Same reason we bailed out the auto industry: nvidia and co now comprise 1/4 of our goddamned economy, and are thusly wearing a C4 vest in the middle of our money system and threatening to blow it to bits if we don't pay up.
The most rational economic system strikes again. The rich get richer. Everyone else gets fucked. Socialism for corporations, capitalism for the workers.
Edit: upon further thought, this has less in common with the bailout of the auto industry, and far more in common with the 2008 housing crash, and subsequent bailouts which went to the bankers, of course, while workers lost their homes in droves.
The meat of my point remains unchanged though. I just sometimes forget which once-in-a-lifetime economic collapse is which, side effect of being alive right now.
Yep.
An American is free to pay $80k for a Tesla while the rest of the world buys a BYD for $25k
An American is free to pay $100 for Insulin while the rest of the world pays $20.
So too will an American pay for OpenAI while they start blocking foreign competitors from the market "for security reasons" the more the government gets entangled into the scheme.
There’s a difference between Americans paying $80K and getting an overpriced car in return and paying a ton of money to OpenAI and getting nothing in return.
People say this, but I disagree. It's all one big slippery slope of zero value - the more a company can charge you for fewer features, the more they profit.
The parent is correct to identify that a lack of free market controls are what's destroying us here. You wouldn't have to pay out the nose like this if the fed didn't build and enforce monopolies for fun. But now we're here, after 20 years of Google's AdSense monopoly and Apple's App Store monopoly, throwing stones at OpenAI for ruining the fun.
Feels like we deserve this, everyone ignored the warning signs and pushed us way up the corporate escalation ladder.
Genuine question :
When companies are bailed out by the taxpayer, why can't we then give ownership of the company to the taxpayer? Effectively 'buying' the company to save it, instead of just gifting it money for it to survive.
Is there a reason not to make the taxpayer (or government) the main shareholder after the bailout?
This has precedent in the US, like when the government nationalized failing freight railroads and merged them into Conrail. But after the more recent bank and auto bailouts I wouldn't expect to see this happen again. The shareholders would really prefer to have money thrown at them but also keep their stake.
> But after the more recent bank and auto bailouts I wouldn't expect to see this happen again. The shareholders would really prefer to have money thrown at them but also keep their stake.
The auto bailouts did not feature shareholders having money thrown at them and keeping their stakes (GM and Chrysler shareholders, for instance, were almost completely wiped out in the bailout, with the new GM owned by the UAW and the US and Canadian governments; the new Chrysler was majority owned by Fiat with minority stakes held by an autoworkers pension fund and the US and Canadian governments.)
Bank bailouts were more protective of shareholders because they were mostly government purchase of distressed assets or extensions of credit,
> Genuine question : When companies are bailed out by the taxpayer, why can't we then give ownership of the company to the taxpayer?
tHaT'S sOcIAlIsM
Though ironically for the first time ever, the people shouting that would actually be correct. Kind of.
If you mean in a broad "is this possible" sense though, sure, absolutely. Entities owned in part or in whole by the state are not uncommon, but anytime such things are proposed in the US, the right loses it's fucking mind.
Edit: hit the comment rate wall.
> I see! But still, I don't get in what sense it is more socialist that just having people actually buy the company to save it (instead of just saving it 'for free'). If anything it makes it more capitalist if the taxpayers invest in the bailout, instead of just giving it away!
Because socialism isn't an economic system in American politics, it's a scary word that the Russians and CHYNA are. It's also completely interoperable with communism because our conservative party here has long since abandoned anything resembling reality, and even when they were here with us, they didn't know the difference between the two.
Doing it this way is capitalist because it's American. Doing it the other way is evil because it's socialist/communist, like the Russians/Chinese/North Koreans do with the lot of this rhetoric absolutely drowning in racism and nationalism. Mind you, all those countries have issues, absolutely. I'm just saying a conservative with a gun to their head couldn't actually explain those issues, they're just evil because they're not American. [ insert eagle screech here ]
Honestly the best distillation is: It's Freedom when private citizens run things, and it's Communism when the government does. The fact that the government sometimes has to give rich private citizens a shit ton of money to keep things afloat is not reflected upon.
If you try and analyze it through a lens of what these words actually mean, yeah it makes no goddamn sense at all.
But not really right? It could happen in the market, company A chooses to bailout company B by buying it and investing money to keep it afloat.
Except company A in this case is the government. No? Why is it that when it is the government doing this action, it has to gift the money instead of potentially profiting from it?
Edit : just saw the edit. I see! But still, I don't get in what sense it is more socialist, instead of just saving the companing 'for free', people actually buy (forcefully invest?) in the company to save it. If anything it makes it more capitalist if the taxpayers invest in the bailout, instead of just giving it away!
Its pure propaganda playing on American fear of socialism.
In other very capitalist economies governments did take stakes in banks in return for bailouts. The first British bank that needed one (Northern Rock) was entirely taken over by the government and shareholders just lost their money. The government bought stakes in others. It was still criticised as being too generous to shareholders and management.
You are right. It is a lot of your retirement funding, though. There is a lot of debt involved here too. Crash the companies, enough stakeholders get burnt, money gets sucked out of everything else.
How many Americans even have money for retirement?
It reminds me of 1990s Russia were the smart people didn't get caught up in capitalism casino and just kept tending their vegetable garden like they had done for centuries.
To quote Bob Dylan
"If you ain't got nothing you got nothing to lose"
> Russia were the smart people didn't get caught up in capitalism casino and just kept tending their vegetable garden like they had done for centuries.
Ha. Most people didn't have any money to get into the crony/gangster capitalism. Just peasants with their vegetable gardens. How many American have vegetable gardens?
this while factually correct probably requires age distribution. I did not have a penny saved for retirement until I hit mid-to-late-30's. I am 50 now and can retire comfortably today
It is insofar as when it does great we get nothing, and when it doesn't we get fired. And when it does bad enough, tons of retirees lose a shit ton of money for literally no reason.
