Will OpenAI someday be seen as the WeWork of AI, as I have suggested several times, as far as back as late 2023? I still think so, and I think that moment is drawing close. They have, by any reasonable standard, seen rough times of late. Google and Anthropic have each largely caught up; various Chinese companies are closing in. More and more questions are being raised about their financing. (And more and more people are agreeing that AGI won’t arrive this decade).
I think they are toast. Anthropic looks better but they are all holding very small moats and AGI they are not. Better search and language translation.. amazing little probabilistic databases but no real intellect.
The lack of moat was clear early and always made the massive spending questionable. The fact that Altman so clearly has a loose relationship with the truth means that few will cry if they go down. Shoutout to Karen Hao’s work.
On an entirely unrelated note, Open AI just announced that the winner of the Best Ever Peace Prize has been determined. "We asked all our employees and 106% were in agreement," gushed Sam I Am. "Soon to be announced!" (Spoiler alert: the award looks just like the FIFA peace prize except that the four grasping hands each have six fingers.)
I think I read the news yesterday that Alphabet’s projected capital spending in 2026 is around $180 billion. To help support this wave of investment, the company has even issued extremely long-dated debt, including a rare 100-year bond.
The total announced capex of Alphabet, Amazon, Meta, and Microsoft for 2026 is aound $650B, reportedly. These are profitable companies, lots of revenue, Open AI is not, keeping up isn't possible for them.
Siince $650B can't practicably be turned into actual compute in a year, maybe the point of the announcements was to scare off Open AI's investors. If so, could be working.
In what way do Anthropic‘s finances differ from those of OpenAI? Both are years from profitability and rapidly burning money. Like OpenAI, Anthropic announces financing deals at fantastic valuations.
November 18 Microsoft & NVIDIA with $13 billion at $183 billion valuation.
This isn’t the behavior of companies nearing positive cash flow, let alone the more difficult bar of profitability. Their increasing need for cash suggests the opposite. These companies are valued based on the latest press releases and investors’ dreams.
I am using Chinese model for coding (GLM) and I quite like it. Even in the benchmarks (which don't always project real world performance, I know) it's only a few points below Opus, but much cheaper (maybe 10x, depends on the plan).
It's also open weight, so if you have your own server farm (or can rent one), you can run it entirely on your own.
The last but not least thing I like about GLM is that I don't even know their CEO.
The old joke really does apply: “We lose money on every sale, but we make it up in volume.”
As the cost of tokens is greater than the revenue acquired, this is a loss on variable costs alone, and nota case of apportioning the overhead. Zitron was hammering away at this issue for ages, and I don't believe anything has fundamentally changed. OpenAI could reduce its fixed costs to $0, and it would still go bankrupt. This is probably why it is now trying to offset costs with advertising, a futile approach given the cost structure of the "business".
We may be past the "gradually" stage of going bust and entering the "fast" stage. Is RIP OpenAI approaching?
Ed Zitron’s latest post (gated) has a brutal assessment of Anthropic. Excerpt, at the end of the section about it:
“Anthropic (per The Information’s reporting) tells investors it will make $18 billion in revenue in 2026 and $55 billion in 2027 — year-over-year increases of 400% and 305% respectively, and is already raising $25 billion after having just closed a $30bn deal. How does Anthropic pay its bills? Why does outlet after outlet print these fantastical numbers without doing the maths of ‘how does Anthropic actually get all this money?’
“Because even with their ridiculous revenue projections, this company is still burning cash, and when you start to actually do the maths around anything in the AI industry, things become genuinely worrying.”
Sam Altman is a pixie bullshitter. The futuristic nonsense that he spouts makes Elon look sober. Not surprised to see OpenAI is having trouble. It looks like a soufflé.
"Talent will make a beeline for competitors." The "talent" is worthless - AI needs people who understand what words mean. Not people who don’t. “he cried out in public” – is that a public outcry? There is a subtlety in words that some people will never get.
AGI needs people to work on large hard problems that are beyond us, not trivial stuff like the Tower of Hanoi, or an AI overview of a phrase. It needs to shake out the people who want to continue as programmers. Maybe the AGI folk could call themselves Complex Systems Engineers to make clear the difference.
You may be pleased that OpenAI is landing with a thud, but did its competitors learn anything in the process – that LLMs weren’t the way? Don’t think so.
without SoftBank openAI might not have made it through 2025. 7 and 1/2 billion early and then another 22 billion just 2 months ago. That would be an awfully quick reversal. Poor Sam.
