AI Giants' Enterprise Revenue: 80% from 1% of Customers, Most Kept Alive by VC
OpenAI and Anthropic get 80% of enterprise revenue from 1% of customers, mostly AI companies burning VC-subsidized tokens; only tech pays big, and when VC dries up, trillion-dollar compute bills blow first.
The video won't play here. Listen to the audio instead:
The argument · tap a timestamp to hear it
80% of enterprise revenue rests on 1% of customers
Ramp's corporate spending data shows that about 80% of OpenAI's and Anthropic's enterprise customer revenue comes from the top 1% of customers, and this ratio has not improved in the past three years. Ramp's chief economist Ara Karazian notes that this 1% is heavily skewed toward the tech industry and AI product/service companies, with a concentration he has not seen in any software category he tracks. The host adds a data boundary: it excludes players like Microsoft and large banks, but it is still representative of the overall AI spending landscape. In other words, what really supports the enterprise revenue of the two leading model makers is not a diversified mass market but a handful of extremely heavy spenders.
— Ed ZitronAI money-burners' money comes from VC, not revenue
This spending structure holds largely because AI startups rely on VC funding to sustain their burn. A common play is selling subscriptions but subsidizing tokens: a user paying $20 per month can burn $30, $40, or even $100 worth of tokens, with the difference covered by capital. That means the money OpenAI/Anthropic receives from such customers is contingent on those customers continuing to raise funds. The host admits this is his intuition: the loudest voices on AI Twitter are mostly VC-backed startups, and the tokens they burn internally for coding may far exceed what external customers pay. So-called real demand, at least a large chunk of it, is actually propped up by 'fundraising ability.'
— Ed ZitronOnly tech itself will pay big for AI
Spreading out these numbers, the conclusion is more awkward than most are willing to admit: AI has dominated every media outlet and every boardroom for three consecutive years, almost every company has tried it, and many have been approved for large AI budgets, but the ones actually willing to spend big turn out to be only the tech industry itself. Industries outside tech and AI hardly pay premium prices for large models, and there is no sign of reversal. Worse, this handful of top customers is itself unstable—they could easily shift to cheaper self-developed open-source models or eventually switch to on-device models. The host points out that this customer structure is the real underlying exposure for OpenAI and Anthropic.
— Ed ZitronCloud giants' fortunes rest on two model companies
The risk does not stop at the two labs themselves. OpenAI and Anthropic have compute procurement commitments exceeding $1.1 trillion, and they are take-or-pay contracts—regardless of actual usage or earnings, the agreed amounts must be paid. Microsoft, Google, Amazon, Oracle, and CoreWeave's compute revenue is all betting that these two can pay. The host cites figures: OpenAI contributes about 70% of Microsoft's FY26 AI revenue; UBS estimates OpenAI and Anthropic will account for 48% of Google Cloud's total revenue next year, corresponding to about $84–100 billion; Barclays also gives forecasts of about $40 billion for AWS and at least $50 billion for Azure in 2027. And these two companies are still losing billions of dollars annually.
— Ed ZitronCompute bills hit a reset wall in 2027
Why hasn't the bubble burst yet? Because the mortgage-like structure hasn't reached its peak repayment period. Such compute contracts typically involve a small upfront payment, data center capacity is not yet online, and the real payment obligations start next month and increase month by month. The host cites an analogy from a Groundbreaker article: it's like the low teaser rate on a mortgage expiring—the rate is extremely low at signing, and after reset, the monthly payment skyrockets to unsustainable levels. Borrowers sign assuming housing prices will keep rising and they can refinance or sell the house, but when everyone tries to run at once, no one can. The massive compute commitments of OpenAI and Anthropic are the subprime mortgages of the AI bubble. The flashy deals give cloud providers huge revenue backlogs, but the ability to pay depends on the belief in continued growth and the constant inflow of VC money.