> Anyways, OpenAI is not in profit-seeking mode, and there's no economic incentives to do so right now.
You disparaged the article, but then immediately agreed with its main point. The fact that there is no economic incentive for OpenAI to run sustainably is a problem. It means they will happily continue to spend trillions of investor, lender, and (soon) government money, most of which is being burned as waste heat radiating from GPUs, in pursuit of an AGI pipedream.
Yeah. No revenue. Nobody wants to hear about revenue! It's not about how much you make, it is about how much you're worth and who is worth the most? Companies that lose money.
The W is important there. If the DAUs were good, they'd report those. I do generally find that LLMs are a weekly thing at best in a chat bot form (the agentic stuff I use more often).
If I operated a vending machine that spat out a dollar every time somebody pressed a button, I'd have just as many weekly users as I put dollar bills into it. You wouldn't call that a winning business despite the exceptionally high rate of 5-star reviews.
The more appropriate analogy is: you have a vending machine that dispenses a paper slip with your fortune on it everytime somebody presses a button. However you can only get one fortune told per day, but you can pay to get more in bulk.
Turns out people find these fortunes super useful, and many are actually paying real money to get more, and each vending machine is actually making money on this.
But now the vending machine industry has also figured out that bigger, more powerful machines produce tell better fortunes which draws in even more people.
So now the industry is investing heavily to build more, bigger vending machines. However, these machines need tons of expensive parts and power, and oh, we can't slow down because China, and so they are racing like crazy to build more.
Unfortunately, there is effectively only one company making a key part, and there's not enough power for all the machines being built, and so very expensive new infrastructure has to be built to meet the forecasted demand for all the fortunes in the world.
And this requires trillions in funding, which gets very expensive to borrow, and so the US government is being asked to provide loan guarantees, because who better would know what interest payments on trillions of debt look like?
> Weird thing to fixate about when the whole industry actually uses MAUs.
Are you kidding me? Like, both Google and Facebook tracked l7, l28 which is the number of days that a person/account logged in over the last 7 or 28 days. Fundamentally, that's the sign of a useful product, in that people keep coming back.
> 800M WAU is a "losing company"?
Not enough information to be sure. If their business model requires selling dollars for fifty cents, then maybe.
I don't doubt the utility of these products (sometimes), but I do doubt the business model behind OpenAI & Anthropic (the "pure" model providers).
Maybe, maybe not. The stat alone doesn't really dictate the answer - it just gives hope they'll be able to turn a profit one day (hopefully soon) without losing their position vs doubt of the same.
But what percentage of those startups are making such grand promises while taking in such unfathomable amounts of capital? The sheer scale of the investment per AI company, the massive tax incentives being given out by states and debt being taken on by energy companies promising massive increase in loads…I mean this is all without precedent. AI speculation is rapidly driving infrastructural changes at the state level across the country.
We know they can’t all be winners. But the price and ripple per failure (at least it seems to me) will be staggering. Am I off base here?
Not being in profit seeking mode is one thing. Being so far away from profit seeking mode that you need a government bailout is an entirely different thing.
If you show revenue, people will ask "How much?" And it will never be enough, but if you have no revenue, you can say you're pre-revenue. You're a potential pure play. It's not about how much you earn, it's about what you're worth. And who's worth the most? Companies that lose money.
Huh? They have committed to about a trillion dollars in infrastructure build out they will need to pay for. How about instead of begging tax payers to back loans to pay for this, they actually produce some profit and pay for it themselves?
I feel like this question needs to be even louder because in the past they would just mumble something about “job creation,” but that’s undermined by the fact that they also tout AI as “job replacement” to investors and clients alike.
What an amazing financial innovation.
If the government involves itself now it'll actually be much cheaper then if the government involves themselves after they've entangled the entire stock market and financial sector into their bet.
Maybe we can call it the pre-bailout.
They also propose that their services will be so key to some sectors (like pharma) that they'll also seek revenue sharing as part of the companies getting the privilege to use the most cutting edge intelligence. Insane stuff honestly.
All I'm going to say is I'm actually really hoping that China stays competitive in this field. Just like they are delivering EVs for $25k to the world it'd be great for all consumers and companies if they can also deliver 90% of the AI performance for 1/10 of the cost.
American tech giants like OpenAI and Nvidia borrow and spend far beyond what they can realistically earn back for decades. Now, short on cash, they’re framing AI as a national security race — warning that without government backing, China will win.
The real play is to socialize their risk while keeping profits private. When markets call this “innovation,” it’s worth remembering: that’s not capitalism working — that’s capture.
Why should we bail out OAI again? It’s hardly sota anymore. Replacing it with a (more) competitive model made by a trillion dollar company with tons of cash is as easy as replacing an API key
Wait, if I reading the original article (https://news.ycombinator.com/item?id=45830380) correctly, this is not a bailout? It's more that the US government acts as a guarantor for loans from other parties, and only steps in if the debt is defaulted on.
Also this is not a loan for what has already been done so far, this is what will be built out in the future, so the US is getting something in return: huge amounts of infrastructure, most of which would be datacenters, chips, and power. (In an alternate timeline this would be clean power, or at least be built out in Australia, cf https://news.ycombinator.com/item?id=45836104 -- but sadly this is not that timeline.)
Worst case, when the bubble pops, we get a glut of power and compute which (ideally) should reduce overall electricity and the obscenely bloated cloud bills. However, at the rates that demand is growing, this seems unlikely.
So to rephrase: The real problem is that the infra needed to support the forecasted growth is very expensive, and requires trillions in funding, which gets very expensive to borrow. And so the US government is being asked to provide loan guarantees, because who better to appreciate what interest payments on trillions of debt look like?