AGI is a rich man's distraction to avoid solving actual issues (malaria, poverty, etc, etc.). The fact that OpenAI keeps dangling this to investors is no different from mister Musk repeating that Mars is the future of mankind, until it's not. That said, I think that negative sentiments towards OpenAI or SpaceX, for that matter, are misplaced. These companies employ human beings who are certainly doing their very best to create value, and we owe them respect.
No it isn’t because it’s not likely achievable and it’s a distraction. Ml/dl value to industry has nothing to do with AGI and anyone who touts AGI (and I do mean anyone, including critics of the current crop of snake oil salespeople) is just hyping a scam.
I think they’ll be sold off for parts and the most valuable part is the user base. It’ll have a different name but those accounts aren’t going anywhere. Altman will slither away, maybe a new billionaire, to start up some other bs vaporware.
You haven't addressed the revenue vs cost issue. Apple was moving towards having its AI operate on the device, not in the cloud, pushing the costs onto the user, albeit likely with a far less potent AI, especially "reasoning" versions. I can see Apple might buy some pieces of OpenAI's models, but not all, unless the "re-tooled" means distilling and refining the AI onto narrower tasks and trimming the model size appropriately. Small "expert" models rather than a single, "god-like ASI" model.
AI makes no financial sense right now. It's all future-hoping. This may be Apple's last best chance. We're talking levels of desperation where reasoning goes out the window....the entire industry is basically in a hysterical FEAR OF MISSING OUT phase where business sense takes a back seat to emotion.
FOMO is not a good way to run a business. Altman is generating FOMO to keep the funding process going. As has been pointed out elsewhere, the value of data centers packed with short-lived GPUs is not high, unlike dark fiber for high-speed data transport. The models may be acquired for cents on the dollar, but the frontier models need the data centers to run. These will remain costly to maintain, at least for now. This is all so reminiscent of the 19th-century railway boom in the US that eventually blew up for all but a few companies. You can still buy worthless railway bonds for their artistic and historical value.
I think they are toast. Anthropic looks better but they are all holding very small moats and AGI they are not. Better search and language translation.. amazing little probabilistic databases but no real intellect.
The lack of moat was clear early and always made the massive spending questionable. The fact that Altman so clearly has a loose relationship with the truth means that few will cry if they go down. Shoutout to Karen Hao’s work.
On an entirely unrelated note, Open AI just announced that the winner of the Best Ever Peace Prize has been determined. "We asked all our employees and 106% were in agreement," gushed Sam I Am. "Soon to be announced!" (Spoiler alert: the award looks just like the FIFA peace prize except that the four grasping hands each have six fingers.)
I think I read the news yesterday that Alphabet’s projected capital spending in 2026 is around $180 billion. To help support this wave of investment, the company has even issued extremely long-dated debt, including a rare 100-year bond.
The total announced capex of Alphabet, Amazon, Meta, and Microsoft for 2026 is aound $650B, reportedly. These are profitable companies, lots of revenue, Open AI is not, keeping up isn't possible for them.
Siince $650B can't practicably be turned into actual compute in a year, maybe the point of the announcements was to scare off Open AI's investors. If so, could be working.
That makes sense!
In what way do Anthropic‘s finances differ from those of OpenAI? Both are years from profitability and rapidly burning money. Like OpenAI, Anthropic announces financing deals at fantastic valuations.
November 18 Microsoft & NVIDIA with $13 billion at $183 billion valuation.
https://www.nytimes.com/2025/09/02/technology/anthropic-funding-ai.html
January 7, $30 billion at $350 billion valuation. Closed today.
https://www.nytimes.com/2026/02/12/technology/anthropic-valuation-380-billion-funding.html
This isn’t the behavior of companies nearing positive cash flow, let alone the more difficult bar of profitability. Their increasing need for cash suggests the opposite. These companies are valued based on the latest press releases and investors’ dreams.
What a shame. Maybe Sam Altman can get his bribe back from Trump
Explains all the recent donations from Brockman. Doubling down on bribes in hopes of a bailout.
I am using Chinese model for coding (GLM) and I quite like it. Even in the benchmarks (which don't always project real world performance, I know) it's only a few points below Opus, but much cheaper (maybe 10x, depends on the plan).
It's also open weight, so if you have your own server farm (or can rent one), you can run it entirely on your own.
The last but not least thing I like about GLM is that I don't even know their CEO.
A Ponzi scam collapses when the supply of new investors slows and the supposed profits evaporate. https://davidhsing.substack.com/p/openai-is-a-ponzi-scam-operation
We are bearing that
The old joke really does apply: “We lose money on every sale, but we make it up in volume.”