— Ed ZitronNVIDIA's card sales rely on customer debt
The top of the chain is equally dangerous. Hyperscalers and neoclouds buy more GPUs because of demand from OpenAI and Anthropic; even those building data centers to sell to third parties only reference the revenue backlogs on the books of companies like CoreWeave and Nebius—and these companies' customers are mainly OpenAI, Anthropic, or the cloud giants behind them. More glaringly, NVIDIA's customers can no longer buy GPUs with operating cash flow alone; every purchase depends on the continued availability of debt. Broadcom's latest earnings also show that Anthropic and OpenAI will become its top two customers. The host closes this layer with irony: 'Everything is normal, everything is fine, no one needs to panic.'
— Ed ZitronBloomberg pops a new mega-deal, no one worries
The host points out that sell-side analysts and financial media will reassure you: the smartest people and the most powerful companies don't burn money for no reason, AI demand is real, and skeptics are just cherry-picking data. But just as his monologue is about to end, a Bloomberg terminal pops up with news: Crusoe and Jane Street sign a cloud computing contract worth about $13 billion. The irony is that Jane Street is itself both a major customer of CoreWeave and a shareholder in CoreWeave—these roles intertwined are a sample of the concentration risk he has been talking about throughout the episode. He judges this to be a catastrophic misallocation of capital and will be the biggest collective miss in news history.
— Ed ZitronIn their own words · checked verbatim
Per data from fintech firm Ramp, 80% of OpenAI and Anthropix Enterprise revenues come from 1% of their customers, a number that hasn't improved over the last three years.
Ed Zitron1:03
Most businesses have been given the green light to spend a bunch of money on AI. And in the end, it seems that the only people the tech industry can get to spend money on AI is the tech industry itself.
Ed Zitron4:08
These are, from what I can tell, take-or-pay agreements where they agree to buy that compute capacity regardless of how much capacity they actually end up using and how much revenue they actually bring in.
Ed Zitron5:09
In other words, OpenAI and Anthropic's massive compute commitments are the subprime mortgages of the AI bubble. They signed big, beautiful deals that helped hyperscalers and neoclouds post massive revenue backlogs under the belief that nothing bad would ever happen.
Ed Zitron7:11
Oh, and NVIDIA's customers are no longer able to buy its GPUs through cash flow alone, so all of those purchases are contingent on their constantly availability of debt.
Ed Zitron8:11
I am literally looking at my fucking Bloomberg terminal, and what just popped up says, Crusoe signs roughly $ 13 billion Jane Street deal for cloud computing. ... Jane Street's a major customer of CoreWeave and an investor in CoreWeave.
Ed Zitron9:12
Figures
| Share of OpenAI/Anthropic enterprise revenue from Top 1% customers | 80%, unchanged in nearly three years | 1:03 |
| Total compute procurement commitments by OpenAI+Anthropic | over $1.1 trillion (take-or-pay) | 4:08 |
| OpenAI's contribution to Microsoft AI revenue (FY26) | about 70% | 5:09 |
| Barclays forecast: OpenAI+Anthropic spending on AWS in 2027 | about $40 billion | 5:09 |
| Barclays forecast: OpenAI+Anthropic spending on Azure in 2027 | at least $50 billion | 5:09 |
| Expected 2027 revenue for three cloud vendors from OpenAI/Anthropic (low-end) | over $174 billion | 5:09 |
| Crusoe-Jane Street cloud computing contract value | about $13 billion | 9:12 |
Glossary
- concentration risk
- Risk that revenue or investment is overly concentrated in a few customers, businesses, or assets, collapsing if they leave.
- take-or-pay agreement
- A long-term procurement contract where the buyer must pay the agreed amount regardless of actual usage.
- neocloud
- A new generation of cloud vendors that rent out GPU compute, such as CoreWeave.
- revenue backlog
- Future revenue already contracted but not yet fulfilled, often used to prove real demand.
How to listen
AI founders, VCs, and enterprise tech buyers: if you want to know how fragile the revenue of model vendors and cloud giants really is, this episode lays out the transmission chain clearly.