> Worst case, when the bubble pops, we get a glut of power and compute which (ideally) should reduce overall electricity and the obscenely bloated cloud bills. However, at the rates that demand is growing, this seems unlikely.
You are making it sound like there is no opportunity cost to using money for this instead of, I don't know, welfare and social issues; or even other kinds of infrastructure projects altogether.
Also important to remember that Google is years ahead of most other AI shops in that they're running on custom silicon. This makes their inference (and maybe training) cheaper then almost any other company. People don't realize this when compared to OpenAI/Anthropic where most folks are utilizing NVIDIA GPUs, Google is completely different in that aspect with their custom TPU platform.
> Also important to remember that Google is years ahead of most other AI shops in that they're running on custom silicon.
Not just the chips, Google's entire datacenter setup seems much more mature (e.g. liquid cooling, networking, etc.). I saw some video of new Amazon datacenter (https://www.youtube.com/watch?v=vnGC4YS36gU) and it looks like a bunch of server racks in a warehouse.
Google’s datacenters are excellent, from what I’ve seen in my career. They genuinely had so many amazingly talented SMEs pushing boundaries for decades without executive intervention or deterrence, and that’s paid dividends in the subsequent tenure under Pichai and external shareholders (in that they have “infinite” runway and cash reserves to squander on moonshots before risking the company’s core businesses). That said, nothing lasts forever, and if their foray into LLMs don’t pay off, their shareholders are going to be pissed.
And not just pushing the boundaries, working with the HW vendors to define them, asking for features and design elements that others don't really even see the point of.
Anthropic uses TPUs as well as nvidia. Compiler bugs in the tooling around the platform caused most of their quality issues and customer churn this year, but I think they've since announced a big expansion in use:
Depends what you mean by deep in the hole. Given how liberal Gemini is with free access it hardly seems like moneymaking is the end goal. Amongst AI labs, I've personally found Deepmind to be consistently the least sensationalist and least concerned with marketing. I think they're doing a great job of commoditising their complement.
ChatGPT has replaced Google Search for many people. If ChatGPT can also replace Gmail, Maps, and YouTube (by generating content for each of these), Google is done.
Yes but ChatGPT has not replaced Google Search at a profit. ChatGPT loses revenue per user. Certainly the hope is that at some point they become profitable but the concern is that might not happen.
Why not? I already use LLMs to generate summaries of my emails, search emails, and perform a host of email related operations as do many people. No reason OpenAI couldn't release their own email service that tightly integrates their LLM with email directly.
It cannot be stressed or even succinctly elaborated how far ahead Google is compared to every other organization doing anything remotely like this, anywhere on the planet. Practically speaking there is no "catching up" to Google; you can only hope to capitalize on any mistakes that they make.
Google has a massive ad revenue that’s funding its AI habit. The difference between Google and OpenAI is that of a rich guy buying an expensive car that he can easily afford and a working class guy having a mid life crisis.
If banks were too big to fail back in '08, it makes total sense that the current AI bubble is also too big to fail, given that it is many multiples of the financial crisis.
Not to mention that it is propping up the stock market. Ain't no way the current admin will allow the biggest financial crisis to unfold on their watch.
Let them fail, and take the overhyped tech economy with them I say. After the utter disaster that were the 9/11 Airline bailouts, the 2008 Bank bailouts, and the COVID bailouts, I am sick and fucking tired of tax dollars supporting rich sociopaths who can’t make rational financial decisions beyond the scope of their personal wealth.
I just hope it's very soon. The longer they go on propping up the house of cards, the more it seeps into the broader economy making the inevitable crash deeper and longer.
sama showed a really bad side of himself in that answer to Brad Gerstner. You can tell Gerstner thought Altman was kidding at first, but when it became apparent he was lashing out, the answer became less about the answer and more about why Sam Altman was so injured by the (very fair) question. Which, if you haven't read TFA or seen it, the question was "How are you going to make money given how much you borrowed" but Sam's answer was essentially "How dare you ask me that question, we're happy to take back your shares if you have a problem with the way things are going". That was my read on it anyway.
Corporate leadership in America has a megalomania problem. Billionaires in general have a megalomania problem. OpenAI specifically, apparently, has a megalomania problem.
There is certainly some concern around what is effectively a giant bet but I’m finding Gary’s aggressive doomer takes a little tiresome lately. Just like Sam can’t know for sure that the bet will work, Gary can’t know it won’t. Conviction levels should reflect that
Every startup begins with lots of debt and little revenue and a all in bet that it’ll change. Sam’s company unsurprisingly follow same playbook
I refuse to accept that a publicly-traded company with a $300B market cap and hundreds of millions of users is a "startup." Nothing about a startup is comparable to OpenAI besides the megalomaniac running the show.
> Every startup begins with lots of debt and little revenue and a all in bet that it’ll change. Sam’s company unsurprisingly follow same playbook
This reads like an intentional misunderstanding of the general concerns around this current boom. The arguments aren't about the odds of the bet, they're pointing out that in fundamental ways the bet itself doesn't make sense.
I also wouldn't describe this post as "doomerism", it's basically pointing out that this "bet" only makes sense if the plan is for the US public to ultimately be the ones paying for the risks being taken.
It's not much of a bet really. In the old days if a general got it wrong he would die on the battlefield if he was lucky or get shot before a firing squad.
These AI clowns will still have their mansions come what may.
Almost all of those startups fail. Doomerism is the objectively correct outlook on any given startup in the absence of strong countervailing evidence. This isn’t “glass half full vs glass half empty”, it’s “wild optimism vs sober consideration”.
Your use of the word ‘bet’ is a useful clue here. If I bet on a horse, then neither of us knows if the horse will win or not. That doesn’t mean that optimists and pessimists about my prospects as a professional gambler have equally justifiable positions.
