test answers
You forgot - cheating on job interviews, writing resumes to be repetitive, and adding an annoying flowery tone to non-native English speakers who think AI wrote something for them that isn't AI-obvious.
I wonder how many market inefficiencies this creates. People with worse education, people who cheated their way to a job opening compared to a better candidate, etc. Basically counteracting the productivity gains AI was supposed to bring.
Many of those things were already at a “good enough” level since GPT-3.5.
There’s probably a good business usecase there for companies wanting to have smoother communication with offshore teams.
Could that be a game-changer? I wouldn’t discount it, but it does sound like something that has to operate at a very low margin and that doesn’t merit a lot more investment.
How many people graduate from a US software engineering degree each year? About 100k? If they (the 100k in the US) earn $100k each in the first year, before gaining the skills to earn more, that's $10 billion a year, every year. If you can capture that market for next 20 years, it's worth $200 billion.
Except… can you capture it? A junior dev is… not exactly someone you want connecting to your business-critical database without supervision, and a real human dev will get better with a predictable rate. Will LLMs get better? The makers are betting on that, but we'll only know after the model releases, and even then after we play with them for a bit to differentiate between the record performance on whichever benchmark and the actual work we want them to do.
Then there's the question of can you really keep an edge for 20 years with investments today: Sometime between 2030-2035, there's likely to be models matching 2025-SOTA performance that run on ${year}'s high-end smartphone.
(Well, unless we all die in WW3 because of Russia getting desperate from its failure to remove Ukraine's sovereignty, or because China has a hot war with Taiwan and/or the USA messing with global consumer electronics supplies, but I don't think those get priced into the market…).
But that's just the USA's software developers in just their first year after graduating. Software devs are 1% of the US job market, the first year after graduation is (66-21=45 years, 1/45 ~= 2%) of a working life, the US is just 4% of the world's population/25% GDP.
For the 1% to matter, there have to be other jobs that LLMs can do as well as a fresh graduate. I don't know, are LLMs like someone the first year out of law school or medical school, or are those schools better than software? Certainly the home robotics' AI are nowhere near ready yet, no plumber, no driver (despite the news about new car AIs), would you trust an Optimus to cut your hair? etc.
For the 2% to matter, depends how seriously you take the projections of improvements. Myself, I do not. Looks like exponential improvements come at exponential costs, and you run out of money to spend for further improvements very quickly.
For the 4% to matter, depends on how fast other economies grow. 4% by population, about 25% by GDP. I believe China is still growing quite fast, likely to continue. Them getting +160% growth, and thus getting 2.6x times the money available to burn on AI tokens, over the next 20 years would be unsurprising.
All in all, I don't think the USA is competent enough at large-scale projects to handle the infrastructure that this kind of AI would need, so I think it's a bubble and will burst before 2030 because of that. China seems to be able to pull off this kind of infrastructure, so may pull ahead after the US does whatever it does.
> For the 1% to matter, there have to be other jobs that LLMs can do as well as a fresh graduate. I don't know, are LLMs like someone the first year out of law school or medical school, or are those schools better than software?
Before looking to medical and law schools, I might look to middle-manager school or salesperson school or bookkeeper school.
I don’t know enough to speculate even beyond those crude guesses, but as I thought about this question, I found it interesting to skim the US’ employment-by-detailed-occupation chart:
That analysis only makes sense if companies value AI tools as much as equivalent human productivity. Hypothetically say you have a company with 100 junior developers. An AI service comes around that doubles the productivity of your junior developers, so you can keep only 50 of them. Would the company pay 5M dollars a year forever for that service?
In my experience, the answer is a resounding no. They’ll nickel and dime some kind of per-seat licensing on a monthly basis that costs less than 100 USD or whatever. So for every 100$ in salary you can automate away, you might get 2$ in subscription payments if you are lucky, at current rates.
The hardware also wears out really fast. And every replacement is more expensive. How long can that party keep going when none of the companies make enough revenue to cover the expenses?
It seems like that at first glance. But in reality, GPUs have had extremely slow adoption for real-world operational meteorology applications. Because of the fundamental design and architecture of most NWP systems, it was very difficult to leverage GPUs as compute accelerators; most efforts barely eked out any performance gains once you account for host/device memory transfers. It really wasn't until some groups started to design new weather modeling systems from the ground up that they could architect things in such a way that GPUs made a significant difference.
Obviously AI / ML weather modeling is a different story.
As someone working in a field that has used NLP for quite some time - yeah, I generally agree that those investments are worth their weight in gold... which is unfortunate because before ChatGPT came along they were viewed as niche unprofitable money-sinks. The astronomical investments we've seen lately have been in general models which can be leveraged to outperform some of our older models but had we wanted purely to improve those models there were much more efficient ways to do so.
Hopefully we can retain a lot of this value when the bubble bursts but I just haven't seen any really good success stories of converting these models into businesses. If you try and build as a middleman where you leverage a model to solve someone's problem they can always just go to the model runner and get the same results for cheaper - and the model runners seem (so far - this may change) to be unable to price model usage at a level that actually makes it sustainable.
Those older models running specialized tasks seem to be trucking along just fine for now - but I remain concerned that when the bubble bursts it's going to starve these necessary investments of capital.
I think it's pretty clear to all the big operators that they will need to go whole hog into ads and take some of the Google/Meta pie. It's just a matter of time.
You're missing the point. Those kind of narrow AI applications are not the motivation for the trillions of dollars being poured into AI. Of course AI has a variety of applications many disciplines, as it has for decades. The motivation behind the massive investment in AI is as forgetfulness said, reap the benefits from "revolutionizing the workplace"
Eh, those applications (incl. protein folding) existed for a decade-plus before LLMs came onto the scene, and there was absolutely nothing like the scale of capex that we're seeing right now. It's like literally 100-1000x larger than what GPU hosting providers were spending previously.
That’s copium, as the kids say nowadays. The massive planetary investment is a 100% for AI chats. All those other things are taking the crumbs where they can.
The really weird thing is that Big Business actually is buying supercomputer clusters to do just that. I can't really talk to the government side but a lot of businesses' early forays into AI was just slapping a chatbot on their product and hoping it'd attract a lot more business. I also think you'd be surprised how integrated really dumb chatbots are into business communication these days.
I think most smart people are looking seriously at different models to try and improve the accuracy of any existing ML uses they had in their business but the new features built post-ChatGPT tend to often just be fancied up chats.
That's happening of course, but that's not really the whole picture. Any org that already invests in R&D is likely considering or already implementing modern AI tech into their existing infrastructure. A big oil or pharmaceutical or materials company likely doesn't care much about chat bots, or any customer-facing tech for that matter.
Actually, big orgs are doing exactly that; slapping a chatbot onto their support ticket backlog. Being really, actually “data driven” is hard, and must happen from the bottom up. So instead there’s chatbots in their frontend and support backend, but the backend doing the actual lifting probably hasn’t changed one bit.
That's the same mistake made with every new, and eventually successful, technology - we haven't found a valuable application yet, so the technology is not valuable.
Finding the valuable application is often the hardest part. That it hasn't happened yet is meaningless. Some technologies sit on the shelf for decades.
AI seems to have a lot of potential: It may be the most valuable technology ever; it may not provide more value than it does now, or something in between. Nobody actually knows. The challenge of innovation is managing that irreduceable risk. It starts by accepting risk, accepting that you don't know. One wrong way is to deny the risk - denying uncertainty - by either saying it's worthless or that success is guaranteed.
> thinking of automating away junior software engineers and low end design work.
And it's really still very arguable imo if it's even doing this
Like you said, it still needs strict supervision. In my opinion it is not a good use of your supervisory time to be babysitting an LLM versus mentoring actual juniors
> In my opinion it is not a good use of your supervisory time to be babysitting an LLM versus mentoring actual juniors
Right. Because at least juniors are supposed to learn and at some point become senior and stop needing this kind of supervision. Also, interacting with people can be more rewarding (or not, depending on the people)…
Ok, ignoring any AGI or massive advances, let's just say an LLM can help the average office worker be 15% more productive... what do you think the economic value of that is?
100 billion a year, with math that’s downright delusional.
So, back of the envelope math, the US GDP is 27.72T USD, 80% (22.18) corresponds to the tertiary sector.
Let’s say that this is a 15% increase over a 10 years period, because YoY a boom like the computer revolution itself looks like 2% increases a year.[1]. This amounts to about 1.5% increases each year.
Let’s just make the huge leap that you can just scale the productivity up of all this just by making typing out reports and emails a faster activity, and summarizing information for which you’re not facing liability if the bot gets it wrong.
Yes, including the nurses, cleaners, truckers, teamsters, all of it.
1.5% of it is a cool 330 billion.
How much of a cut of that productivity increase could AI companies take? 30%? That’s 100 billion in one year there.
So with pie-in-the-sky math, they could break even if their obligations throughout the decade don’t amount to 1 trillion (since those 1.5T in bonds issued this year mature in longer periods)
Not only that, but it’s devaluing the sacred cows of the very same companies that are investing heavily.
Search is dead or dying
Social media is dead or dying
Content creation is dead or dying
If they cant make AI work, then they are left with AI at a level that continues to devalue their core business.
They have no choice. They made a deal with the devil. And the devil means to collect.
This is why I think Apple is lucky their attempt failed so bad. They dodged a bullet. They have an opportunity to guide a lost tech industry through a post AI bubble world.
The value of LLM is reliable high quality translation between all languages. The economic value of this is at least trillions of dollars per year. The cultural and humanitarian value is equally gigantic, even if it can't be measured in dollars and cents.
That is of immense social value - but manual translation services for commercial projects (like application localization) is already dirt cheap to do and automatic casual translation services for consumers would be incredibly difficult to monetize.
I will freely admit this - but I have led a localization effort on a mobile game and worked on localization for a desktop application. I am also lucky enough to travel abroad quite a bit and am quite familiar with consumer offerings.
So I would fairly limit my experience to consumer and medium-sized business uses - I have no experience with large corporate translation efforts (the largest would probably be Ubisoft or the Mouse-Ears company's gaming divisions if you consider them large) and even the small mobile game company I worked at had a budget in the millions range. It certainly hasn't been a focus of my career but I feel comfortable standing by my statement above.
I notice that angry hackers are going nuts with their down votes. Make an argument instead if you have something to say.
Tourism and travel is one of the worlds largest industries, and companies in that sector can improve their revenue with double digits if they can reach and communicate with more customers. That's real impact, which is worth something, while hacker votes are worth nothing.
Alternate viewpoint - The value of modern AI seems very high. That nobody knows how high, that they still haven't figured out applications, and that the technology and its tools are still far from refined, might be a sign that the technology shouldn't be valued this highly at all
> That nobody knows how high, that they still haven't figured out applications, and that the technology and its tools are still far from refined, might be a sign that the technology shouldn't be valued this highly at all
That might be true, but it also might be true that it's tremendously valuable. The conditions you identify exist for every new technology; obviously the outcomes vary greatly.
> The value of modern AI seems very high. That nobody knows how high, that they still haven't figured out applications, and that the technology and its tools are still far from refined, is normal for any new technology.
The same could be said for the internet. But, and I know this will be hard for younger readers to believe, I seem to recall the value proposition of the Internet being more immediately apparent at the time.
From what I've read, people spent a long time looking for the 'killer application'. Google didn't know how to make money off its search engine. Social media didn't exist in any mass form. Internet access on phones, beyond email, didn't exist in any usable way.
Proprietary, walled garden services, with their own dial-up numbers, such as AOL and CompuServe, were seen as the future. Microsoft thought their similar service (MSN?) was the way forward and didn't integrate the Internet into Windows 95. That was after Netscape's browser was released and (relative to the time, I'm sure) very popular.
The value was obvious almost immediately, but it was obscured by flashy get rich quick attempts. Google chose not to make money at first but that was a business decision, not a necessity: people were paying tons of money for ads and referrals years before they were founded.
That’s part of the difference with AI now: the internet immediately generated significant value almost instantly for relatively modest investments, so there were a ton of small and mid-sized companies who saw almost immediate profits from moving catalog sales or support online, switching software distribution from mailing diskettes to downloads, etc. and some people were able to cut into established markets where the existing companies were slow to change.
AI is different in two key ways: the first is the extremely high cost, which prevents a lot of the bottom-up growth we saw on the web, but the second is that it’s not a reliable technology like the web was so you can’t safely use it in the most valuable contexts. In the 90s, some people screwed up form validation badly but most people didn’t. Right now, companies would love to replace things like customer service or sales with chatbots but nobody on the market can make a system which is both safe and useful because things like prompt injection require theoretical breakdowns, not just attention to detail.
> The value was obvious almost immediately, but it was obscured by flashy get rich quick attempts. Google chose not to make money at first but that was a business decision, not a necessity: people were paying tons of money for ads and referrals years before they were founded.
All I can say is, that was the opposite of what I've heard. Did you participate?
Also, Google was many years after Netscape's IPO, seen as the first SV boom. And what about Windows 95, if the Internet's ROI was so obvious?
There's a lot of variety in the quality of writing from that period. At the time, a lot of “analysts” were doing lazy number-go-up stories rather than the harder work of actually looking at business fundamentals—I remember so many stories which were basically trying to rationalize the greater fool theory—and there was also a tendency to lump all of the internet companies together. Amazon.com was an interesting example of the latter phenomenon where people would talk about them being unprofitable because they were just looking at the total numbers but completely missing the distinction between the dotcoms which were losing money on every sale with no way to close the gap and Amazon, which was expanding rapidly but was profitable in each business segment within a couple years (e.g. books were cash-positive circa 1995) and could go from reporting losses to profits any time they wanted simply by slowing their expansion plans.
In Google's case, the key thing was that they were out-competing in what was already an established market. Search companies were making money from ads and sales referrals, so the questions were things like marketshare and efficiency rather than the ability to make revenue at all, whereas Netscape struggled with that once Microsoft used their Windows monopoly to set the price for browsers at $0. They tried to sell server software but that was a different market with limited synergy for browsers and there was a ton of competition even if they'd been better at writing software.