As the cost of tokens is greater than the revenue acquired, this is a loss on variable costs alone, and nota case of apportioning the overhead. Zitron was hammering away at this issue for ages, and I don't believe anything has fundamentally changed. OpenAI could reduce its fixed costs to $0, and it would still go bankrupt. This is probably why it is now trying to offset costs with advertising, a futile approach given the cost structure of the "business".
We may be past the "gradually" stage of going bust and entering the "fast" stage. Is RIP OpenAI approaching?
About “Anthropic {is} doing just fine”
Ed Zitron’s latest post (gated) has a brutal assessment of Anthropic. Excerpt, at the end of the section about it:
“Anthropic (per The Information’s reporting) tells investors it will make $18 billion in revenue in 2026 and $55 billion in 2027 — year-over-year increases of 400% and 305% respectively, and is already raising $25 billion after having just closed a $30bn deal. How does Anthropic pay its bills? Why does outlet after outlet print these fantastical numbers without doing the maths of ‘how does Anthropic actually get all this money?’
“Because even with their ridiculous revenue projections, this company is still burning cash, and when you start to actually do the maths around anything in the AI industry, things become genuinely worrying.”
Sam Altman is a pixie bullshitter. The futuristic nonsense that he spouts makes Elon look sober. Not surprised to see OpenAI is having trouble. It looks like a soufflé.
"Talent will make a beeline for competitors." The "talent" is worthless - AI needs people who understand what words mean. Not people who don’t. “he cried out in public” – is that a public outcry? There is a subtlety in words that some people will never get.
AGI needs people to work on large hard problems that are beyond us, not trivial stuff like the Tower of Hanoi, or an AI overview of a phrase. It needs to shake out the people who want to continue as programmers. Maybe the AGI folk could call themselves Complex Systems Engineers to make clear the difference.
You may be pleased that OpenAI is landing with a thud, but did its competitors learn anything in the process – that LLMs weren’t the way? Don’t think so.
without SoftBank openAI might not have made it through 2025. 7 and 1/2 billion early and then another 22 billion just 2 months ago. That would be an awfully quick reversal. Poor Sam.
I have no sympathy. And he better NOT get a govt. bailout from the taxpayers.
AGI is a rich man's distraction to avoid solving actual issues (malaria, poverty, etc, etc.). The fact that OpenAI keeps dangling this to investors is no different from mister Musk repeating that Mars is the future of mankind, until it's not. That said, I think that negative sentiments towards OpenAI or SpaceX, for that matter, are misplaced. These companies employ human beings who are certainly doing their very best to create value, and we owe them respect.
Germans have a word for rejoicing about someone else's misery: "Schadenfreude"(https://en.wikipedia.org/wiki/Schadenfreude). It's a very natural feeling, best avoided!
A wider view... with the way this has gone down... with LLMs... is the quest for AGI... really worthwhile.
A trillion bucks on "look at this"!?
No it isn’t because it’s not likely achievable and it’s a distraction. Ml/dl value to industry has nothing to do with AGI and anyone who touts AGI (and I do mean anyone, including critics of the current crop of snake oil salespeople) is just hyping a scam.
I think they’ll be sold off for parts and the most valuable part is the user base. It’ll have a different name but those accounts aren’t going anywhere. Altman will slither away, maybe a new billionaire, to start up some other bs vaporware.
But how would they support that user base if it costs more to service each AI request than the request returns?
AppleGPT......Later re-tooled, re-tuned into something like "Songbird" you heard it here first.
You haven't addressed the revenue vs cost issue. Apple was moving towards having its AI operate on the device, not in the cloud, pushing the costs onto the user, albeit likely with a far less potent AI, especially "reasoning" versions. I can see Apple might buy some pieces of OpenAI's models, but not all, unless the "re-tooled" means distilling and refining the AI onto narrower tasks and trimming the model size appropriately. Small "expert" models rather than a single, "god-like ASI" model.
AI makes no financial sense right now. It's all future-hoping. This may be Apple's last best chance. We're talking levels of desperation where reasoning goes out the window....the entire industry is basically in a hysterical FEAR OF MISSING OUT phase where business sense takes a back seat to emotion.
FOMO is not a good way to run a business. Altman is generating FOMO to keep the funding process going. As has been pointed out elsewhere, the value of data centers packed with short-lived GPUs is not high, unlike dark fiber for high-speed data transport. The models may be acquired for cents on the dollar, but the frontier models need the data centers to run. These will remain costly to maintain, at least for now. This is all so reminiscent of the 19th-century railway boom in the US that eventually blew up for all but a few companies. You can still buy worthless railway bonds for their artistic and historical value.
American government having good sense is an oxymoron at this time.