Funding model aside this has little in common with them. The average SaaS isn't in the right spot at the right time for a once in a lifetime general-purpose technology (along with money to execute at scale)
>doesn’t mean [..] equally justifiable positions.
Indeed. And I'm not confident it'll be enough to cover Sam's 1 trillion either but the range of possible outcomes here seem very wide to me and a good chunk of it being positive. Seems entirely plausible to me that the odds are 60% in favour here or whatever. We don't really know, but we're reading an article that seems very certain.
> possible outcomes here seem very wide to me and a good chunk of it being positive.
The sheer scale of this makes it nothing like other startups, so the same high risk / high return reasoning doesn't apply. Worse, the 'AI bet' isn't like a diverse VC portfolio approach. OAI, NVidia, MSFT, Oracle, etc are all incestuously inter-related parties propping each other up with future commitments they won't be able to meet at their current revenue growth rates. They are so over-leveraged their revenue growth rates would have to grow at increasingly implausible multiples to cover. So, you now have to believe a sequence of individually extraordinary events will all occur and do so in less than 5 years.
The financial engineering propping up OAI and it's unprecedented market cap makes it a speculative bet with an astronomically high break even and a fixed near-term expiration date. To compensate for that much risk, any sane VC would want at least a somewhat plausible path to 1000x returns - but at this mega-scale there's a serious question if that's even possible from one company in this time frame. That turns OAI into an incredibly specific prop bet which is increasingly untethered from the long-term ability of AI to become a transformative technology beneficial to humankind or even a sustainable growth industry. Those are still both plausible positive outcomes but OAI returning just 100x to their investors (or even surviving a crash of its own making) is increasingly implausible. A year ago I was trying to come up with a good way to short the bubble but couldn't find a pure play-enough bet. Now it's grown so over-leveraged I'm starting to worry the crash is going to seriously impact the broader economy well beyond tech and equities. So now we all have financial exposure to this crazy prop bet, even though we're not betting.
Every startup says that it’s not like other startups. If the founder of the company’s response to being asked how they’re going to obtain one trillion dollars in revenue is (fluff aside) essentially equivalent to “trust me bro”, then I am very comfortable putting long odds on this happening (unlesss via some kind of government bailout).
And yes, obviously there is a wide range of possible outcomes. However, there is not a uniform probability distribution over them.
I'm just going to copy my comment on a previous post about this topic:
You think that Trump won't demand something in return for this?
I keep telling everyone I know that AI will be enshitified just like every other internet business. Tell me why the incentives will be different this time around. Putting yourself in hock to an aspiring authoritarian is certainly one way to supercharge that process.
What do you think OpenAI's output about Jan 6 will be one year from now if this goes through?
The China "AI" threat? Xi has warned of a bubble, has shut down the Internet during exam time and is in no hurry at all to push this nonsense.
Perhaps the government should loan OpenAI $1 billion so it can donate to a four times larger ballroom, upon which it will receive $100 billion in public money to fund the TikTok Sora videos.
Absolutely not. Why should taxpayers bail out a company that is actively putting people out of jobs?
Maybe they should buy it out and convert it into a non-profit.
You mean a non-profit like it was originally supposed to be?
What, and repeat that cycle every couple of years? ;)
How will taxpayers even pay if they don't have jobs?
With all their WORLDCOIN obviously
Because “China”. Also because the people in charge of that in the government won’t understand the technology, and only listen to the PR.
And the people in government and their cronies are getting a cut
Because it doesn't put Dump and his richies out of jobs just the plebs who have zero control over anything, least of all the spending of taxes
Same reason we bailed out the auto industry: nvidia and co now comprise 1/4 of our goddamned economy, and are thusly wearing a C4 vest in the middle of our money system and threatening to blow it to bits if we don't pay up.
The most rational economic system strikes again. The rich get richer. Everyone else gets fucked. Socialism for corporations, capitalism for the workers.
Edit: upon further thought, this has less in common with the bailout of the auto industry, and far more in common with the 2008 housing crash, and subsequent bailouts which went to the bankers, of course, while workers lost their homes in droves.
The meat of my point remains unchanged though. I just sometimes forget which once-in-a-lifetime economic collapse is which, side effect of being alive right now.
Yep. An American is free to pay $80k for a Tesla while the rest of the world buys a BYD for $25k An American is free to pay $100 for Insulin while the rest of the world pays $20. So too will an American pay for OpenAI while they start blocking foreign competitors from the market "for security reasons" the more the government gets entangled into the scheme.
There’s a difference between Americans paying $80K and getting an overpriced car in return and paying a ton of money to OpenAI and getting nothing in return.
What do you mean, nothing? They have an app that they can ask to read wikipedia for them and turn it into bullet points.
People say this, but I disagree. It's all one big slippery slope of zero value - the more a company can charge you for fewer features, the more they profit.
The parent is correct to identify that a lack of free market controls are what's destroying us here. You wouldn't have to pay out the nose like this if the fed didn't build and enforce monopolies for fun. But now we're here, after 20 years of Google's AdSense monopoly and Apple's App Store monopoly, throwing stones at OpenAI for ruining the fun.
Feels like we deserve this, everyone ignored the warning signs and pushed us way up the corporate escalation ladder.
What if Americans started voting in their own interests.
Either that or hating the minorities, so... yeah.
Genuine question : When companies are bailed out by the taxpayer, why can't we then give ownership of the company to the taxpayer? Effectively 'buying' the company to save it, instead of just gifting it money for it to survive.
Is there a reason not to make the taxpayer (or government) the main shareholder after the bailout?
This has precedent in the US, like when the government nationalized failing freight railroads and merged them into Conrail. But after the more recent bank and auto bailouts I wouldn't expect to see this happen again. The shareholders would really prefer to have money thrown at them but also keep their stake.
