The AI bubble has a name – and it’s OpenAI
For three years, Sam Altman has worn the badge of greatest pitchman in the history of tech.
He sold the world a vision of artificial general intelligence – a machine that thinks, reasons and creates – and sold investors the bill to build it.
The man raised more capital in a single funding round than any other firm has ever raised going public.
He bagged $122bn at a valuation of $852bn, making the ChatGPT-maker the most highly valued private firm in Silicon Valley history, promised investors a trillion dollar IPO. And for a while, the market believed every word.
But this week, shares in the companies tied to OpenAI’s fate were pummelled, following a Wall Street Journal report saying the firm had missed internal targets for both revenue and user growth.
Softbank, the Tokyo-based investor that has ploughed over $34bn into OpenAI since September 2024, fell by nearly 12 per cent.
Cloud infrastructure startup Coreweave fell by 7.1 per cent, Oracle dropped 6.5, Nvidia and Broadcom between three and four per cent.
The communal tumble showed that if OpenAI sneezes, the entire network it has spent the last three years assembling catches a cold.
“If revenue growth doesn’t re-accelerate, those contracts become the most expensive fixed-cost bet in technology history”, said Harrison Rolfes, senior analyst at Pitchbook.
The Anthropic problem
This isn’t a story of consumers falling out of love with AI, and demand hasn’t yet showed signs of slowing.
But what has happened is something more dangerous for OpenAI specifically: a decisive shift in where that demand is going.
The WSJ’s report said OpenAI missed multiple monthly revenue targets after ceding ground to Anthropic in coding and enterprise – precisely the highest-value segments in the entire industry.
ChatGPT’s share of generative AI web traffic has slumped from 86.7 per cent a year ago to 64.5 per cent in January 2026, while Google’s Gemini surged from 5.7 per cent to 21.5 per cent.
Anthropic’s Claude Code reached more than $2.5bn in annualised revenue, a figure that would have seemed impossible 18 months ago.
OpenAI is still huge, having topped $25bn in annualised revenue and currently housing nearly 900 million weekly users, but it failed to reach its internal target of one billion weekly active users by the end of 2025 – a bad sign for a firm built almost entirely on the audacity of its own targets.
OpenAI’s circular deals
OpenAI sits at the centre of one of the most unusual financial arrangements in corporate history.
Nvidia invests $100bn in OpenAI; OpenAI buys Nvidia chips. Oracle spends $300bn building OpenAI data centres; OpenAI pays Oracle for cloud capacity.
SoftBank raises $100bn to fund OpenAI infrastructure. CoreWeave borrows heavily, builds data centres for OpenAI, receives $350m in OpenAI equity.
The arrangement has been likened to the dot-com era’s circular advertising deals, whilst defenders have dubbed it ‘legitimate’ vendor financing – companies funding a customer they believe in.
The difference, as this week has shown, is that when the customer misses its targets, every link in the chain jitters simultaneously.
Missed target aside, one of the most alarming details in the WSJ story was an internal report by OpenAI’s chief financial officer.
Sarah Friar, who joined the firm in 2024, told colleagues that the company may struggle to fund future computing contracts if revenue growth does not accelerate soon – a statement of remarkable gravity for a company valued at $852bn.
OpenAI has signed AI infrastructure deals totalling more than $1.15 trillion across Oracle, Microsoft and Amazon, projects spending $600bn on compute through 2030, and Deutsche Bank has estimated it could post $143bn in negative cumulative free cash flow between 2024 and 2029.
Altman and Friar issued a joint statement insisting they are “totally aligned on buying as much compute as we can.” Any suggestion of internal division was “ridiculous,” they said.
Altman’s mega IPO
OpenAI has been laying the groundwork for a stock market listing that would value it at more than $1 trillion – perhaps the largest debut in history, eclipsing even Saudi Aramco’s $1.7 trillion float in 2019.
Aramco was priced at roughly five times the revenue; OpenAI, at $1 trillion, would be valued at something approaching 80 times last year’s revenues.
For that to succeed, OpenAI needs acceleration, and Friar has raised doubts about whether the company can even meet the financial reporting standards that public markets demand, while Altman pushes for speed.
While the chief financial officer urges caution, the chief executive urges speed. It seems that markets this week sided with caution.
“I think it’s going to be a tough year for all these companies that want to go public at $800bn valuations, when the company might only be worth $400bn,” said Ross Gerber, founder and boss of Gerber Kawasaki.
What next for OpenAI
The AI industry, with OpenAI at its centre, has constructed a financial architecture unlike anything in tech history.
The technology is real, the demand is there, the long-term potential remains enormous, and the capability curve continues upward.
But this week has exposed that the most hype-inflated expression of the AI boom – a single private firm valued at $852bn that has never turned a profit, committed to spending more than a trillion dollars on infrastructure, locked in circular financial relationships with its own suppliers – may not survive intact long enough to capture the rewards it has promised its investors.
If revenues don’t catch OpenAI’s ambitions, the most likely outcome is absorption by one of the cash-rich hyperscalers – Microsoft, Amazon, or Google – that have already bet so heavily on its success.
Sam Altman remains confident that everything is on track, and that the compute is being bought.
But this week forces us to consider: what will happen if the world’s greatest AI pitchman can no longer outrun the numbers?