I mention all of that because one thing to remember is that the dotcoms weren't the only game in town. A lot of people focused on them because the IPOs and market valuations were wild by traditional standards but most of the real money and energy was in established businesses moving online since there were clear and immediate benefits from doing so. When the bubble popped, that stuff didn't really change and a ton of people moved from dotcom jobs to traditional companies.
I worked for a web development company at the time and we were turning away business for years because there were so many companies looking to add a web presence to their existing business. It made quite an interesting contrast to our startup clients who tended to be a lot more focused on stock prices rather than business fundamentals. I remember being on a call with pets.com people who were basically openly stating that they were basically frittering time away waiting for the IPO to make them rich, whereas our customers who were in manufacturing, banking, insurance, etc. were a lot more realistic about working within a budget to do something they would actually turn a profit from.
Re: Windows 95, the web drove a lot of its success. That was around the time normal people could see a lot of immediate personal benefit from being online–shopping, recipes, stock trading, research, forums, games, dating, porn—and Win95 was the first version of Windows to ship with built-in TCP/IP and a web browser, not to mention being a lot closer to the competition on usability. For a lot of PC users, especially in business, the web was what sold them on the need to pay more to have a GUI, mouse, etc. and while that was possible with Windows 3.1 it really went mainstream around the time Windows 95 came out (Windows NT was too demanding on hardware for most budgets outside of engineering/development).
Thanks for all the detail from someone who was there.
> completely missing the distinction between the dotcoms which were losing money on every sale with no way to close the gap and Amazon, which was expanding rapidly but was profitable in each business segment within a couple years (e.g. books were cash-positive circa 1995) and could go from reporting losses to profits any time they wanted simply by slowing their expansion plans.
Great insight. And of course Amazon would raise any funds possible to expand into this new frontier.
> Win95 was the first version of Windows to ship with built-in TCP/IP and a web browser, not to mention being a lot closer to the competition on usability.
Per Wikipedia:
"Windows 95 originally shipped without Internet Explorer, and the default network installation did not include TCP/IP .... At the release date of Windows 95, Internet Explorer 1.0 was available,[48] but only in the Plus! add-on pack for Windows 95, which was a separate product. ...
Windows 95 OEM Service Release 1 was the first release of Windows to include Internet Explorer (version 2.0) with the OS."
My point is not to nitpick, but that the Internet's value wasn't obvious to Microsoft at that point.
There was a recent Bezos talk about the fact that (much like the dot-com) we've overspending on infrastructure that'll bubble and implode but then we'll have all this amazing infra for companies to build off of... but the process of that overspend and implosion is essentially a massive debt erasure - a lot of people are currently propping up this market with their capital and the companies they're propping up will collapse and those obligations liquidated - and that will result in massive society-wide pain. We may end up in a better place for the next generation because of this investment - but if you're a retail investor don't expect your 401k accounts to weather that burst gracefully... and, unlike the boomers, this will largely hit Millenials and Gen Z both of whom are currently under massive financial stress.
Arguably, it's exactly this mindset creating the problem. It's not the value of the technology that matters. It's the value of the companies. If one company was the only one that had a particular valuable technology, then that would matter, but otherwise 90% of them likely end up worthless even if the sector itself doesn't.
How that washes out on net we'll have to see. I'm not gonna pretend I know more than anyone else. Just keep in mind that a major difference between now and 2000 is companies stay private a lot longer. An IPO forces you to open the books and sustain public scrutiny of the broad investor class. A still-private startup only needs to convince one funder of their value. That inevitably leads to higher variance and a greater risk of failure. That doesn't mean the whole sector is necessarily a bubble, but if it is, it can be sustained a lot longer without us knowing. A small number of people with $50 billion they need to park somewhere and no other obvious options can keep shitty ideas afloat in a way that wouldn't be possible if they had to be subjected to broader exposure. We like to believe people with $50 billion can't be wrong, but the wisdom of crowds always beats the wisdom of individual genius.
the real answer is that the applications for the military, surveillance, and population control are proven, and the pathways to scale those capacities are clear, so the money will pour in no matter what. the implication is that we had better come up with some more consumer/humanity friendly applications that create comparable value, or that is all we will get.
The value can be incredibly high... just like the value of the internet was incredibly high.. and it can still be a finanfial bubble. I feel like the people arguing against the bubble perspective are saying look the internet was actually valuable and so too will this.. and that's totally a valid perspective. The bubble is not about whether there's value but about whether or not the market will come down. All of the above can totally happen
Very useful, clearly. But valuable? In aggregate it seems clear that it is. But where does that value acrue? It seems to me the value will be thinly spread while the costs are concentrated.
It does not seem possible that any conceivable business can pay for all the announced plans for developing data centres, nor energy available to power them.
If AI systems can be developed to be trust worthy enough to act on their own, none so far (?), then I can see where the value could acrue, but as things stand?
> It seems to me the value will be thinly spread while the costs are concentrated.
Won't the value will be concentrated in the few with the scale to build competitive AIs?
In a sense, costs are spread thinly across the content creators whose work has been stolen. (But of course, it is indeed incredibly concentrated in building datacenters.)
> Won't the value will be concentrated in the few with the scale to build competitive AIs?
No.
Open AI and their ilk can sell tokens, but how much business is there in that? Trillions? If not, they're bankrupt.
Dead man walking. Zombie
The open and free models are very capable and getting better.
If there is a breakthrough beyond transformers and neural nets that means trustworthy AI that can be agentic (does not exist, the "trustworthy" part) then it will be exponentially more valuable. But that was as likely in 2020 as in 2030.
Meanwhile something like 10% of the USA electricity supply is being "committed" to datacenters and no practical plans to supply it
Meanwhile the Chinese are leaping ahead and in a more community orientated open way. The cognitive dissonance is quite loud for that one!
Current IP laws are too restrictive, but scraping public facing websites is not theft even under those crazy laws
I have no moral objection to the use of Anna's Archive to train models. A good use, if illegal. But I do recogise and condemn the hypocrisy of Meta doing it. Not surprising, just disgusting
The wasting of bandwidth is annoying, so silly. We should be better organized
Calling it theft is buying into the mad bad and sad IP legal thinking that has become the dominant paradigm in the West
In a way, I agree with your last sentence. But there is a difference, and that is power.
As a simplistic hypothesis: The already-powerful and super-wealthy taking all that IP from the small content creators is theft, IMHO. The small content creators who can't afford JSTOR finding some free source for a paper, etc. is democratic, empowering, and enriches our society.
I worry that it is not a valid perspective and that the bubble dynamics are being driven exactly by this sentiment.
We are pricing in the hollywood version of AI we don't have as if it is the internet.
I communicated with people exactly like what I am doing right now in this post on usenet in 1995. Message boards by 1997. The internet bubble wasn't based on some wild virtual reality version of the internet that hadn't been invented yet.
It is a categorically different process at work here. This bubble is far more insane and speculative since we don't have what we are pricing in. We don't even really know what we are pricing in besides some vague notation of an inevitable "AGI".
Yep. It seems somewhat inevitable to have a readjustment at some point. Both Web 2.0 (online retail) and AI are (rightfully, I believe) creating a lot of buzz and stimulating a lot of investment. It's all very new and exciting and there is bound to be a lot of hit and miss activity. Once the dust settles, the misguided bets will become more obvious and there will be a culling, just as we had during the dot com bubble.
Absolutely nothing in your list prevents the AI from being a bubble. The Internet is an absolute marvel of engineering that completely changed the life of the majority of people in the world, and yet it crashed 25 years ago, and crashed hard. Both can be true at the same time.
> To say whether it's a bubble, we need to know the value of the technology.
Not really. I mean, not only. The value of the web is immense. And yet, the dot-com bubble was indeed a bubble. What matters is the value in the short term compared to the value of the companies in the current context. Even if AI is huge 20 years from now, it can still crash dramatically tomorrow.
I was going to say that. Also the bubble can be in some areas but not in others. I think the proposed data center build out is a bubble as in not economically justified presently, but I think money spent on AI research is reasonable.
I have a noob question. How can developed economies where populations levels have plateaued continue to be expected to post positive GDPs (and therefore add net new goods and services) yoy?
Homes as assets should pass on, higher cost services of today would be replaced by lower cost which only temporarily would increase units sold(but should eventually plateau because #units/person is not going to change). No component of the GDP will move.
If the fertility rate of developed economies is less than 2.1 then there should be wealth and asset accumulation among the younger people over time. The demand for newer goods and services from the people who get richer is inelastic: most goods and service's prices dont matter to the rich they buy them any ways, and it continues to keep becoming more inelastic.
So within like a several years the demand should just collapse as wealth accumulates a lot and people work less and become more price insensitive. Immigration is set to remain low to developed nations for the next 3-5 years.
This seems to be quite evident in Japan and EU already(though in the EU if you adjust the productivity for work hours the GDP becomes same as America's).
So why do people assume developed countries would even buy this much more new stuff. 1.5 trillion$ worth of new things over the next 5 years?
GDP does not require that the goods or services were sold for money. GDP measures the market value adjusted to constant quality of all final goods and services produced. GDP can grow even with fewer workers, fewer hours, fewer buyers, and fewer units sold.
So the way I understood it, productivity, efficiency and quality gains can increase GDP year over year.
Say we are a country of 1 person. If that person can make a car in 2025, but in 2026 manages to make both a car and a house, the GDP has more than doubled. Doesn't matter that nobody paid the money for them.
Another weird thing is, say that 1 person country makes a car in 2025, and in 2026 makes a similarly priced car, if the car is higher quality, it counts as a higher GDP, because they'll measure its value as greater than last year's model, even if it sells for the same price, because the old model would now be worth less.
But this only addresses the supply side. The demand side has to be buying those two cars. The 1 person country car that one or two car would need to be bought by someone else there is no GDP consumption recorded
I have a theory (actually I wanted at some point to write more seriously about it) about how GDP also expands by consuming life itself. If you're burnt out, or don't have time to see friends and are depressed have to pay for therapy, GDP grows. If you don't have time to care for your children so you have to keep them busy with million classes, or paid for caretaking, GDP expands. If you don't have time or confidence to meet people IRL and have to pay for Tinder, GDP expands. If you don't have time to cook (even if you enjoy it) so UberEats all the time, GDP expands. When you replace public services by extractivist and more inefficient private ones (see healthcare, education) GDP often grows. The list goes on and on. The more every aspect of human existence is replace by a transaction in the market place, the more GDP grows. We replace a variety of motivations to do and to be based of human relationships and affection, by cold self-interested exchanges between strangers.
GDP is a fucked up way to look at life. It's go to way to look at whether a country is doing good, but it's consuming our environment and our own sanity in so many ways.
That's one of the reasons why I think that actually limiting the working hours (bringing it down to 30, and eventually to 20 hours a week) should be one of the main agenda points for this coming century. It's important for our environment, but also for our own sanity.
One could argue that given a sufficiently large GDP, one can make the individual choice to earn less and have more time. But that's sadly not how things work, since having more workforce available also devalues work relative to subsistence goods (e.g., you can't afford a roof without a full time job). Also, individualism is such a powerful ideology in a market driven economy. Maximizing individualism itself can help you get ahead in the marketplace, and spreads through society via marketing, private media. At some point, we even stop seeing how to behave differently than to maximize our own profit. We need democratic instruments outside of the marketplace to steer our society in a way that improves our lives, regardless of what that does to GDP.
> How can developed economies where populations levels have plateaued continue to be expected to post positive GDPs (and therefore add net new goods and services) yoy?
Think about the unsatisfied needs and desires most people have. In extremely low income areas, it may be a roof over your head or knowing where your next meal comes from. Moving up a tier, it might be the ability to send your children to education or better clothing. In wealthier areas it might be things like a better car or higher quality food even though you're not in danger of going hungry. For the extremely wealthy it might be more leisure time, art, and new experiences.
When GDP increases, broadly, those are the areas you see expand. Looking at life today in a baseline American household, the things which are mass produced are far more available and affordable than they were a century ago - in the 1930s households spent about 10-12% of their income on clothing.
Sadly, the rate of improvement for non-mass-produced items like college tuition, medical care, and especially housing has ballooned compared to median income, so life doesn't feel inexpensive, certainly, but GDP has a lot of room to grow in a lot of areas.
Why not? It has been changing forever. It was on a pretty consistent upward trend in the US since there was a US, roughly doubling since 1790, but it has started to decrease in the last few years.
Total GDP can keep rising so long as technologist can improve efficiency through robotics, inventions and scientific breakthroughs.
GDP just describes peoples amount of activity. People will always build bigger buildings or monuments (see egypt pyramids, dubai skyscrapers, cambodia angkor wat). These are actaully not inelastic as megaprojects will quickly hit real limits regardless of the amount of capital. (I can always add 10 more floors to the tallest skyscraper, or 10 feet to the longest wall, or 10 facets to the most ornate church)
There has never once in history been a point where people decieded that they where going to stop innovating or producing permanently. This is equivalent with death.
So the people with the most money at some point will decide to build things which they spend all their money on which increases GDP.
Also the actual "number" of gdp if heavily controlled by USA inflation rates. So we should always look at gdp in regards to inflation adjusted dollars to get a clearer picture.
You would just think there has got to be end to it all... technology needing to increase so much, indefinitely, the imagination falters tbh. Its like we have to become space gods by year 2500 or we are going to go bankrupt as species anyway.
Even if the populations and quantities of stuff sold by weight peak, people can still by fancier stuff - Ferraris rather than Fiestas etc. GDP just measures the amount spent basically.
Your entire premise is simply incorrect, through and through. Discard it entirely and begin again.
You have a hundred people, they have a GDP of N. Tomorrow, their productivity doubles because of a technological innovation. Your GDP is 2N.
Prices matter extensively to the rich and poor. The cost of a given compute capacity has gone from "literally the entire United States can't afford it" to "my lightbulb has this much compute because it's cheaper than choosing a dumber processor".
What happens tomorrow if eg ChatGPT 5.1 performance becomes doable for $500 of tech? $50? Swap this for grain harvesting, waste bin collection, etc if you don't like the LLM case.
No, it wouldn't as the whole reason people were giving Openai that 500 dollars is because they thought they could make more than 500 dollars from it.
So now that value is just shifted into the companies that were going to purchase from openai.
It would just hurt the investors who have exposure to openai/anthropic/google/microsoft.