> But after the more recent bank and auto bailouts I wouldn't expect to see this happen again. The shareholders would really prefer to have money thrown at them but also keep their stake.
The auto bailouts did not feature shareholders having money thrown at them and keeping their stakes (GM and Chrysler shareholders, for instance, were almost completely wiped out in the bailout, with the new GM owned by the UAW and the US and Canadian governments; the new Chrysler was majority owned by Fiat with minority stakes held by an autoworkers pension fund and the US and Canadian governments.)
Bank bailouts were more protective of shareholders because they were mostly government purchase of distressed assets or extensions of credit,
Thanks for the context. Nuts that this is not seen as a strictly better outcome vs a bailout.
Isn't that what just happened with Intel?
> Genuine question : When companies are bailed out by the taxpayer, why can't we then give ownership of the company to the taxpayer?
tHaT'S sOcIAlIsM
Though ironically for the first time ever, the people shouting that would actually be correct. Kind of.
If you mean in a broad "is this possible" sense though, sure, absolutely. Entities owned in part or in whole by the state are not uncommon, but anytime such things are proposed in the US, the right loses it's fucking mind.
Edit: hit the comment rate wall.
> I see! But still, I don't get in what sense it is more socialist that just having people actually buy the company to save it (instead of just saving it 'for free'). If anything it makes it more capitalist if the taxpayers invest in the bailout, instead of just giving it away!
Because socialism isn't an economic system in American politics, it's a scary word that the Russians and CHYNA are. It's also completely interoperable with communism because our conservative party here has long since abandoned anything resembling reality, and even when they were here with us, they didn't know the difference between the two.
Doing it this way is capitalist because it's American. Doing it the other way is evil because it's socialist/communist, like the Russians/Chinese/North Koreans do with the lot of this rhetoric absolutely drowning in racism and nationalism. Mind you, all those countries have issues, absolutely. I'm just saying a conservative with a gun to their head couldn't actually explain those issues, they're just evil because they're not American. [ insert eagle screech here ]
Honestly the best distillation is: It's Freedom when private citizens run things, and it's Communism when the government does. The fact that the government sometimes has to give rich private citizens a shit ton of money to keep things afloat is not reflected upon.
If you try and analyze it through a lens of what these words actually mean, yeah it makes no goddamn sense at all.
But not really right? It could happen in the market, company A chooses to bailout company B by buying it and investing money to keep it afloat.
Except company A in this case is the government. No? Why is it that when it is the government doing this action, it has to gift the money instead of potentially profiting from it?
Edit : just saw the edit. I see! But still, I don't get in what sense it is more socialist, instead of just saving the companing 'for free', people actually buy (forcefully invest?) in the company to save it. If anything it makes it more capitalist if the taxpayers invest in the bailout, instead of just giving it away!
Its pure propaganda playing on American fear of socialism.
In other very capitalist economies governments did take stakes in banks in return for bailouts. The first British bank that needed one (Northern Rock) was entirely taken over by the government and shareholders just lost their money. The government bought stakes in others. It was still criticised as being too generous to shareholders and management.
That's crazy to me...
The stock market is not the economy
You are right. It is a lot of your retirement funding, though. There is a lot of debt involved here too. Crash the companies, enough stakeholders get burnt, money gets sucked out of everything else.
How many Americans even have money for retirement?
It reminds me of 1990s Russia were the smart people didn't get caught up in capitalism casino and just kept tending their vegetable garden like they had done for centuries.
To quote Bob Dylan
"If you ain't got nothing you got nothing to lose"
> Russia were the smart people didn't get caught up in capitalism casino and just kept tending their vegetable garden like they had done for centuries.
Ha. Most people didn't have any money to get into the crony/gangster capitalism. Just peasants with their vegetable gardens. How many American have vegetable gardens?
About half of Americans have zero retirement savings
Virtually no one under the age of 18 has any. So that's a lot of Americans. Should they have any?
this while factually correct probably requires age distribution. I did not have a penny saved for retirement until I hit mid-to-late-30's. I am 50 now and can retire comfortably today
The median for ages 45-54 is 115,000, for 55-64 185,000
It is insofar as when it does great we get nothing, and when it doesn't we get fired. And when it does bad enough, tons of retirees lose a shit ton of money for literally no reason.
At least auto industry produced tangible things of value (even though america's car culture is itself a problem). AI firms can't even point to that
"Socialism for corporations, capitalism for the workers." ../n this was in my mind but couldn't phrase it. thank you
Here's an entertaining take that flows with this topic: https://syntheticauth.ai/posts/synthetic-auth-the-power-prob...
https://officechai.com/ai/microsoft-has-ai-chips-sitting-unu...
>When reality writes better satire than satirists can invent, perhaps it's time to ask some harder questions.
Oof
Wouldn't be a gary marcus post without congratulating himself for tweeting something that all but confirms a disaster that hasn't happened yet.
Anyways, OpenAI is not in profit-seeking mode, and there's no economic incentives to do so right now.
> Anyways, OpenAI is not in profit-seeking mode, and there's no economic incentives to do so right now.
You disparaged the article, but then immediately agreed with its main point. The fact that there is no economic incentive for OpenAI to run sustainably is a problem. It means they will happily continue to spend trillions of investor, lender, and (soon) government money, most of which is being burned as waste heat radiating from GPUs, in pursuit of an AGI pipedream.
No economic incentives to seek profit?
Yeah there certainly aren’t when you can sucker everyone else into paying for your money losing company and cash out in the secondary market.
Yeah. No revenue. Nobody wants to hear about revenue! It's not about how much you make, it is about how much you're worth and who is worth the most? Companies that lose money.
800M WAU is a "losing company"?
The W is important there. If the DAUs were good, they'd report those. I do generally find that LLMs are a weekly thing at best in a chat bot form (the agentic stuff I use more often).