Much of the value of this AI boom is not from the direct model companies but its from companies which use their technology.
Although the government could be stupid and bail out these companies which WOULD hurt all us citizens and the inflation caused by money printing due to that could cause a recession.
Here's what I think would happen if anyone, by tomorrow, could download GPT 5.1 for free and run it performantly on something like a $500 laptop:
* It would stop datacenter- and other related infrastructure construction, making huge investments effectively worthless for companies like Oracle and Amazon, and of course hurt the construction sector.
* It would hurt the companies you mention, plus a many more including NVidia, likely in ways that would lead to large-scale layoffs.
* It would seriously hurt corporate and VC investors and likely make them much less interested in large investments for quite some time, thus affecting other sectors as well.
* It would seriously hurt index funds and pension funds.
A number of years down the line, if LLMs are indeed capable of significantly boosting productivity, I'm sure we'd see a recovery, but when large bubbles suddenly burst there's usually some pretty serious fallout.
Why is that an impossible question? I've moved from investing mostly in the SP500 (which is something like %25 big AI companies) to investing into other indices which are not so AI-heavy -- healthcare, infrastructure, etc. I've also put a lot into TSLS (inverse etf for tesla), because I think it is particularly overvalued. I still have a fairly high risk tolerance, but trying to reduce the amount that I'm tied to AI. I'm sure there are even better ways this could be done, but this is a fairly simple way to do roughly what I want.
Edit: Notably, you don't have to fully stop investing in the SP500. If you were previously 60% weighted towards the SP500, you can still reduce your exposure by changing that to 20%, or something like that.
I have been 'raising cash' by selling stuff in my portfolio that seems high risk and is up. I've learned from the past that I end up feeling physically unwell when the portfolio sinks and stays low for a year, some days 1-5% which make me stop looking. Now I'm reducing my risk exposure - yes, that means not literally 'cash' but either a money market fund and a tax-free bond fund. For years I had a lot of gold miners and a gold mining index which only recently popped, I thought that was a 'hedge' but now I am convinced the only hedge is cash or if you know what you are doing, options. Everything is prone to pops and crashes, from big giant index funds, to gold, to bitcoin, to large caps, all the stuff people said was supposed to be a hedge.
For this round I've bought (well, bought and sold, and now waiting to buy via an order) puts on a semiconductor ETF. Since I own some of the stocks, if it goes sky high still I win, if it crashes I win, and if nothing happens I'll have to sell it 30 days before it expires to avoid the theta decay, which starts to really bite in that last month.
No one knows anything - which is the environment I'm learning to operate it.
A very conservative investor, say a boglehead or a 'value investor' would tell you to buy in an index fund and not look, unless you know for sure what you are doing and have done insane research on a company and it's priced low. They would say, buy VTI or VOO if you don't need the money within 5 years, and stop looking at it. Oh, and DCA into it versus all at once.
It’s probably most important to take a long term view of asset allocations, stick to the plan, and be tax efficient. Once you cover that off, the tactical positions can maybe help a touch.
Diversification is the bedrock of portfolio management, meaning owning a range of assets that collectively perform acceptably under a range of scenarios. But it’s generally not sexy, not something you catch individuals bragging about - avoiding permanent loss of capital.
Think about what risk you’re willing to take, in the context of your job/career prospects, current investments etc.
These are things for you to decide. Wouldn’t trust anyone who says buy x without knowing your individual circumstances.
What I can say is there are consistent patterns for many successful investors, and the media will tend to focus on the outliers / lottery winners, which by definition are difficult to emulate / replicate. Be wary of survivorship bias and the narrative fallacy.
If you look at most stock markets over a very long timeline, on average it goes up because the economy is growing. So a very broad portfolio of stocks or ETFs.
Gold is only for very rich people if you want to have something in the case the whole economy goes to the bin.
A friend of mine who is decidedly not rich put her money in gold (via an ETF) two or three years ago and has done very well. She wasn't even trying to do well, just have something that wouldn't get eaten by inflation or crash in a hurry.
During covid you probably wanted a 70/30 split so you didn't get eaten by inflation, but I think 60/40 is probably a safer bet for the average person right now. I'd consider pulling back to 50/50 over the next 6 months to a year so you can loss harvest and skip the capital gains. I'd also consider moving money you don't expect to need anytime in the next decade into real estate.
I'm not really an investor btw, this is just my intuition. I'm curious what others here are doing.
The indices may be inflated but in this environment, treasury bonds guarantee losses (because of explicit inflationary policy. Just look at the real rate of return.) Besides, a 60/40 allocation has not made sense since the early 80s. Morgan Stanley e.g. advised a 60/20/20 with gold: https://www.reuters.com/markets/wealth/morgan-stanley-cio-fa...
> I'm not really an investor btw, this is just my intuition
Don't give advice when you don't know the subject. For a start, there are more than 2 assets. For some, certain corporate bonds look nice right now. Although indices are overpriced and carry risk due to the bubble, much of the wider market is extremely cheap. Developed and undeveloped foreign markets have also been outperforming the US. And commodities...
It’s pretty unpopular, but a properly designed whole life insurance policy is a good stable, growing asset. You can borrow against it at reasonable rates, and it makes a good rainy day fund and risk buffer. You can generally annuitize them later if you wish, providing a stress reduction in retirement which is correlated with longer life, iirc, though that could simply be due to selection bias.
Yeah, not sure how valuable lots of real estate is if there aren't any high-paying jobs nearby, in part due to AI.
And this isn't even about the total number of high-paying jobs. Even having too much income concentration (fewer, but higher paying jobs) will mean that there's less demand at the margin. To put it another way, if the job growth in say, silicon valley, starts to reverse because of AI, there will still be newcomers, but not enough to buy out the available housing at an ever increasing price.
If the price trend ever reverses and holds that way long enough to seem like a new normal, I suspect the price will suddenly correct downwards. Everyone holding on to real estate as an investment will have a great reason to sell once it becomes a depreciating asset. If it goes on long enough, people will be underwater on housing and start walking away.
The price trend is already somewhat flattened, which reduces FOMO. Why buy now when AI is uncertain and the price seems pretty flat?
Those people still need to rent. So as long as the rental income covers the mortgage, you’re ahead of the game and someone is paying an asset down for you.
I'm not sure it's sensible but I have a lot of cash at the moment. I think there'll be a bit of a crash and then it'll be good to invest in stocks bitcoin or whatever. Non crazy might be income generating boring companies like utilities.
I've always used value-tilted indexes, and am hoping that they will suffer less when the bubble pops. With that and a healthy dose of fixed income (which I chose years ago based on an assumption that the equity portion of the portfolio could drop 50% in a downturn at any time) I'm trying to stick to the plan and not try to time the market. Even though it very much feels like the bubble is near its peak (if not just past it!) I do still believe on some level that market timing is a fool's game, so I'm trying to stay convinced that the steps I've already taken are all one can rationally do.
What sort of fixed income that doesn't require much research? A broad bond fund?
What are the implications of bond prices in this dubious interest rate environment? It seems no one knows what the Fed should or wants to do, including the Fed. And if the economy is on shaky ground, won't that be bad for bonds if companies can default?
At least historically, the research I've seen is that one is better off keeping risk in the equity side of a portfolio, not trying to eek out gains on the fixed income side. So that would suggest sticking with intermediate duration, government bond (funds).
You're not expecting it to earn much, but it should hold its value over the long term. This will reduce the expected return of the portfolio, but the goal is to get volatility to a level you can stomach, allowing you to ride out fluctuations on the equity side. Because even though we can all look at the current situation and say the stock market appears overvalued, we can't know how much higher it will go, when we've hit the top, or once it starts declining, when we've hit the bottom. Even experts do no better than luck would dictate at that game.
Agreed on globally diversified value stocks/funds. Personally I still like to have a fixed income cushion as well, though there are certainly arguments both ways on that. (And on whether to globally diversify the bonds you do hold.)
> And Coreweave, the former crypto miner turned data center service provider, unveiled in its IPO documents earlier this year that it has borrowed so much money that its debt payments represent 25 percent of its revenues.
Wow, what a sentence. It starts out bad and just gets worse.
Ok, so if that prediction is true what actions did you take?
Identify the therefore part of the prediction and enumerate the three highest priority steps.
Have you determined that the stock market will crash and bought positions accordingly? Have you sold all your nvidia stock?
What are the implications in the broader economy and what steps have you taken?
Do you have to? I'll be 45 in a few days, but even at that age, I'm still not expecting to need anything I have invested for at least 20 years, the ratio of the value of the broad overall market then to now is still likely to be a larger multiple than for any other asset class I can expose myself to. If there's really a bubble, maybe that won't be true in 5 years, but I've yet to see a downturn that didn't recover in 20. Even if your predictions are directionally correct eventually, if you're not clairvoyant, trying to time the market is still not a great idea.
I've thought about and asked this question. One potential idea: AI was sold as the "death of SaaS" and hit their share prices accordingly. If you think this is not true (at least in the near to mid-term) you could buy stock at (what you perceive as) a discount. I might look at a company like Constellation (a big vertically integrated portfolio of mature, decent revenue-generating SaaS) as a proxy for a fund focused on this strategy.
This sort of approach is pretty aggressive and definitely contrarian to the general market, but some alternatives to mitigate the coming pain (if you believe it's near) would be shifting into a more liquid position to take advantage of tightening liquidity, i.e. big companies may see their shares tank but will likely still be able to service their debt. In the meantime you'll miss out on a market that still seems pretty strong.
Recognizing a bubble and timing what will happen are two really different things. Will it pop tomorrow or 4 years from now? Will the market go up another 70% before the eventual pop?
What would it take to make money? Double or triple the pricing. Most of these companies messed on pricing initially. ChatGPT Plus/ Claude Plus (or whatever it is called) are $20/month. These should be $80 to $100/month products. Probably even more, $200.
But then, how do you get the users and growth figures that justify the current boom in infrastructure build-up? A high equilibrium might be profitable to a few companies and useful to a few rich customers, but it’s far from justifying current valuations.
For the most part, current AI tools do tasks for people, not their jobs. And for the most part it does rather poor quality, low-depth work (but with superhuman speed and superhuman breadth).
The closest thing I can think of is fiverr type gigs, and I'm still paying someone on fiverr for like, UI assets/design for an extremely simple game with 5 screens, and it's not particularly close.
Only for the most basic of requests. I have interacted with a fair share of AI front loaded customer service chat portals and they are often misleading, sending outright incorrect info (telling us that the dev team would work on it even when they weren't going to) and I almost always just want to talk to an agent. Yes, it's a good first layer to prevent people who haven't even bothered to read any FAQ or informational pages, but it's not doing real customer service work.
This sort of reactionary post is a tiring as the motivating situation. Here's a big difference: aside from all the circular investments the companies at the root of AI have huge earnings and did so before the boom as well.It's convenient to say "Look at all these declines in the past month! The end is nigh!" - maybe expand your chart to 12 months or even 5 years. You can argue against Nvidia but if you do they've made you look foolish 3 pivots in a row. Are companies very expensive right now? Absolutely. Are there a lot of frothy valuations and looming failures? Probably - I think so. Is this just like the Internet bubble? Beyond the generic basics, no.
Ah. Yep. They mention NVDA, MSFT, GOOG, etc. I was thinking of the startup world and OpenAI, Grok, etc. The giant cash machines of FAMOG will probably come out well positioned when the dust settles.
Idk, we seem to be the at the cusp of autonomous driving. Transportation is like ~8% of world's GDP. Payroll is, what, 30% of that? It seems like we can already have the return on all AI investment by just conquering this one application.
Maybe guard your investments, but the big question back then was "is the internet really that good, will it stick around" and the reality is it won argument hands down. The same thing will prove true about AI.
One difference is most of the companies listed are old, long standing, well-performing businesses. Pets.com had potential, Apple, Amazon, and Google had large amounts of revenue even 3 years ago.
The bubble may deflate but every company mentioned will still be standing, whereas in 1999 many of them were basically Ponzi schemes relying on further investor dollars to subsidize losses. All this AI spending will hurt some investors if the bubble pops but just make for a few bad quarters for the big tech cos.
> One difference is most of the companies listed are old, long standing, well-performing businesses. Pets.com had potential, Apple, Amazon, and Google had large amounts of revenue even 3 years ago.
Plenty of old companies spiked in 2000. Companies like Microsoft, Intel, or Cisco. Shovel sellers with a history of decent profits. I mean, the NASDAQ crashed, an it is not all made of start ups. You sound in denial, but there are more similarities than you seem to realise.
Hmm. The comparison makes sense in some ways, but the thing is - as much as I dislike AI, I think AI will stay, whereas the internet bubble seemed to subside without that much left from it, more or less. While I hate most of AI, it does have some use cases - the current hype will eventually subside or collapse, but I think the core use cases will most likely remain.
"All but Alphabet have seen big share declines in the past month. Microsoft is down 12 percent, Amazon is down 14 percent, Meta is down 22 percent, Oracle is down 24 percent"
That fluctuates anyway and profits are made by some - nothing new here.
I think the world needs to detach from the stock markets though. That may not seem realistic right now, but if you look at the current US president and the ties to superrich, we really have a huge problem now. A few parasitize on the masses. That can not be sustained. It is not ethical.
"It will spark a generation of innovations that we can’t yet even imagine."
I am not so sure. So far I successfully avoided AI, including becoming dependent on AI - I am not. So that is good.
I can not really see what "innovation" would make me want to embrace AI. Perhaps I can be forced into it, but right now I am happy avoiding it.
"Because we humans are pretty good at predicting the impact of technology revolutions beyond seven to ten years."
No, we really are not.
"Not only does the AI bubble in 2025 feel like the internet bubble in 1999"
It is still not the same.
I feel the article is falling apart there. It tries too much to compare to the 1999, but it is not the same.
I wonder if we need a new term rather than bubble for these types of short-term over-invested sectors.
The Tulip mania (if it existed) was a bubble. It popped, it's never coming back.
BeanieBabies was a bubble for the same reason. NFT art remains to be seen, but I think it was likely a bubble.
The internet was over-invested and decreased significantly when looking on a timeline of 5-7 years.
But outside of that, if you were to take the valuation of internet companies, it is greater than during the period we call the bubble.
We'll see the same with AI. There will most likely be a drop in valuations when looking at a medium term timescale, but long-term, I expect AI companies will be worth far more than they are today.
When the markets drop on a monthly scale, but then rebound, we don't call that a bubble, so why do we call this sub-decade decline a bubble?
Do you think we should have another term for this?
Same with the housing bubble, yes, it popped and lots of people got hurt, but those who were able to hang-on, I think, ended up ok, and on a moderately decent timeline.
Poor old much maligned tulips. The mania isn't coming back but the tulip industry had endured for nearly four hundred years and Holland produces billions of tulip bulbs each year. Some of the more conservative tulip buyers in mania days probably even did ok out of it.
Revenues are up massively, across the board for the big companies but interestingly enough also for startups like ChatGPT, Loveable, Cursor etc. Which is again totally different than last time where mostly startups were driving the insane valuations.
For the shovels part: Actually Global Crossing tried to sell shovels. But failed since there was no demand and no moat. Now compare that to NVIDIA.
What’s really scary is the money that’s being invested vs the .COM bubble. In the end 90s 100 million was a huge investment. Now it seems every AI company immediately a billion dollar and more valuation. There was a time when this was called “unicorn”.
I grew up in the 90's and remember being a computer nerd BEFORE the .com bubble, and starting my career precisely after it burst. Let me tell you one difference I personally am experiencing that differentiates the two. Note, this is a pro-'we're in a bubble' stance:
In the late 90's and early 2000's, businesses were SALIVATING to get online, individuals were finding new ways to benefit and profit from it, and massive investment was being made to facilitate what was inevitable: an interconnected network of everything. Do you remember faxing things ? Paper mail? People half in the know were pushing the Boomers and older Gen-X types to get on with it and modernize.
Now? Not only are people not CLAMORING for more AI, it's that The Powers that Be are forcing it down our throats. At work, we have mandatory AI training, we have people getting promoted for promoting extremely dubious AI solutions both internally and on our product. I log into ANY web site now, whether I'm shopping for a vacuum cleaner or logging into a vendor website, and I get AI shoved in my face, from from "assistant" I have to interact with before typing "agent" to new features I don't care about.
Is there some truth, some merit? Absolutely. But my red flag I'm trying to raise is this: never did it feel FORCED in the 90's, there was a salivation from individuals to get more online, and a reluctance from institutions like elementary and high schools to get with the program.
Now? Corporations large and small are shoving it down our throats. Why? Well, to justify the crazy spending.
I'm no prophet and the world (and economy) are fundamentally unpredictable. But I'll say this. I'm putting my money where my mouth is, and I've put in an order to buy a 5 digit dollar amount of puts on a big 'AI' type ETF, that'll expire in the Spring. It's already wobbling and if you can't beat them, profit off of them when they stumble.
There's absolutely something wrong in the current moment, even if AI is somehow the future. For one thing, the U.S. economy would probably be in a recession right now if it weren't for this insane AI spending. It's wobbled with the recent Nvidia earnings release, and I think it is going to dip (not crash, but start dipping) soon.
The author wrote:
"None of these companies has proven yet that AI is a good enough business to justify all this spending. But the first four are now each spending $70 billion to $100 billion a year to fund data centers and other capital intensive AI expenses. Oracle is spending roughly $20 billion a year. "
In my opinion it's a theoretical arms race 'just because my competitor might win', and not based on anything certain.
I know this is semantics pedantry, but there has never been an „Internet bubble“, only an „Internet-related stock market bubble“.
It’s not as if there were ten million people using and/or building on the Internet, and then this bubble popped and for some years there were only ten thousand people on the Internet.
And I think the same is true for „AI usage/adoption“.
From what I remember, in past bubbles that burst, everyone thought stocks will go up forever.
So, if people are wondering out loud (with widespread traction) if there's a bubble, there's probably not one. It may be self-fulfilling. The awareness of a potential bubble influences us to be more cautious.
No, there's never been a time where everyone was saying "stocks will go up forever" then. Even in the late 90's, as momentum was accumulating towards what would eventually burst as the dot-com bubble, a Very Central Figure in the whole thing was warning about "irrational exuberance" in the market but didn't have safe tools do anything about it.
For the last most explosive bubbles -- "dot-com" (~2000) and "housing" (~2008) -- you can go back and find all sorts of warnings and critiques and fretting and skepticism in newspapers, academic and industry journals, blogs, television, personal ephemera, etc. And not too far from almost all the examples you find, you'll also find somebody blowing off the skepticism.
The tricky part is that you can more or less find similar dialog in all those sources even when there didn't turn out to be a looming collapse. There's just always people worrying about collapse or hard correction when the markets are running hot and always people blowing them off, and sometimes worriers prove right. Eventually. And sometimes they don't, and the underlying economy either catches up with the market or government policy successfully props it up for long enough that some other concern dominates or some other industry can absorb the over-allocated investment.
Basically, it turns out your heuristic for spotting a bubble faces too much noise and isn't actually all that informative at all.
The fact that I'm getting rate-limited every day even though my company has enterprise-level subscriptions to gemini, chatgpt, etc. tells me this bubble is far from popping. I predict within a year 80%+ of devs will be using LLMs to write most of their code for them.
If it is so certain that will happen this year, why didn't it happen already last year, or the year before that? By most peoples measures the codegen hasn't improved. Some believe the newly released models are even a regression on the old in favor of saving costs.
The tooling for autonomous programming agents are still improving and being iterated upon. Developing integrations with various systems and learning how to most effectively make use of agents is an ongoing and open problem. Adoption has actually been reasonably fast.
This post argues the burst is not only inevitable it's right around the corner - never mind their principal argument being "we're terrible at predicting technology revolutions in the near term".
This seems like a variation on the President's (and my CTO's) process of 1. go on record both supporting and denouncing all positions, but commit to none, 2. wait... 3. spin your previous wisdoms based on the actual outcome. A super-inconsistent & confusing track record is actually a major asset when executing this strategy!
This is another doomer article about AI. The author openly admits to using AI for the article and the irony isn't lost on me. How long will we be hearing about the AI bubble?
People continued to use the internet during and after the very real ".com" bubble burst.
LLMs can be useful tools in the right situations and the valuations for companies involved with them can be wildly and irrationally inflated at the same time.
In the 1990s, there were plenty of closets running websites, BBSs and other internet available resources. As a percentage of total internet users, it was higher than those doing so today. So, money was paid to run portions of the internet at home by ISP subscribers.
Pets.com→Chewy
Webvan → Instacart, DoorDash, Amazon Fresh
Kozmo.com → Postmates, Uber Eats, Gopuff
Boo.com (fashion) → Farfetch, Net-a-Porter, ASOS
Broadcast.com → YouTube, Netflix, Twitch
The dot-com bubble didn’t prove the internet was a fad — it proved the internet was inevitable, but the valuations assumed adoption would happen in 2 years instead of 15–20. To me it feels like the AI inevitability will be much quicker.
> To me it feels like the AI inevitability will be much quicker.
Based on what? We're only seeing linear improvements for increasing spending. There's no new algorithm ideas on the horizon, just more and more hardware, in the hopes that if we throw enough RAM and CPU at the problem, it will suddenly become "AGI."
No one has their eye on power budgets or sustainability or durability of the system. The human brain has such a high degree of energy efficiency that I don't think people understand the realities of competing with it digitally.
The main problem "AI" seems to solve is that humans get bored with certain tasks. The language models obviously don't, but they do hallucinate, and checking for hallucinations is an exceedingly boring task. It's a coffin corner of bad ideas.
>no new algorithm ideas on the horizon
The LLM algorithms seem pretty clunky to me, like a hack designed for text translation that surprised everyone by getting quite smart. The reason the human brain is so much more energy efficient is quite likely better design/algorithms. I was watching some video comparing brain and LLM function and almost tempted to try building something myself (https://youtu.be/3SUqBUGlyh8). I'm sure there are many more competent people looking at similar things.
Everyone says "those autoregressive transformer LLMs are obviously flawed", and then fails to come up with anything that outperforms them.
I'm not too bullish on architectural gains. There are efficiencies to be had, but far closer to "+5% a year" than "+5000% in a single breakthrough".
You can try to build a novel AI architecture, at a small scale. Just be ready. This field will kick your teeth in. ML doesn't like grand ideas and isn't kind to high aspirations.
Physics is obviously incomplete and yet nobody can solve quantum gravity. Being obviously flawed doesn't mean the solution is obvious. That's the whole problem.
I think in this case, people tend to underrate just how capable and flexible the basic LLM architecture is. And, also, underrate how many gains are there in better training vs better architecture.
Not obvious but the brain manages to think in ways LLMs don't really and the design is presumably of fairly finite complexity to be encoded in DNA.
Most people are not ML researchers. Most of the AI industry is not AI researchers. Most of the AI spending is not going to AI researchers.
AI researchers came up with an architectural improvement that made a lot of previously impossible stuff barely possible. Then, industry ran with it. Scaling that particular trick to the limits by throwing as much raw compute and data at it as humanly possible.
You don't need to be an AI expert to know that there are probably more advances to be had and that funding foundational research is the way to get them.
The results from "funding foundational research" are, too, middling at best.
It's not certain if something like JEPA would ever reach production grade AI models.
[dead]
> Based on what?
Internet did not have enough devices to reach people. At the height of 2002 only a fraction of people worldwide had an already expensive computer and an internet to go with it.
I ran a e-commerce startup from 2005-2010. Having access to demand is a thing.
Today everyone has access in their pockets. Go to small city in Africa, India and China, and observe how they use AI. See how perplexity has put AI answers in hundreds of millions of people's hands before Google in a matter of months.
Forgive me for saying — but "Based on what" for comparing accelerated adoption between 2005 and 2025 — is discarding many huge elephants in the room, starting with that small thing you're reading this in your hand with, and the invisible thing that's sending you this comment.
The current LLM models are too fallible and inefficient for everyday use. Energy requirements have become a major concern, bucking the trend of banking exponential computing gains through the efficiency improvements. Until recently, year-on-year global energy demands of data-centers were increasing linearly despite the exponential increase in computing power.
This has changed since the cost of equipment and infrastructure has been close to "free" for the large corporations running a cycle of funny-money investment in a bubble. This has allowed them sell access to the computing and models for just above operational energy costs (to the extent of increasing global energy prices), whilst offering free accounts to harvest data. No small competitor could possibly compete with that model.
The calculation for profitability (useful output for humans in a cost-benefit analysis) of the current setup is broken, and trades on our dreams of the future. Scaling computing forever simply does not work. It will never be profitable without further leaps forward in the technology--either more efficient models or hardware. By this time, the extent to the excessive investment in new data-centers will be clear.
"The dot-com bubble didn't prove the internet was a fad -"
The internet is a medium, an interconnection of autonomous computer networks
A "web" of hyperlinked HTML pages is one use of an internet
However this internet is more than a handful of popular websites incorporated as companies
Perhaps that's why it was called the "dot-com" bubble and not the "internet" bubble
The internet predated dot-coms by two or three decades and wasn't very bubbly - mostly government funded links between academic and military institutions. It was only when commerce got in there in the 90s that things started getting busy and then bubbly.
A lot of those early .com companies could have been profitable. They chose to go for rapid growth instead. Some people here probably remember discussion of users mattering more than revenue.
You see that same pattern with AI now. Products are being provided for free or nearly free, and plenty is being spent on marketing.
Thats because both then and now there’s a perception that it’s a land grab.
That the key to success is developing a brand and user base faster than the next person can.
And that’s where AI makes things so hard. Creating a protective moat.
also people forget that user base can change, yahoo -> gmail, altavista -> google search etc
A lot of companies are shocked to find user loyalty is about as good as employer
The minute Amazon stops being good about refunds or a cheaper delivery app comes out than DoorDash I’m gone. No loyalty.
And both of those are on the verge as they focus on profits over customer experience.
Technology for all the above existed in rudimentary form, faster Internet, faster machines and adoption was missing. But Current bets are assuming AGI. No one knows how soon. To predict that would be foolish.
If technology had been the same for the past 20 years, basically none of these would have existed, or would be even close as large as today. We needed way faster cable and mobile internet, and smartphones. Probably even smaller laptops. It was possible to predict these more or less, however, it was impossible to predict when or whether people start to really utilize the internet. Even now, we needed COVID to have another shift regarding this. The general acceptance of “internet first” kind of worldview maybe would have never happened without forcing us to have.
Inevitability of what, chatbots that mark poisonous mushrooms as edible in every product?
How does this add to the discussion? Is the goal to make HN as toxic as everywhere else online? If you have something to say, say it. Otherwise this performative negativity and cynicism is boring honestly.
What I have to say is in the comment above: chatbots are mediocre at best
How is it toxic to back against the irrational infatuation with barely useful stochastic parrots?
On the RHS, post hype, the second movers could work on the boring, unsexy problems in those domains nobody wanted to solve. And solve them extremely well. Then build a moat around that.
There is also a customer adoption curve of technology that lags far behind the technologist adoption curve. For example video on the Web failed a long time, until it didn't, when Youtube began to succeed. The problem became "boring" to technologists in some ways, but consumers gradually caught up.
Blockchain --> ?
Blockchain the the internet seem to have stabilized. The internet as fast video capable links between computers, blockchain as a tech for speculation, gambling and some criminal stuff. AI has not and is still on the exponential bit of the S curve.
Most people would agree with this, question is just how much faster.
> To me it feels like the AI inevitability will be much quicker.
AI is accelerating "let them eat cake" at rates never seen before in history, so I imagine the violence will follow soon after
Yeah but AI can also generate a picture of any cake you prompt it.
Aside from a belief that the AI adoption will happen very quickly, which maybe that’s your main point, you’re not really disagreeing with the article:
> All this means two things to us: 1)The AI revolution will indeed be one of the biggest technology shifts in history. It will spark a generation of innovations that we can’t yet even imagine. 2) It’s going to take way longer to see those changes than we think it’s going to take right now.
The internet was monumental and valuable, offering instant conveyance of media and data. AI is monumental, allowing instant access to near infinite data; but whether instant access is as valuable as instant conveyance yet alone five times the value, appears to be the question
Honestly anyone who thinks AI has intrinsic value to rival the GDP of nations is a bagholder in denial and I'll be happy to buy your puts.