Weird thing to fixate about when the whole industry actually uses MAUs.
Even if all GPT users only logged in once a week (which I highly highly doubt), the question still stands.
800M WAU is a "losing company"?
If I operated a vending machine that spat out a dollar every time somebody pressed a button, I'd have just as many weekly users as I put dollar bills into it. You wouldn't call that a winning business despite the exceptionally high rate of 5-star reviews.
The more appropriate analogy is: you have a vending machine that dispenses a paper slip with your fortune on it everytime somebody presses a button. However you can only get one fortune told per day, but you can pay to get more in bulk.
Turns out people find these fortunes super useful, and many are actually paying real money to get more, and each vending machine is actually making money on this.
But now the vending machine industry has also figured out that bigger, more powerful machines produce tell better fortunes which draws in even more people.
So now the industry is investing heavily to build more, bigger vending machines. However, these machines need tons of expensive parts and power, and oh, we can't slow down because China, and so they are racing like crazy to build more.
Unfortunately, there is effectively only one company making a key part, and there's not enough power for all the machines being built, and so very expensive new infrastructure has to be built to meet the forecasted demand for all the fortunes in the world.
And this requires trillions in funding, which gets very expensive to borrow, and so the US government is being asked to provide loan guarantees, because who better would know what interest payments on trillions of debt look like?
> Weird thing to fixate about when the whole industry actually uses MAUs.
Are you kidding me? Like, both Google and Facebook tracked l7, l28 which is the number of days that a person/account logged in over the last 7 or 28 days. Fundamentally, that's the sign of a useful product, in that people keep coming back.
> 800M WAU is a "losing company"?
Not enough information to be sure. If their business model requires selling dollars for fifty cents, then maybe.
I don't doubt the utility of these products (sometimes), but I do doubt the business model behind OpenAI & Anthropic (the "pure" model providers).
It is a losing company if they aren't making money.
It's a losing company if they aren't making money and have no feasible path to do so.
Maybe, maybe not. The stat alone doesn't really dictate the answer - it just gives hope they'll be able to turn a profit one day (hopefully soon) without losing their position vs doubt of the same.
“Money losing company” as in the company loses money, not “losing company” like their softball team got eliminated from the tournament
"money-losing"
Almost every Silicon Valley startup goes through a period where they seek growth, not profit and spend their investors money to maintain it.
Yes, at about 1% of this scale. OpenAI's obligations are not something they can just run to daddy VC to pay for; he can't afford it either
Sure, most of then lose millions of dollars, but uh, not trillions
But what percentage of those startups are making such grand promises while taking in such unfathomable amounts of capital? The sheer scale of the investment per AI company, the massive tax incentives being given out by states and debt being taken on by energy companies promising massive increase in loads…I mean this is all without precedent. AI speculation is rapidly driving infrastructural changes at the state level across the country.
We know they can’t all be winners. But the price and ripple per failure (at least it seems to me) will be staggering. Am I off base here?
Not being in profit seeking mode is one thing. Being so far away from profit seeking mode that you need a government bailout is an entirely different thing.
If you show revenue, people will ask "How much?" And it will never be enough, but if you have no revenue, you can say you're pre-revenue. You're a potential pure play. It's not about how much you earn, it's about what you're worth. And who's worth the most? Companies that lose money.
Is that why they just changed to become a for-profit company?
Huh? They have committed to about a trillion dollars in infrastructure build out they will need to pay for. How about instead of begging tax payers to back loans to pay for this, they actually produce some profit and pay for it themselves?
I feel like this question needs to be even louder because in the past they would just mumble something about “job creation,” but that’s undermined by the fact that they also tout AI as “job replacement” to investors and clients alike.
There's no ability for them to do so. They have a pathetic conversion rate and lose 3x as much money as they make.
What an amazing financial innovation. If the government involves itself now it'll actually be much cheaper then if the government involves themselves after they've entangled the entire stock market and financial sector into their bet. Maybe we can call it the pre-bailout.
They also propose that their services will be so key to some sectors (like pharma) that they'll also seek revenue sharing as part of the companies getting the privilege to use the most cutting edge intelligence. Insane stuff honestly.
All I'm going to say is I'm actually really hoping that China stays competitive in this field. Just like they are delivering EVs for $25k to the world it'd be great for all consumers and companies if they can also deliver 90% of the AI performance for 1/10 of the cost.
The buildout is already taxpayer funded: see the big beautiful bill, which seems like it was engineered to make this effectively free.
They can blend leverage, 100% 1st year depreciation with using the hardware itself as a financing asset and dozens more financial engineering steps -
Their actual cost of financing is probably incredibly low already.
I worry that they are thinking of trying to run the company just ahead of debt/lease payments or something, otherwise this is just a distraction
So they want even more?
Depreciation only helps if you have revenue of the same basic scale to offset (to avoid a paper profit).
It looks like they might not get the bailout they're hoping for - https://x.com/DavidSacks/status/1986476840207122440
the "If you owe the bank 1M, you have a problem. If you owe the bank 1B, the bank has a problem" adagium scaled up ?
If you owe the bank $1B, the bank has a problem (they will work with you on repayment terms).
If you owe the bank $1T, the government has a problem (you're too big to fail and will get bailed out).
American tech giants like OpenAI and Nvidia borrow and spend far beyond what they can realistically earn back for decades. Now, short on cash, they’re framing AI as a national security race — warning that without government backing, China will win.
The real play is to socialize their risk while keeping profits private. When markets call this “innovation,” it’s worth remembering: that’s not capitalism working — that’s capture.
Why should we bail out OAI again? It’s hardly sota anymore. Replacing it with a (more) competitive model made by a trillion dollar company with tons of cash is as easy as replacing an API key
There is a difference between a model, a product, and a network.
OAI has a solid handle on the latter two.