"instant access to near infinite data" is Google search.
AI if it gets better will be thinking / intellectual work which is a different category.
I remember in the 90s people would say the internet will bring instant conveyance of media including video and people were like yeah right I can just about get one page of crappy text in a minute over the unreliable dial up modem. It's like that now with AI thinking.
To say whether it's a bubble, we need to know the value of the technology.
The value of modern AI seems very high. That nobody knows how high, that they still haven't figured out applications, and that the technology and its tools are still far from refined, is normal for any new technology.
If you add the value of the potential political power gained by controlling AI, then the value to the owners and investors is astronomical. Many of the investors have demonstrated a strong motivation to sacrifice money for political power, for example by supporting nationalism that undermines a global economy that they benefit enormously from. Somewhere, I read someone explaining their investment by saying 'it's the greatest transfer of power in (modern) history'. Also see: https://news.ycombinator.com/item?id=45983700
The implied promise was that these things were going to “revolutionize the workplace” i.e. massively automate middle class office jobs
A couple of years down the road, their useful applications still are summarizing text, transferring style to text, generating code under strict supervision, and generating images that need retouching.
That’s a lot to get out of a tool, but it’s dubious that investors were pouring trillions of dollars into these things thinking of automating away junior software engineers and low end design work.
Edit: I forgot their other niche, that of generating homework and school test answers
You forgot - cheating on job interviews, writing resumes to be repetitive, and adding an annoying flowery tone to non-native English speakers who think AI wrote something for them that isn't AI-obvious.
I wonder how many market inefficiencies this creates. People with worse education, people who cheated their way to a job opening compared to a better candidate, etc. Basically counteracting the productivity gains AI was supposed to bring.
There’s a corollary here. People with worse educations may be able to do much higher quality work.
Many of those things were already at a “good enough” level since GPT-3.5.
There’s probably a good business usecase there for companies wanting to have smoother communication with offshore teams.
Could that be a game-changer? I wouldn’t discount it, but it does sound like something that has to operate at a very low margin and that doesn’t merit a lot more investment.
How many people graduate from a US software engineering degree each year? About 100k? If they (the 100k in the US) earn $100k each in the first year, before gaining the skills to earn more, that's $10 billion a year, every year. If you can capture that market for next 20 years, it's worth $200 billion.
Except… can you capture it? A junior dev is… not exactly someone you want connecting to your business-critical database without supervision, and a real human dev will get better with a predictable rate. Will LLMs get better? The makers are betting on that, but we'll only know after the model releases, and even then after we play with them for a bit to differentiate between the record performance on whichever benchmark and the actual work we want them to do.
Then there's the question of can you really keep an edge for 20 years with investments today: Sometime between 2030-2035, there's likely to be models matching 2025-SOTA performance that run on ${year}'s high-end smartphone.
(Well, unless we all die in WW3 because of Russia getting desperate from its failure to remove Ukraine's sovereignty, or because China has a hot war with Taiwan and/or the USA messing with global consumer electronics supplies, but I don't think those get priced into the market…).
> If you can capture that market for next 20 years, it's worth $200 billion.
that's like 5% of NVIDIA's current market cap. sounds like peanuts when you lay it out like that
Perhaps.
But that's just the USA's software developers in just their first year after graduating. Software devs are 1% of the US job market, the first year after graduation is (66-21=45 years, 1/45 ~= 2%) of a working life, the US is just 4% of the world's population/25% GDP.
For the 1% to matter, there have to be other jobs that LLMs can do as well as a fresh graduate. I don't know, are LLMs like someone the first year out of law school or medical school, or are those schools better than software? Certainly the home robotics' AI are nowhere near ready yet, no plumber, no driver (despite the news about new car AIs), would you trust an Optimus to cut your hair? etc.
For the 2% to matter, depends how seriously you take the projections of improvements. Myself, I do not. Looks like exponential improvements come at exponential costs, and you run out of money to spend for further improvements very quickly.
For the 4% to matter, depends on how fast other economies grow. 4% by population, about 25% by GDP. I believe China is still growing quite fast, likely to continue. Them getting +160% growth, and thus getting 2.6x times the money available to burn on AI tokens, over the next 20 years would be unsurprising.
All in all, I don't think the USA is competent enough at large-scale projects to handle the infrastructure that this kind of AI would need, so I think it's a bubble and will burst before 2030 because of that. China seems to be able to pull off this kind of infrastructure, so may pull ahead after the US does whatever it does.
> For the 1% to matter, there have to be other jobs that LLMs can do as well as a fresh graduate. I don't know, are LLMs like someone the first year out of law school or medical school, or are those schools better than software?
Before looking to medical and law schools, I might look to middle-manager school or salesperson school or bookkeeper school.
I don’t know enough to speculate even beyond those crude guesses, but as I thought about this question, I found it interesting to skim the US’ employment-by-detailed-occupation chart:
https://www.bls.gov/cps/cpsaat11b.htm
That analysis only makes sense if companies value AI tools as much as equivalent human productivity. Hypothetically say you have a company with 100 junior developers. An AI service comes around that doubles the productivity of your junior developers, so you can keep only 50 of them. Would the company pay 5M dollars a year forever for that service?
In my experience, the answer is a resounding no. They’ll nickel and dime some kind of per-seat licensing on a monthly basis that costs less than 100 USD or whatever. So for every 100$ in salary you can automate away, you might get 2$ in subscription payments if you are lucky, at current rates.
The hardware also wears out really fast. And every replacement is more expensive. How long can that party keep going when none of the companies make enough revenue to cover the expenses?
GPUs have massive applications such as Alphafold, CRISPR, Medical Imaging, Meteorology.
The massive planetary investment is not to make more AI chats that summarize text. That's just short sighted.
> Meteorology
It seems like that at first glance. But in reality, GPUs have had extremely slow adoption for real-world operational meteorology applications. Because of the fundamental design and architecture of most NWP systems, it was very difficult to leverage GPUs as compute accelerators; most efforts barely eked out any performance gains once you account for host/device memory transfers. It really wasn't until some groups started to design new weather modeling systems from the ground up that they could architect things in such a way that GPUs made a significant difference.
Obviously AI / ML weather modeling is a different story.
As someone working in a field that has used NLP for quite some time - yeah, I generally agree that those investments are worth their weight in gold... which is unfortunate because before ChatGPT came along they were viewed as niche unprofitable money-sinks. The astronomical investments we've seen lately have been in general models which can be leveraged to outperform some of our older models but had we wanted purely to improve those models there were much more efficient ways to do so.
Hopefully we can retain a lot of this value when the bubble bursts but I just haven't seen any really good success stories of converting these models into businesses. If you try and build as a middleman where you leverage a model to solve someone's problem they can always just go to the model runner and get the same results for cheaper - and the model runners seem (so far - this may change) to be unable to price model usage at a level that actually makes it sustainable.
Those older models running specialized tasks seem to be trucking along just fine for now - but I remain concerned that when the bubble bursts it's going to starve these necessary investments of capital.
> converting these models into businesses.
I think it's pretty clear to all the big operators that they will need to go whole hog into ads and take some of the Google/Meta pie. It's just a matter of time.
> The massive planetary investment is not to make more AI chats that summarize text. That's just short sighted.
Yup. It's to make already rich people richer.
You're missing the point. Those kind of narrow AI applications are not the motivation for the trillions of dollars being poured into AI. Of course AI has a variety of applications many disciplines, as it has for decades. The motivation behind the massive investment in AI is as forgetfulness said, reap the benefits from "revolutionizing the workplace"
Eh, those applications (incl. protein folding) existed for a decade-plus before LLMs came onto the scene, and there was absolutely nothing like the scale of capex that we're seeing right now. It's like literally 100-1000x larger than what GPU hosting providers were spending previously.
That’s copium, as the kids say nowadays. The massive planetary investment is a 100% for AI chats. All those other things are taking the crumbs where they can.
Big business and government aren't buying supercomputer clusters and licensing models to run chats.
The really weird thing is that Big Business actually is buying supercomputer clusters to do just that. I can't really talk to the government side but a lot of businesses' early forays into AI was just slapping a chatbot on their product and hoping it'd attract a lot more business. I also think you'd be surprised how integrated really dumb chatbots are into business communication these days.
I think most smart people are looking seriously at different models to try and improve the accuracy of any existing ML uses they had in their business but the new features built post-ChatGPT tend to often just be fancied up chats.
> just slapping a chatbot on their product
That's happening of course, but that's not really the whole picture. Any org that already invests in R&D is likely considering or already implementing modern AI tech into their existing infrastructure. A big oil or pharmaceutical or materials company likely doesn't care much about chat bots, or any customer-facing tech for that matter.
Actually, big orgs are doing exactly that; slapping a chatbot onto their support ticket backlog. Being really, actually “data driven” is hard, and must happen from the bottom up. So instead there’s chatbots in their frontend and support backend, but the backend doing the actual lifting probably hasn’t changed one bit.
I can talk for the gov. site in my European home country: they too are buying GPUs for chat ...
That's the same mistake made with every new, and eventually successful, technology - we haven't found a valuable application yet, so the technology is not valuable.
Finding the valuable application is often the hardest part. That it hasn't happened yet is meaningless. Some technologies sit on the shelf for decades.
AI seems to have a lot of potential: It may be the most valuable technology ever; it may not provide more value than it does now, or something in between. Nobody actually knows. The challenge of innovation is managing that irreduceable risk. It starts by accepting risk, accepting that you don't know. One wrong way is to deny the risk - denying uncertainty - by either saying it's worthless or that success is guaranteed.
[dead]
> thinking of automating away junior software engineers and low end design work.
And it's really still very arguable imo if it's even doing this
Like you said, it still needs strict supervision. In my opinion it is not a good use of your supervisory time to be babysitting an LLM versus mentoring actual juniors
> In my opinion it is not a good use of your supervisory time to be babysitting an LLM versus mentoring actual juniors
Right. Because at least juniors are supposed to learn and at some point become senior and stop needing this kind of supervision. Also, interacting with people can be more rewarding (or not, depending on the people)…
its helping people be more productive but its not helping firing people which was the wet dream
Don't forget they are great at generating speech to text. That's all, TBH. You summarized everything.
> The implied promise was that these things were going to “revolutionize the workplace” i.e. massively automate middle class office jobs
Promise by who? I think the bet is that it would lay off nearly everyone white collar and let AI take its place.
Ok, ignoring any AGI or massive advances, let's just say an LLM can help the average office worker be 15% more productive... what do you think the economic value of that is?
100 billion a year, with math that’s downright delusional.
So, back of the envelope math, the US GDP is 27.72T USD, 80% (22.18) corresponds to the tertiary sector.
Let’s say that this is a 15% increase over a 10 years period, because YoY a boom like the computer revolution itself looks like 2% increases a year.[1]. This amounts to about 1.5% increases each year.
Let’s just make the huge leap that you can just scale the productivity up of all this just by making typing out reports and emails a faster activity, and summarizing information for which you’re not facing liability if the bot gets it wrong.
Yes, including the nurses, cleaners, truckers, teamsters, all of it.
1.5% of it is a cool 330 billion.
How much of a cut of that productivity increase could AI companies take? 30%? That’s 100 billion in one year there.
So with pie-in-the-sky math, they could break even if their obligations throughout the decade don’t amount to 1 trillion (since those 1.5T in bonds issued this year mature in longer periods)
1. https://www.bls.gov/opub/mlr/2021/article/the-us-productivit...
They'll waste 15% more of the day.
And propaganda.
Not only that, but it’s devaluing the sacred cows of the very same companies that are investing heavily.
Search is dead or dying
Social media is dead or dying
Content creation is dead or dying
If they cant make AI work, then they are left with AI at a level that continues to devalue their core business.
They have no choice. They made a deal with the devil. And the devil means to collect.
This is why I think Apple is lucky their attempt failed so bad. They dodged a bullet. They have an opportunity to guide a lost tech industry through a post AI bubble world.
The value of LLM is reliable high quality translation between all languages. The economic value of this is at least trillions of dollars per year. The cultural and humanitarian value is equally gigantic, even if it can't be measured in dollars and cents.
That is of immense social value - but manual translation services for commercial projects (like application localization) is already dirt cheap to do and automatic casual translation services for consumers would be incredibly difficult to monetize.
I think your perspective might be severely limited. There exists millions of businesses outside of big IT enterprises.
I will freely admit this - but I have led a localization effort on a mobile game and worked on localization for a desktop application. I am also lucky enough to travel abroad quite a bit and am quite familiar with consumer offerings.
So I would fairly limit my experience to consumer and medium-sized business uses - I have no experience with large corporate translation efforts (the largest would probably be Ubisoft or the Mouse-Ears company's gaming divisions if you consider them large) and even the small mobile game company I worked at had a budget in the millions range. It certainly hasn't been a focus of my career but I feel comfortable standing by my statement above.
You're only taking about IT stuff, while tourism is one of the largest global industries.
Not to mention hundreds of other sectors which will benefit from better international communication.
I notice that angry hackers are going nuts with their down votes. Make an argument instead if you have something to say.
Tourism and travel is one of the worlds largest industries, and companies in that sector can improve their revenue with double digits if they can reach and communicate with more customers. That's real impact, which is worth something, while hacker votes are worth nothing.
Alternate viewpoint - The value of modern AI seems very high. That nobody knows how high, that they still haven't figured out applications, and that the technology and its tools are still far from refined, might be a sign that the technology shouldn't be valued this highly at all
> That nobody knows how high, that they still haven't figured out applications, and that the technology and its tools are still far from refined, might be a sign that the technology shouldn't be valued this highly at all
That might be true, but it also might be true that it's tremendously valuable. The conditions you identify exist for every new technology; obviously the outcomes vary greatly.
> The value of modern AI seems very high. That nobody knows how high, that they still haven't figured out applications, and that the technology and its tools are still far from refined, is normal for any new technology.
The same could be said for the internet. But, and I know this will be hard for younger readers to believe, I seem to recall the value proposition of the Internet being more immediately apparent at the time.