800,000,000 users a month last I saw them report. 10% of the world population a month is no joke.
That isn’t to say they are worth it, I am commenting that “the model” isn’t the important part.
There is more to a car than its engine. Even among our people, no one really cares about engines anymore.
If they go out of business tomorrow, those 800M users aren't going to just stop using LLMs. They'll shift to other providers.
>If they go out of business [...]
Hold your breath, then.
I feel like inferred tokens are a commodity and will be subject to commodity pricing pressure.
Which indicates to me that the value add will come from somewhere else, which would seem to be whales and addicts.
which is a real problem if a significant part of being in the population of whale or addict involves AI psychosis
He's defensive but not agitated: https://www.youtube.com/watch?v=WEnBCfzsgyE
It struck me as performative for whatever reason.
Buried lede there is Natella explaining how there isn't enough electric power in the world for AI!!
Wait, if I reading the original article (https://news.ycombinator.com/item?id=45830380) correctly, this is not a bailout? It's more that the US government acts as a guarantor for loans from other parties, and only steps in if the debt is defaulted on.
Also this is not a loan for what has already been done so far, this is what will be built out in the future, so the US is getting something in return: huge amounts of infrastructure, most of which would be datacenters, chips, and power. (In an alternate timeline this would be clean power, or at least be built out in Australia, cf https://news.ycombinator.com/item?id=45836104 -- but sadly this is not that timeline.)
Worst case, when the bubble pops, we get a glut of power and compute which (ideally) should reduce overall electricity and the obscenely bloated cloud bills. However, at the rates that demand is growing, this seems unlikely.
So to rephrase: The real problem is that the infra needed to support the forecasted growth is very expensive, and requires trillions in funding, which gets very expensive to borrow. And so the US government is being asked to provide loan guarantees, because who better to appreciate what interest payments on trillions of debt look like?
> Worst case, when the bubble pops, we get a glut of power and compute which (ideally) should reduce overall electricity and the obscenely bloated cloud bills. However, at the rates that demand is growing, this seems unlikely.
You are making it sound like there is no opportunity cost to using money for this instead of, I don't know, welfare and social issues; or even other kinds of infrastructure projects altogether.
How can I help destroy the company?
Is Google also deep in the hole with Gemini, or is it just OpenAI that can't make ends meet?
Also important to remember that Google is years ahead of most other AI shops in that they're running on custom silicon. This makes their inference (and maybe training) cheaper then almost any other company. People don't realize this when compared to OpenAI/Anthropic where most folks are utilizing NVIDIA GPUs, Google is completely different in that aspect with their custom TPU platform.
> Also important to remember that Google is years ahead of most other AI shops in that they're running on custom silicon.
Not just the chips, Google's entire datacenter setup seems much more mature (e.g. liquid cooling, networking, etc.). I saw some video of new Amazon datacenter (https://www.youtube.com/watch?v=vnGC4YS36gU) and it looks like a bunch of server racks in a warehouse.
Google’s datacenters are excellent, from what I’ve seen in my career. They genuinely had so many amazingly talented SMEs pushing boundaries for decades without executive intervention or deterrence, and that’s paid dividends in the subsequent tenure under Pichai and external shareholders (in that they have “infinite” runway and cash reserves to squander on moonshots before risking the company’s core businesses). That said, nothing lasts forever, and if their foray into LLMs don’t pay off, their shareholders are going to be pissed.
And not just pushing the boundaries, working with the HW vendors to define them, asking for features and design elements that others don't really even see the point of.
Anthropic uses TPUs as well as nvidia. Compiler bugs in the tooling around the platform caused most of their quality issues and customer churn this year, but I think they've since announced a big expansion in use:
https://www.anthropic.com/engineering/a-postmortem-of-three-...
Depends what you mean by deep in the hole. Given how liberal Gemini is with free access it hardly seems like moneymaking is the end goal. Amongst AI labs, I've personally found Deepmind to be consistently the least sensationalist and least concerned with marketing. I think they're doing a great job of commoditising their complement.
https://gwern.net/complement
Google has enough income that they can self fund Gemini at a loss on whatever timespan they want. OpenAI does not.
ChatGPT has replaced Google Search for many people. If ChatGPT can also replace Gmail, Maps, and YouTube (by generating content for each of these), Google is done.
Yes but ChatGPT has not replaced Google Search at a profit. ChatGPT loses revenue per user. Certainly the hope is that at some point they become profitable but the concern is that might not happen.
Replace Gmail? Meaning instead of real emails from my actual contacts, I’ll read AI generated emails?
Why not? I already use LLMs to generate summaries of my emails, search emails, and perform a host of email related operations as do many people. No reason OpenAI couldn't release their own email service that tightly integrates their LLM with email directly.
Google has the revenue to cover their spending and then some.
It cannot be stressed or even succinctly elaborated how far ahead Google is compared to every other organization doing anything remotely like this, anywhere on the planet. Practically speaking there is no "catching up" to Google; you can only hope to capitalize on any mistakes that they make.
Google has a massive ad revenue that’s funding its AI habit. The difference between Google and OpenAI is that of a rich guy buying an expensive car that he can easily afford and a working class guy having a mid life crisis.
Related:
OpenAI asks U.S. for loan guarantees to fund $1T AI expansion
https://news.ycombinator.com/item?id=45830380
If banks were too big to fail back in '08, it makes total sense that the current AI bubble is also too big to fail, given that it is many multiples of the financial crisis.
Not to mention that it is propping up the stock market. Ain't no way the current admin will allow the biggest financial crisis to unfold on their watch.
Let them fail, and take the overhyped tech economy with them I say. After the utter disaster that were the 9/11 Airline bailouts, the 2008 Bank bailouts, and the COVID bailouts, I am sick and fucking tired of tax dollars supporting rich sociopaths who can’t make rational financial decisions beyond the scope of their personal wealth.
Let. It. Burn.
I just hope it's very soon. The longer they go on propping up the house of cards, the more it seeps into the broader economy making the inevitable crash deeper and longer.
Too bad, you elected the guy that is best friends with all the people who will profit from an AI bailout.
What did you expect?
Glad I'm not american!
Mm, I'm not sure if we're saved outside the US from this crash either...
"You" as in Americans. Not me.
sama showed a really bad side of himself in that answer to Brad Gerstner. You can tell Gerstner thought Altman was kidding at first, but when it became apparent he was lashing out, the answer became less about the answer and more about why Sam Altman was so injured by the (very fair) question. Which, if you haven't read TFA or seen it, the question was "How are you going to make money given how much you borrowed" but Sam's answer was essentially "How dare you ask me that question, we're happy to take back your shares if you have a problem with the way things are going". That was my read on it anyway.
Corporate leadership in America has a megalomania problem. Billionaires in general have a megalomania problem. OpenAI specifically, apparently, has a megalomania problem.
Time to sell shares and buy bonds?
If the US government has to find 1T$ somewhere, it isn't very good for US treasuries you know.
Gold having quite a year!
There is certainly some concern around what is effectively a giant bet but I’m finding Gary’s aggressive doomer takes a little tiresome lately. Just like Sam can’t know for sure that the bet will work, Gary can’t know it won’t. Conviction levels should reflect that
Every startup begins with lots of debt and little revenue and a all in bet that it’ll change. Sam’s company unsurprisingly follow same playbook
I refuse to accept that a publicly-traded company with a $300B market cap and hundreds of millions of users is a "startup." Nothing about a startup is comparable to OpenAI besides the megalomaniac running the show.
> Every startup begins with lots of debt and little revenue and a all in bet that it’ll change. Sam’s company unsurprisingly follow same playbook
This reads like an intentional misunderstanding of the general concerns around this current boom. The arguments aren't about the odds of the bet, they're pointing out that in fundamental ways the bet itself doesn't make sense.
I also wouldn't describe this post as "doomerism", it's basically pointing out that this "bet" only makes sense if the plan is for the US public to ultimately be the ones paying for the risks being taken.
It's not much of a bet really. In the old days if a general got it wrong he would die on the battlefield if he was lucky or get shot before a firing squad.
These AI clowns will still have their mansions come what may.
Almost all of those startups fail. Doomerism is the objectively correct outlook on any given startup in the absence of strong countervailing evidence. This isn’t “glass half full vs glass half empty”, it’s “wild optimism vs sober consideration”.
Your use of the word ‘bet’ is a useful clue here. If I bet on a horse, then neither of us knows if the horse will win or not. That doesn’t mean that optimists and pessimists about my prospects as a professional gambler have equally justifiable positions.
>Almost all of those startups fail.
Funding model aside this has little in common with them. The average SaaS isn't in the right spot at the right time for a once in a lifetime general-purpose technology (along with money to execute at scale)
>doesn’t mean [..] equally justifiable positions.
Indeed. And I'm not confident it'll be enough to cover Sam's 1 trillion either but the range of possible outcomes here seem very wide to me and a good chunk of it being positive. Seems entirely plausible to me that the odds are 60% in favour here or whatever. We don't really know, but we're reading an article that seems very certain.
> possible outcomes here seem very wide to me and a good chunk of it being positive.
The sheer scale of this makes it nothing like other startups, so the same high risk / high return reasoning doesn't apply. Worse, the 'AI bet' isn't like a diverse VC portfolio approach. OAI, NVidia, MSFT, Oracle, etc are all incestuously inter-related parties propping each other up with future commitments they won't be able to meet at their current revenue growth rates. They are so over-leveraged their revenue growth rates would have to grow at increasingly implausible multiples to cover. So, you now have to believe a sequence of individually extraordinary events will all occur and do so in less than 5 years.
The financial engineering propping up OAI and it's unprecedented market cap makes it a speculative bet with an astronomically high break even and a fixed near-term expiration date. To compensate for that much risk, any sane VC would want at least a somewhat plausible path to 1000x returns - but at this mega-scale there's a serious question if that's even possible from one company in this time frame. That turns OAI into an incredibly specific prop bet which is increasingly untethered from the long-term ability of AI to become a transformative technology beneficial to humankind or even a sustainable growth industry. Those are still both plausible positive outcomes but OAI returning just 100x to their investors (or even surviving a crash of its own making) is increasingly implausible. A year ago I was trying to come up with a good way to short the bubble but couldn't find a pure play-enough bet. Now it's grown so over-leveraged I'm starting to worry the crash is going to seriously impact the broader economy well beyond tech and equities. So now we all have financial exposure to this crazy prop bet, even though we're not betting.
Every startup says that it’s not like other startups. If the founder of the company’s response to being asked how they’re going to obtain one trillion dollars in revenue is (fluff aside) essentially equivalent to “trust me bro”, then I am very comfortable putting long odds on this happening (unlesss via some kind of government bailout).
And yes, obviously there is a wide range of possible outcomes. However, there is not a uniform probability distribution over them.
I'm just going to copy my comment on a previous post about this topic:
You think that Trump won't demand something in return for this?
I keep telling everyone I know that AI will be enshitified just like every other internet business. Tell me why the incentives will be different this time around. Putting yourself in hock to an aspiring authoritarian is certainly one way to supercharge that process.
What do you think OpenAI's output about Jan 6 will be one year from now if this goes through?
LLM technology is the endgame and holy grail of advertising and propaganda.
The China "AI" threat? Xi has warned of a bubble, has shut down the Internet during exam time and is in no hurry at all to push this nonsense.
Perhaps the government should loan OpenAI $1 billion so it can donate to a four times larger ballroom, upon which it will receive $100 billion in public money to fund the TikTok Sora videos.