From what I've read, people spent a long time looking for the 'killer application'. Google didn't know how to make money off its search engine. Social media didn't exist in any mass form. Internet access on phones, beyond email, didn't exist in any usable way.
Proprietary, walled garden services, with their own dial-up numbers, such as AOL and CompuServe, were seen as the future. Microsoft thought their similar service (MSN?) was the way forward and didn't integrate the Internet into Windows 95. That was after Netscape's browser was released and (relative to the time, I'm sure) very popular.
The value was obvious almost immediately, but it was obscured by flashy get rich quick attempts. Google chose not to make money at first but that was a business decision, not a necessity: people were paying tons of money for ads and referrals years before they were founded.
That’s part of the difference with AI now: the internet immediately generated significant value almost instantly for relatively modest investments, so there were a ton of small and mid-sized companies who saw almost immediate profits from moving catalog sales or support online, switching software distribution from mailing diskettes to downloads, etc. and some people were able to cut into established markets where the existing companies were slow to change.
AI is different in two key ways: the first is the extremely high cost, which prevents a lot of the bottom-up growth we saw on the web, but the second is that it’s not a reliable technology like the web was so you can’t safely use it in the most valuable contexts. In the 90s, some people screwed up form validation badly but most people didn’t. Right now, companies would love to replace things like customer service or sales with chatbots but nobody on the market can make a system which is both safe and useful because things like prompt injection require theoretical breakdowns, not just attention to detail.
> The value was obvious almost immediately, but it was obscured by flashy get rich quick attempts. Google chose not to make money at first but that was a business decision, not a necessity: people were paying tons of money for ads and referrals years before they were founded.
All I can say is, that was the opposite of what I've heard. Did you participate?
Also, Google was many years after Netscape's IPO, seen as the first SV boom. And what about Windows 95, if the Internet's ROI was so obvious?
There's a lot of variety in the quality of writing from that period. At the time, a lot of “analysts” were doing lazy number-go-up stories rather than the harder work of actually looking at business fundamentals—I remember so many stories which were basically trying to rationalize the greater fool theory—and there was also a tendency to lump all of the internet companies together. Amazon.com was an interesting example of the latter phenomenon where people would talk about them being unprofitable because they were just looking at the total numbers but completely missing the distinction between the dotcoms which were losing money on every sale with no way to close the gap and Amazon, which was expanding rapidly but was profitable in each business segment within a couple years (e.g. books were cash-positive circa 1995) and could go from reporting losses to profits any time they wanted simply by slowing their expansion plans.
In Google's case, the key thing was that they were out-competing in what was already an established market. Search companies were making money from ads and sales referrals, so the questions were things like marketshare and efficiency rather than the ability to make revenue at all, whereas Netscape struggled with that once Microsoft used their Windows monopoly to set the price for browsers at $0. They tried to sell server software but that was a different market with limited synergy for browsers and there was a ton of competition even if they'd been better at writing software.
I mention all of that because one thing to remember is that the dotcoms weren't the only game in town. A lot of people focused on them because the IPOs and market valuations were wild by traditional standards but most of the real money and energy was in established businesses moving online since there were clear and immediate benefits from doing so. When the bubble popped, that stuff didn't really change and a ton of people moved from dotcom jobs to traditional companies.
I worked for a web development company at the time and we were turning away business for years because there were so many companies looking to add a web presence to their existing business. It made quite an interesting contrast to our startup clients who tended to be a lot more focused on stock prices rather than business fundamentals. I remember being on a call with pets.com people who were basically openly stating that they were basically frittering time away waiting for the IPO to make them rich, whereas our customers who were in manufacturing, banking, insurance, etc. were a lot more realistic about working within a budget to do something they would actually turn a profit from.
Re: Windows 95, the web drove a lot of its success. That was around the time normal people could see a lot of immediate personal benefit from being online–shopping, recipes, stock trading, research, forums, games, dating, porn—and Win95 was the first version of Windows to ship with built-in TCP/IP and a web browser, not to mention being a lot closer to the competition on usability. For a lot of PC users, especially in business, the web was what sold them on the need to pay more to have a GUI, mouse, etc. and while that was possible with Windows 3.1 it really went mainstream around the time Windows 95 came out (Windows NT was too demanding on hardware for most budgets outside of engineering/development).
Thanks for all the detail from someone who was there.
> completely missing the distinction between the dotcoms which were losing money on every sale with no way to close the gap and Amazon, which was expanding rapidly but was profitable in each business segment within a couple years (e.g. books were cash-positive circa 1995) and could go from reporting losses to profits any time they wanted simply by slowing their expansion plans.
Great insight. And of course Amazon would raise any funds possible to expand into this new frontier.
> Win95 was the first version of Windows to ship with built-in TCP/IP and a web browser, not to mention being a lot closer to the competition on usability.
Per Wikipedia:
"Windows 95 originally shipped without Internet Explorer, and the default network installation did not include TCP/IP .... At the release date of Windows 95, Internet Explorer 1.0 was available,[48] but only in the Plus! add-on pack for Windows 95, which was a separate product. ...
Windows 95 OEM Service Release 1 was the first release of Windows to include Internet Explorer (version 2.0) with the OS."
My point is not to nitpick, but that the Internet's value wasn't obvious to Microsoft at that point.
You can have brilliant technology and still go broke.
The internet was amazingly valuable, but many of the early companies failed. Investing in internet companies was hardly a sure bet.
There was a recent Bezos talk about the fact that (much like the dot-com) we've overspending on infrastructure that'll bubble and implode but then we'll have all this amazing infra for companies to build off of... but the process of that overspend and implosion is essentially a massive debt erasure - a lot of people are currently propping up this market with their capital and the companies they're propping up will collapse and those obligations liquidated - and that will result in massive society-wide pain. We may end up in a better place for the next generation because of this investment - but if you're a retail investor don't expect your 401k accounts to weather that burst gracefully... and, unlike the boomers, this will largely hit Millenials and Gen Z both of whom are currently under massive financial stress.
Arguably, it's exactly this mindset creating the problem. It's not the value of the technology that matters. It's the value of the companies. If one company was the only one that had a particular valuable technology, then that would matter, but otherwise 90% of them likely end up worthless even if the sector itself doesn't.
How that washes out on net we'll have to see. I'm not gonna pretend I know more than anyone else. Just keep in mind that a major difference between now and 2000 is companies stay private a lot longer. An IPO forces you to open the books and sustain public scrutiny of the broad investor class. A still-private startup only needs to convince one funder of their value. That inevitably leads to higher variance and a greater risk of failure. That doesn't mean the whole sector is necessarily a bubble, but if it is, it can be sustained a lot longer without us knowing. A small number of people with $50 billion they need to park somewhere and no other obvious options can keep shitty ideas afloat in a way that wouldn't be possible if they had to be subjected to broader exposure. We like to believe people with $50 billion can't be wrong, but the wisdom of crowds always beats the wisdom of individual genius.
Heck, AI itself taught us that same lesson!
Both are great points.
>still haven't figured out applications
the real answer is that the applications for the military, surveillance, and population control are proven, and the pathways to scale those capacities are clear, so the money will pour in no matter what. the implication is that we had better come up with some more consumer/humanity friendly applications that create comparable value, or that is all we will get.
What are the clear military applications that differ from civilian ones?
The value can be incredibly high... just like the value of the internet was incredibly high.. and it can still be a finanfial bubble. I feel like the people arguing against the bubble perspective are saying look the internet was actually valuable and so too will this.. and that's totally a valid perspective. The bubble is not about whether there's value but about whether or not the market will come down. All of the above can totally happen
> The value of modern AI seems very hig
Very useful, clearly. But valuable? In aggregate it seems clear that it is. But where does that value acrue? It seems to me the value will be thinly spread while the costs are concentrated.
It does not seem possible that any conceivable business can pay for all the announced plans for developing data centres, nor energy available to power them.
If AI systems can be developed to be trust worthy enough to act on their own, none so far (?), then I can see where the value could acrue, but as things stand?
> It seems to me the value will be thinly spread while the costs are concentrated.
Won't the value will be concentrated in the few with the scale to build competitive AIs?
In a sense, costs are spread thinly across the content creators whose work has been stolen. (But of course, it is indeed incredibly concentrated in building datacenters.)
> Won't the value will be concentrated in the few with the scale to build competitive AIs?
No.
Open AI and their ilk can sell tokens, but how much business is there in that? Trillions? If not, they're bankrupt.
Dead man walking. Zombie
The open and free models are very capable and getting better.
If there is a breakthrough beyond transformers and neural nets that means trustworthy AI that can be agentic (does not exist, the "trustworthy" part) then it will be exponentially more valuable. But that was as likely in 2020 as in 2030.
Meanwhile something like 10% of the USA electricity supply is being "committed" to datacenters and no practical plans to supply it
Meanwhile the Chinese are leaping ahead and in a more community orientated open way. The cognitive dissonance is quite loud for that one!
> content creators whose work has been stolen.
The law is an ass.
Current IP laws are too restrictive, but scraping public facing websites is not theft even under those crazy laws
I have no moral objection to the use of Anna's Archive to train models. A good use, if illegal. But I do recogise and condemn the hypocrisy of Meta doing it. Not surprising, just disgusting
The wasting of bandwidth is annoying, so silly. We should be better organized
Calling it theft is buying into the mad bad and sad IP legal thinking that has become the dominant paradigm in the West
In a way, I agree with your last sentence. But there is a difference, and that is power.
As a simplistic hypothesis: The already-powerful and super-wealthy taking all that IP from the small content creators is theft, IMHO. The small content creators who can't afford JSTOR finding some free source for a paper, etc. is democratic, empowering, and enriches our society.
I worry that it is not a valid perspective and that the bubble dynamics are being driven exactly by this sentiment.
We are pricing in the hollywood version of AI we don't have as if it is the internet.
I communicated with people exactly like what I am doing right now in this post on usenet in 1995. Message boards by 1997. The internet bubble wasn't based on some wild virtual reality version of the internet that hadn't been invented yet.
It is a categorically different process at work here. This bubble is far more insane and speculative since we don't have what we are pricing in. We don't even really know what we are pricing in besides some vague notation of an inevitable "AGI".
Yep. It seems somewhat inevitable to have a readjustment at some point. Both Web 2.0 (online retail) and AI are (rightfully, I believe) creating a lot of buzz and stimulating a lot of investment. It's all very new and exciting and there is bound to be a lot of hit and miss activity. Once the dust settles, the misguided bets will become more obvious and there will be a culling, just as we had during the dot com bubble.
An essential point.
Absolutely nothing in your list prevents the AI from being a bubble. The Internet is an absolute marvel of engineering that completely changed the life of the majority of people in the world, and yet it crashed 25 years ago, and crashed hard. Both can be true at the same time.
Also, video games in the 1980s.
> To say whether it's a bubble, we need to know the value of the technology.
Not really. I mean, not only. The value of the web is immense. And yet, the dot-com bubble was indeed a bubble. What matters is the value in the short term compared to the value of the companies in the current context. Even if AI is huge 20 years from now, it can still crash dramatically tomorrow.
I was going to say that. Also the bubble can be in some areas but not in others. I think the proposed data center build out is a bubble as in not economically justified presently, but I think money spent on AI research is reasonable.
I have a noob question. How can developed economies where populations levels have plateaued continue to be expected to post positive GDPs (and therefore add net new goods and services) yoy?
Homes as assets should pass on, higher cost services of today would be replaced by lower cost which only temporarily would increase units sold(but should eventually plateau because #units/person is not going to change). No component of the GDP will move.
If the fertility rate of developed economies is less than 2.1 then there should be wealth and asset accumulation among the younger people over time. The demand for newer goods and services from the people who get richer is inelastic: most goods and service's prices dont matter to the rich they buy them any ways, and it continues to keep becoming more inelastic.
So within like a several years the demand should just collapse as wealth accumulates a lot and people work less and become more price insensitive. Immigration is set to remain low to developed nations for the next 3-5 years.
This seems to be quite evident in Japan and EU already(though in the EU if you adjust the productivity for work hours the GDP becomes same as America's).
So why do people assume developed countries would even buy this much more new stuff. 1.5 trillion$ worth of new things over the next 5 years?
GDP does not require that the goods or services were sold for money. GDP measures the market value adjusted to constant quality of all final goods and services produced. GDP can grow even with fewer workers, fewer hours, fewer buyers, and fewer units sold.
So the way I understood it, productivity, efficiency and quality gains can increase GDP year over year.
Say we are a country of 1 person. If that person can make a car in 2025, but in 2026 manages to make both a car and a house, the GDP has more than doubled. Doesn't matter that nobody paid the money for them.
Another weird thing is, say that 1 person country makes a car in 2025, and in 2026 makes a similarly priced car, if the car is higher quality, it counts as a higher GDP, because they'll measure its value as greater than last year's model, even if it sells for the same price, because the old model would now be worth less.
But this only addresses the supply side. The demand side has to be buying those two cars. The 1 person country car that one or two car would need to be bought by someone else there is no GDP consumption recorded
GDP doesn't measure the consumption side, it's just measuring production.
That's why in my 1 person country example, GDP still goes up, nobody has to buy the cars.
I have a theory (actually I wanted at some point to write more seriously about it) about how GDP also expands by consuming life itself. If you're burnt out, or don't have time to see friends and are depressed have to pay for therapy, GDP grows. If you don't have time to care for your children so you have to keep them busy with million classes, or paid for caretaking, GDP expands. If you don't have time or confidence to meet people IRL and have to pay for Tinder, GDP expands. If you don't have time to cook (even if you enjoy it) so UberEats all the time, GDP expands. When you replace public services by extractivist and more inefficient private ones (see healthcare, education) GDP often grows. The list goes on and on. The more every aspect of human existence is replace by a transaction in the market place, the more GDP grows. We replace a variety of motivations to do and to be based of human relationships and affection, by cold self-interested exchanges between strangers.
GDP is a fucked up way to look at life. It's go to way to look at whether a country is doing good, but it's consuming our environment and our own sanity in so many ways.
That's one of the reasons why I think that actually limiting the working hours (bringing it down to 30, and eventually to 20 hours a week) should be one of the main agenda points for this coming century. It's important for our environment, but also for our own sanity.
One could argue that given a sufficiently large GDP, one can make the individual choice to earn less and have more time. But that's sadly not how things work, since having more workforce available also devalues work relative to subsistence goods (e.g., you can't afford a roof without a full time job). Also, individualism is such a powerful ideology in a market driven economy. Maximizing individualism itself can help you get ahead in the marketplace, and spreads through society via marketing, private media. At some point, we even stop seeing how to behave differently than to maximize our own profit. We need democratic instruments outside of the marketplace to steer our society in a way that improves our lives, regardless of what that does to GDP.
Monkeys invented numbers. And then created a world where numbers are all that matters.
A human life, it's worth what? 1 millions usd? 10million usd? 20 million usd? Is it worth saving lives of GDP goes down?
We're due for a reckoning.
> How can developed economies where populations levels have plateaued continue to be expected to post positive GDPs (and therefore add net new goods and services) yoy?
Think about the unsatisfied needs and desires most people have. In extremely low income areas, it may be a roof over your head or knowing where your next meal comes from. Moving up a tier, it might be the ability to send your children to education or better clothing. In wealthier areas it might be things like a better car or higher quality food even though you're not in danger of going hungry. For the extremely wealthy it might be more leisure time, art, and new experiences.
When GDP increases, broadly, those are the areas you see expand. Looking at life today in a baseline American household, the things which are mass produced are far more available and affordable than they were a century ago - in the 1930s households spent about 10-12% of their income on clothing.
Sadly, the rate of improvement for non-mass-produced items like college tuition, medical care, and especially housing has ballooned compared to median income, so life doesn't feel inexpensive, certainly, but GDP has a lot of room to grow in a lot of areas.
>#units/person is not going to change
Why not? It has been changing forever. It was on a pretty consistent upward trend in the US since there was a US, roughly doubling since 1790, but it has started to decrease in the last few years.
People don't stop wanting things.
Total GDP can keep rising so long as technologist can improve efficiency through robotics, inventions and scientific breakthroughs.
GDP just describes peoples amount of activity. People will always build bigger buildings or monuments (see egypt pyramids, dubai skyscrapers, cambodia angkor wat). These are actaully not inelastic as megaprojects will quickly hit real limits regardless of the amount of capital. (I can always add 10 more floors to the tallest skyscraper, or 10 feet to the longest wall, or 10 facets to the most ornate church)
There has never once in history been a point where people decieded that they where going to stop innovating or producing permanently. This is equivalent with death.
So the people with the most money at some point will decide to build things which they spend all their money on which increases GDP.
Also the actual "number" of gdp if heavily controlled by USA inflation rates. So we should always look at gdp in regards to inflation adjusted dollars to get a clearer picture.
You would just think there has got to be end to it all... technology needing to increase so much, indefinitely, the imagination falters tbh. Its like we have to become space gods by year 2500 or we are going to go bankrupt as species anyway.
Even if the populations and quantities of stuff sold by weight peak, people can still by fancier stuff - Ferraris rather than Fiestas etc. GDP just measures the amount spent basically.
Maybe because the advertising and marketing machine will continue to churn and induldge and shame people into consuming ever more stuff.
Think of GDP as a country’s output.
Your entire premise is simply incorrect, through and through. Discard it entirely and begin again.
You have a hundred people, they have a GDP of N. Tomorrow, their productivity doubles because of a technological innovation. Your GDP is 2N.
Prices matter extensively to the rich and poor. The cost of a given compute capacity has gone from "literally the entire United States can't afford it" to "my lightbulb has this much compute because it's cheaper than choosing a dumber processor".
What happens tomorrow if eg ChatGPT 5.1 performance becomes doable for $500 of tech? $50? Swap this for grain harvesting, waste bin collection, etc if you don't like the LLM case.
> What happens tomorrow if eg ChatGPT 5.1 performance becomes doable for $500 of tech? $50?
The bubble would burst and the US economy would face a recession?
No, it wouldn't as the whole reason people were giving Openai that 500 dollars is because they thought they could make more than 500 dollars from it.
So now that value is just shifted into the companies that were going to purchase from openai.
It would just hurt the investors who have exposure to openai/anthropic/google/microsoft.
Much of the value of this AI boom is not from the direct model companies but its from companies which use their technology.
Although the government could be stupid and bail out these companies which WOULD hurt all us citizens and the inflation caused by money printing due to that could cause a recession.
Here's what I think would happen if anyone, by tomorrow, could download GPT 5.1 for free and run it performantly on something like a $500 laptop:
* It would stop datacenter- and other related infrastructure construction, making huge investments effectively worthless for companies like Oracle and Amazon, and of course hurt the construction sector.
* It would hurt the companies you mention, plus a many more including NVidia, likely in ways that would lead to large-scale layoffs.
* It would seriously hurt corporate and VC investors and likely make them much less interested in large investments for quite some time, thus affecting other sectors as well.
* It would seriously hurt index funds and pension funds.
A number of years down the line, if LLMs are indeed capable of significantly boosting productivity, I'm sure we'd see a recovery, but when large bubbles suddenly burst there's usually some pretty serious fallout.
With hardware getting more performant and LLMs getting more efficient, what's preventing this outcome from being an inevitability?
Sure, the average person has no interest in self hosting, but businesses will at least run the numbers.
There's no guarantee hardware will ever improve to such an extent that running your own ChatGPT 5 is cheap.
I think I'm asking for the impossible here... But is anyone trying to hedge risk in their personal finances?
I'm not a real "investor" (index funds only) but I am feeling more willing to forgo gains to be more risk averse just based on my own neuroses.
Maybe I cash out and buy T-bills? Gold? Bullets? What's the non-crazy person equivalent?
Why is that an impossible question? I've moved from investing mostly in the SP500 (which is something like %25 big AI companies) to investing into other indices which are not so AI-heavy -- healthcare, infrastructure, etc. I've also put a lot into TSLS (inverse etf for tesla), because I think it is particularly overvalued. I still have a fairly high risk tolerance, but trying to reduce the amount that I'm tied to AI. I'm sure there are even better ways this could be done, but this is a fairly simple way to do roughly what I want.
Edit: Notably, you don't have to fully stop investing in the SP500. If you were previously 60% weighted towards the SP500, you can still reduce your exposure by changing that to 20%, or something like that.
I have been 'raising cash' by selling stuff in my portfolio that seems high risk and is up. I've learned from the past that I end up feeling physically unwell when the portfolio sinks and stays low for a year, some days 1-5% which make me stop looking. Now I'm reducing my risk exposure - yes, that means not literally 'cash' but either a money market fund and a tax-free bond fund. For years I had a lot of gold miners and a gold mining index which only recently popped, I thought that was a 'hedge' but now I am convinced the only hedge is cash or if you know what you are doing, options. Everything is prone to pops and crashes, from big giant index funds, to gold, to bitcoin, to large caps, all the stuff people said was supposed to be a hedge.
For this round I've bought (well, bought and sold, and now waiting to buy via an order) puts on a semiconductor ETF. Since I own some of the stocks, if it goes sky high still I win, if it crashes I win, and if nothing happens I'll have to sell it 30 days before it expires to avoid the theta decay, which starts to really bite in that last month.
No one knows anything - which is the environment I'm learning to operate it.
A very conservative investor, say a boglehead or a 'value investor' would tell you to buy in an index fund and not look, unless you know for sure what you are doing and have done insane research on a company and it's priced low. They would say, buy VTI or VOO if you don't need the money within 5 years, and stop looking at it. Oh, and DCA into it versus all at once.
It’s probably most important to take a long term view of asset allocations, stick to the plan, and be tax efficient. Once you cover that off, the tactical positions can maybe help a touch.
Diversification is the bedrock of portfolio management, meaning owning a range of assets that collectively perform acceptably under a range of scenarios. But it’s generally not sexy, not something you catch individuals bragging about - avoiding permanent loss of capital.
Think about what risk you’re willing to take, in the context of your job/career prospects, current investments etc.
These are things for you to decide. Wouldn’t trust anyone who says buy x without knowing your individual circumstances.
What I can say is there are consistent patterns for many successful investors, and the media will tend to focus on the outliers / lottery winners, which by definition are difficult to emulate / replicate. Be wary of survivorship bias and the narrative fallacy.
If you look at most stock markets over a very long timeline, on average it goes up because the economy is growing. So a very broad portfolio of stocks or ETFs.
Gold is only for very rich people if you want to have something in the case the whole economy goes to the bin.
Bullets if you are paranoid.
A friend of mine who is decidedly not rich put her money in gold (via an ETF) two or three years ago and has done very well. She wasn't even trying to do well, just have something that wouldn't get eaten by inflation or crash in a hurry.
During covid you probably wanted a 70/30 split so you didn't get eaten by inflation, but I think 60/40 is probably a safer bet for the average person right now. I'd consider pulling back to 50/50 over the next 6 months to a year so you can loss harvest and skip the capital gains. I'd also consider moving money you don't expect to need anytime in the next decade into real estate.
I'm not really an investor btw, this is just my intuition. I'm curious what others here are doing.
The indices may be inflated but in this environment, treasury bonds guarantee losses (because of explicit inflationary policy. Just look at the real rate of return.) Besides, a 60/40 allocation has not made sense since the early 80s. Morgan Stanley e.g. advised a 60/20/20 with gold: https://www.reuters.com/markets/wealth/morgan-stanley-cio-fa...
> I'm not really an investor btw, this is just my intuition
Don't give advice when you don't know the subject. For a start, there are more than 2 assets. For some, certain corporate bonds look nice right now. Although indices are overpriced and carry risk due to the bubble, much of the wider market is extremely cheap. Developed and undeveloped foreign markets have also been outperforming the US. And commodities...
It’s pretty unpopular, but a properly designed whole life insurance policy is a good stable, growing asset. You can borrow against it at reasonable rates, and it makes a good rainy day fund and risk buffer. You can generally annuitize them later if you wish, providing a stress reduction in retirement which is correlated with longer life, iirc, though that could simply be due to selection bias.
Property. It is the one thing they cannot produce more of.
Real Estate is currently trading at the highest price to earnings ratios ever recorded.
Yeah, not sure how valuable lots of real estate is if there aren't any high-paying jobs nearby, in part due to AI.
And this isn't even about the total number of high-paying jobs. Even having too much income concentration (fewer, but higher paying jobs) will mean that there's less demand at the margin. To put it another way, if the job growth in say, silicon valley, starts to reverse because of AI, there will still be newcomers, but not enough to buy out the available housing at an ever increasing price.
If the price trend ever reverses and holds that way long enough to seem like a new normal, I suspect the price will suddenly correct downwards. Everyone holding on to real estate as an investment will have a great reason to sell once it becomes a depreciating asset. If it goes on long enough, people will be underwater on housing and start walking away.
The price trend is already somewhat flattened, which reduces FOMO. Why buy now when AI is uncertain and the price seems pretty flat?
Those people still need to rent. So as long as the rental income covers the mortgage, you’re ahead of the game and someone is paying an asset down for you.
I don't think the rent will cover most mortgages in California, especially not at higher interest rates.
Depends on equity and interest rate of course.
That's what they said in 2007.
Diversification is the only free lunch on risk management
I'm not sure it's sensible but I have a lot of cash at the moment. I think there'll be a bit of a crash and then it'll be good to invest in stocks bitcoin or whatever. Non crazy might be income generating boring companies like utilities.
Let me know if you find an answer. Same question.
Particularly as it seems contagion is more AB’s more likely. That the implosion won’t be limited to the mag 7.
I've always used value-tilted indexes, and am hoping that they will suffer less when the bubble pops. With that and a healthy dose of fixed income (which I chose years ago based on an assumption that the equity portion of the portfolio could drop 50% in a downturn at any time) I'm trying to stick to the plan and not try to time the market. Even though it very much feels like the bubble is near its peak (if not just past it!) I do still believe on some level that market timing is a fool's game, so I'm trying to stay convinced that the steps I've already taken are all one can rationally do.
What sort of fixed income that doesn't require much research? A broad bond fund?
What are the implications of bond prices in this dubious interest rate environment? It seems no one knows what the Fed should or wants to do, including the Fed. And if the economy is on shaky ground, won't that be bad for bonds if companies can default?
At least historically, the research I've seen is that one is better off keeping risk in the equity side of a portfolio, not trying to eek out gains on the fixed income side. So that would suggest sticking with intermediate duration, government bond (funds).
You're not expecting it to earn much, but it should hold its value over the long term. This will reduce the expected return of the portfolio, but the goal is to get volatility to a level you can stomach, allowing you to ride out fluctuations on the equity side. Because even though we can all look at the current situation and say the stock market appears overvalued, we can't know how much higher it will go, when we've hit the top, or once it starts declining, when we've hit the bottom. Even experts do no better than luck would dictate at that game.
I would stay away from US fixed income due to low spreads, higher than usual inflation and a devaluing currency in forex.
I'd say ex-US international value stocks, especially EU, are a better hedge.
Agreed on globally diversified value stocks/funds. Personally I still like to have a fixed income cushion as well, though there are certainly arguments both ways on that. (And on whether to globally diversify the bonds you do hold.)
[dead]
> And Coreweave, the former crypto miner turned data center service provider, unveiled in its IPO documents earlier this year that it has borrowed so much money that its debt payments represent 25 percent of its revenues.
Wow, what a sentence. It starts out bad and just gets worse.
Ok, so if that prediction is true what actions did you take?
Identify the therefore part of the prediction and enumerate the three highest priority steps.
Have you determined that the stock market will crash and bought positions accordingly? Have you sold all your nvidia stock? What are the implications in the broader economy and what steps have you taken?
Do you have to? I'll be 45 in a few days, but even at that age, I'm still not expecting to need anything I have invested for at least 20 years, the ratio of the value of the broad overall market then to now is still likely to be a larger multiple than for any other asset class I can expose myself to. If there's really a bubble, maybe that won't be true in 5 years, but I've yet to see a downturn that didn't recover in 20. Even if your predictions are directionally correct eventually, if you're not clairvoyant, trying to time the market is still not a great idea.
I've thought about and asked this question. One potential idea: AI was sold as the "death of SaaS" and hit their share prices accordingly. If you think this is not true (at least in the near to mid-term) you could buy stock at (what you perceive as) a discount. I might look at a company like Constellation (a big vertically integrated portfolio of mature, decent revenue-generating SaaS) as a proxy for a fund focused on this strategy.
This sort of approach is pretty aggressive and definitely contrarian to the general market, but some alternatives to mitigate the coming pain (if you believe it's near) would be shifting into a more liquid position to take advantage of tightening liquidity, i.e. big companies may see their shares tank but will likely still be able to service their debt. In the meantime you'll miss out on a market that still seems pretty strong.
Recognizing a bubble and timing what will happen are two really different things. Will it pop tomorrow or 4 years from now? Will the market go up another 70% before the eventual pop?
You even have to time the market twice : when it'll go down so you sell just before and when it'll go up so you don't lose on the gains.
Cop out
I’ve got 15-20 years to retirement so I just watch the market but don’t do anything other than keep buying index funds and rebalancing periodically.
Fundamentally they seem different. The web looked like it had the potential to make lots of money but no one knew exactly how.
AI literally does people's jobs for them. There's not much imagination required.
But currently, nobody's actually making money on AI.
It's also not doing peoples' jobs for them, for the most part. AI's supporters do very loudly proclaim this, though.
What would it take to make money? Double or triple the pricing. Most of these companies messed on pricing initially. ChatGPT Plus/ Claude Plus (or whatever it is called) are $20/month. These should be $80 to $100/month products. Probably even more, $200.
But then, how do you get the users and growth figures that justify the current boom in infrastructure build-up? A high equilibrium might be profitable to a few companies and useful to a few rich customers, but it’s far from justifying current valuations.
For the most part, current AI tools do tasks for people, not their jobs. And for the most part it does rather poor quality, low-depth work (but with superhuman speed and superhuman breadth).
If the task is normally "hire a person to do X" it comes out the same.
People aren't usually hired just for some tasks.
The closest thing I can think of is fiverr type gigs, and I'm still paying someone on fiverr for like, UI assets/design for an extremely simple game with 5 screens, and it's not particularly close.
Which jobs is it doing, exactly?
From what I can tell at this point it's a solution looking for a problem. Incredibly impressive, not so useful
Customer service.
Only for the most basic of requests. I have interacted with a fair share of AI front loaded customer service chat portals and they are often misleading, sending outright incorrect info (telling us that the dev team would work on it even when they weren't going to) and I almost always just want to talk to an agent. Yes, it's a good first layer to prevent people who haven't even bothered to read any FAQ or informational pages, but it's not doing real customer service work.
Is it? https://www.forbes.com/sites/quickerbettertech/2025/05/18/bu...
You cannot trust GenAI to do this job, so no, it's not doing this job
Anything that requires reasoning over data and applying experience, such as technical troubleshooting, is outside of this scope.
This sort of reactionary post is a tiring as the motivating situation. Here's a big difference: aside from all the circular investments the companies at the root of AI have huge earnings and did so before the boom as well.It's convenient to say "Look at all these declines in the past month! The end is nigh!" - maybe expand your chart to 12 months or even 5 years. You can argue against Nvidia but if you do they've made you look foolish 3 pivots in a row. Are companies very expensive right now? Absolutely. Are there a lot of frothy valuations and looming failures? Probably - I think so. Is this just like the Internet bubble? Beyond the generic basics, no.
Most of them have big revenues; not earnings.
that's not true; check the (edit: historical) PE ratios for the big names referenced in this post.
Ah. Yep. They mention NVDA, MSFT, GOOG, etc. I was thinking of the startup world and OpenAI, Grok, etc. The giant cash machines of FAMOG will probably come out well positioned when the dust settles.
I am very grateful how the title of the actual article linked saved 40 minutes reading the whole thing.
Idk, we seem to be the at the cusp of autonomous driving. Transportation is like ~8% of world's GDP. Payroll is, what, 30% of that? It seems like we can already have the return on all AI investment by just conquering this one application.
Transportation AI does not remotely require any of the heavy investments obligated to run LLMs.
Maybe guard your investments, but the big question back then was "is the internet really that good, will it stick around" and the reality is it won argument hands down. The same thing will prove true about AI.
One difference is most of the companies listed are old, long standing, well-performing businesses. Pets.com had potential, Apple, Amazon, and Google had large amounts of revenue even 3 years ago.
The bubble may deflate but every company mentioned will still be standing, whereas in 1999 many of them were basically Ponzi schemes relying on further investor dollars to subsidize losses. All this AI spending will hurt some investors if the bubble pops but just make for a few bad quarters for the big tech cos.
> One difference is most of the companies listed are old, long standing, well-performing businesses. Pets.com had potential, Apple, Amazon, and Google had large amounts of revenue even 3 years ago.
Plenty of old companies spiked in 2000. Companies like Microsoft, Intel, or Cisco. Shovel sellers with a history of decent profits. I mean, the NASDAQ crashed, an it is not all made of start ups. You sound in denial, but there are more similarities than you seem to realise.
Hmm. The comparison makes sense in some ways, but the thing is - as much as I dislike AI, I think AI will stay, whereas the internet bubble seemed to subside without that much left from it, more or less. While I hate most of AI, it does have some use cases - the current hype will eventually subside or collapse, but I think the core use cases will most likely remain.
"All but Alphabet have seen big share declines in the past month. Microsoft is down 12 percent, Amazon is down 14 percent, Meta is down 22 percent, Oracle is down 24 percent"
That fluctuates anyway and profits are made by some - nothing new here.
I think the world needs to detach from the stock markets though. That may not seem realistic right now, but if you look at the current US president and the ties to superrich, we really have a huge problem now. A few parasitize on the masses. That can not be sustained. It is not ethical.
"It will spark a generation of innovations that we can’t yet even imagine."
I am not so sure. So far I successfully avoided AI, including becoming dependent on AI - I am not. So that is good.
I can not really see what "innovation" would make me want to embrace AI. Perhaps I can be forced into it, but right now I am happy avoiding it.
"Because we humans are pretty good at predicting the impact of technology revolutions beyond seven to ten years."
No, we really are not.
"Not only does the AI bubble in 2025 feel like the internet bubble in 1999"
It is still not the same.
I feel the article is falling apart there. It tries too much to compare to the 1999, but it is not the same.
I wonder if we need a new term rather than bubble for these types of short-term over-invested sectors.
The Tulip mania (if it existed) was a bubble. It popped, it's never coming back. BeanieBabies was a bubble for the same reason. NFT art remains to be seen, but I think it was likely a bubble.
The internet was over-invested and decreased significantly when looking on a timeline of 5-7 years.
But outside of that, if you were to take the valuation of internet companies, it is greater than during the period we call the bubble.
We'll see the same with AI. There will most likely be a drop in valuations when looking at a medium term timescale, but long-term, I expect AI companies will be worth far more than they are today.
When the markets drop on a monthly scale, but then rebound, we don't call that a bubble, so why do we call this sub-decade decline a bubble?
Do you think we should have another term for this?
Same with the housing bubble, yes, it popped and lots of people got hurt, but those who were able to hang-on, I think, ended up ok, and on a moderately decent timeline.
>... Tulip mania ... never coming back.
Poor old much maligned tulips. The mania isn't coming back but the tulip industry had endured for nearly four hundred years and Holland produces billions of tulip bulbs each year. Some of the more conservative tulip buyers in mania days probably even did ok out of it.
No, it doesn’t. Go and actually read the crazy story about startups like Global Crossing.
Crazy amount of funding, little to no revenue, no competitive moat, no demand.
Compare that to today, and you see only the crazy amount of funding part is the same.
How are the revenues these days? Let’s not count Nvidia since they’re the ones selling the shovels. How are the moats?
Revenues are up massively, across the board for the big companies but interestingly enough also for startups like ChatGPT, Loveable, Cursor etc. Which is again totally different than last time where mostly startups were driving the insane valuations.
For the shovels part: Actually Global Crossing tried to sell shovels. But failed since there was no demand and no moat. Now compare that to NVIDIA.
> Crazy amount of funding, little to no revenue, no competitive moat, no demand.
Doesn't this also describe Thinking Machines Lab?
What is its stock trading at?
It being a general purpose tech is what gives me pause
What’s really scary is the money that’s being invested vs the .COM bubble. In the end 90s 100 million was a huge investment. Now it seems every AI company immediately a billion dollar and more valuation. There was a time when this was called “unicorn”.
I grew up in the 90's and remember being a computer nerd BEFORE the .com bubble, and starting my career precisely after it burst. Let me tell you one difference I personally am experiencing that differentiates the two. Note, this is a pro-'we're in a bubble' stance:
In the late 90's and early 2000's, businesses were SALIVATING to get online, individuals were finding new ways to benefit and profit from it, and massive investment was being made to facilitate what was inevitable: an interconnected network of everything. Do you remember faxing things ? Paper mail? People half in the know were pushing the Boomers and older Gen-X types to get on with it and modernize.
Now? Not only are people not CLAMORING for more AI, it's that The Powers that Be are forcing it down our throats. At work, we have mandatory AI training, we have people getting promoted for promoting extremely dubious AI solutions both internally and on our product. I log into ANY web site now, whether I'm shopping for a vacuum cleaner or logging into a vendor website, and I get AI shoved in my face, from from "assistant" I have to interact with before typing "agent" to new features I don't care about.
Is there some truth, some merit? Absolutely. But my red flag I'm trying to raise is this: never did it feel FORCED in the 90's, there was a salivation from individuals to get more online, and a reluctance from institutions like elementary and high schools to get with the program.
Now? Corporations large and small are shoving it down our throats. Why? Well, to justify the crazy spending.
I'm no prophet and the world (and economy) are fundamentally unpredictable. But I'll say this. I'm putting my money where my mouth is, and I've put in an order to buy a 5 digit dollar amount of puts on a big 'AI' type ETF, that'll expire in the Spring. It's already wobbling and if you can't beat them, profit off of them when they stumble.
There's absolutely something wrong in the current moment, even if AI is somehow the future. For one thing, the U.S. economy would probably be in a recession right now if it weren't for this insane AI spending. It's wobbled with the recent Nvidia earnings release, and I think it is going to dip (not crash, but start dipping) soon.
The author wrote:
"None of these companies has proven yet that AI is a good enough business to justify all this spending. But the first four are now each spending $70 billion to $100 billion a year to fund data centers and other capital intensive AI expenses. Oracle is spending roughly $20 billion a year. "
In my opinion it's a theoretical arms race 'just because my competitor might win', and not based on anything certain.
Work keeps forcing it on us. I’m hoping the bubble bursts just so I don’t have to hear about it at work anymore and can actually get some work done.
Same. We even have AI hackathons where your project must use AI in some way. I have a serious case of AI fatigue.
Agreed
I know this is semantics pedantry, but there has never been an „Internet bubble“, only an „Internet-related stock market bubble“.
It’s not as if there were ten million people using and/or building on the Internet, and then this bubble popped and for some years there were only ten thousand people on the Internet.
And I think the same is true for „AI usage/adoption“.
From what I remember, in past bubbles that burst, everyone thought stocks will go up forever.
So, if people are wondering out loud (with widespread traction) if there's a bubble, there's probably not one. It may be self-fulfilling. The awareness of a potential bubble influences us to be more cautious.
No, there's never been a time where everyone was saying "stocks will go up forever" then. Even in the late 90's, as momentum was accumulating towards what would eventually burst as the dot-com bubble, a Very Central Figure in the whole thing was warning about "irrational exuberance" in the market but didn't have safe tools do anything about it.
For the last most explosive bubbles -- "dot-com" (~2000) and "housing" (~2008) -- you can go back and find all sorts of warnings and critiques and fretting and skepticism in newspapers, academic and industry journals, blogs, television, personal ephemera, etc. And not too far from almost all the examples you find, you'll also find somebody blowing off the skepticism.
The tricky part is that you can more or less find similar dialog in all those sources even when there didn't turn out to be a looming collapse. There's just always people worrying about collapse or hard correction when the markets are running hot and always people blowing them off, and sometimes worriers prove right. Eventually. And sometimes they don't, and the underlying economy either catches up with the market or government policy successfully props it up for long enough that some other concern dominates or some other industry can absorb the over-allocated investment.
Basically, it turns out your heuristic for spotting a bubble faces too much noise and isn't actually all that informative at all.
The fact that I'm getting rate-limited every day even though my company has enterprise-level subscriptions to gemini, chatgpt, etc. tells me this bubble is far from popping. I predict within a year 80%+ of devs will be using LLMs to write most of their code for them.
If it is so certain that will happen this year, why didn't it happen already last year, or the year before that? By most peoples measures the codegen hasn't improved. Some believe the newly released models are even a regression on the old in favor of saving costs.
The tooling for autonomous programming agents are still improving and being iterated upon. Developing integrations with various systems and learning how to most effectively make use of agents is an ongoing and open problem. Adoption has actually been reasonably fast.
If they are rate limiting you, it means they already started enshitfying the product before it is even profitable
I interpret it more that they're losing money on it and are trying to limit the bleeding.
Why would they enshittifiy?
Because that's peak extraction for the enshittifier
writing a “bubble” post is new monad tutorial - gotta write one to get some street cred
This post argues the burst is not only inevitable it's right around the corner - never mind their principal argument being "we're terrible at predicting technology revolutions in the near term".
This seems like a variation on the President's (and my CTO's) process of 1. go on record both supporting and denouncing all positions, but commit to none, 2. wait... 3. spin your previous wisdoms based on the actual outcome. A super-inconsistent & confusing track record is actually a major asset when executing this strategy!
There’s a bubble bubble and it will pop in 26.
This is another doomer article about AI. The author openly admits to using AI for the article and the irony isn't lost on me. How long will we be hearing about the AI bubble?
People continued to use the internet during and after the very real ".com" bubble burst.
LLMs can be useful tools in the right situations and the valuations for companies involved with them can be wildly and irrationally inflated at the same time.
[dead]
Yeah except I pay hard cash for multiple AI services and use them all day long.
I paid for my internet and used it all day long.
The money paid was to access the internet not run it, that's a world of difference.
In the 1990s, there were plenty of closets running websites, BBSs and other internet available resources. As a percentage of total internet users, it was higher than those doing so today. So, money was paid to run portions of the internet at home by ISP subscribers.