OpenAI’s expanding reach across the global industry has become one of the defining stories of the decade, yet the company behind ChatGPT is still wrestling with one financial truth that refuses to go away: the numbers don’t add up.
According to a report by Fortune, HSBC Global Investment Research now projects that OpenAI will remain unprofitable through 2030, even as adoption spreads across nearly half of the world’s adult population and revenues soar into territory that would place it among the world’s largest tech companies.
The bank’s projection has quickly turned into one of the most consequential assessments of the AI industry this year, because it confronts the question everyone from Wall Street to Washington to Silicon Valley has tiptoed around: can generative AI ever make enough money to justify the astronomical spending required to run it?
Register for Tekedia Mini-MBA edition 19 (Feb 9 – May 2, 2026): big discounts for early bird.
Tekedia AI in Business Masterclass opens registrations.
Join Tekedia Capital Syndicate and co-invest in great global startups.
Register for Tekedia AI Lab: From Technical Design to Deployment (next edition begins Jan 24 2026).
HSBC’s semiconductor research team, led by Nicolas Cote-Colisson, updated its model after factoring in OpenAI’s new multiyear compute agreements, including a $250 billion cloud commitment with Microsoft and a $38 billion deal with Amazon. Those agreements came without new capital injections, deepening the financial strain.
Their updated conclusion is blunt. By 2030, OpenAI’s revenues are projected to exceed $213 billion. Its customer base is expected to include roughly 44% of the world’s adults, up from 10% in 2025. Yet the company will still be losing money against a wall of infrastructure costs that HSBC estimates will reach $792 billion between now and the end of the decade, with a data-center rental bill of $620 billion alone. Cumulative free cash flow remains sharply negative, leaving a funding hole of about $207 billion that OpenAI will need to fill through fresh debt, equity, or aggressive monetization.
The bank models total compute commitments rising to $1.4 trillion by 2033. OpenAI itself has referenced that same figure over an eight-year horizon. That scale has no precedent in the history of commercial technology, raising questions about the carrying capacity of the global capital markets.
The data-center electricity burden is enormous
The scale of compute OpenAI is now targeting adds a second layer of concern. The company’s goal of 36 gigawatts of compute capacity by 2030 requires a physical and electrical buildout that rivals that of a small U.S. state. One gigawatt powers roughly 750,000 homes, meaning OpenAI’s compute footprint would require electricity consumption comparable to a state between the size of Florida and Texas.
This buildout is reshaping power markets, construction timelines, real-estate valuations, and even utility-regulator planning cycles. Analysts tracking power-grid expansion say the U.S. has not faced a comparable single-industry demand surge since the rise of heavy manufacturing in the mid-twentieth century. Some energy economists now argue that the strain created by AI data centers is becoming a structural factor in future electricity pricing, something HSBC cites as a growing cost pressure on long-term compute models.
It explains why Altman’s recent comment—asked whether OpenAI could ever have “enough” compute—came out in a single exasperated word: “Enough.”
Debt markets are showing signs of AI fatigue
The other growing tension comes from the credit market. HSBC warns that debt is “possibly the most challenging avenue” for OpenAI to pursue right now. Oracle and Meta have both raised substantial amounts of debt this year to finance their own AI expansions. Those raises triggered visible market unease, including a sharp rise in Oracle’s credit default swaps, which Morgan Stanley’s Lisa Shalett flagged as a concerning signal. Even JPMorgan strategist Michael Cembalest noted that hyperscalers traditionally funded themselves through free cash flow rather than borrowing, making the current shift unusual.
Against this backdrop, investors are starting to question whether the economics of AI data centers can support debt loads of this magnitude, especially when returns remain uncertain, and many of the industry’s leading players are years away from reliable profitability.
OpenAI’s need for continuous capital puts it in a difficult position. It cannot slow its infrastructure buildout without risking competitive disadvantage. Yet the debt markets are flashing caution, and the equity markets may balk at the size of future raises if profitability milestones remain far over the horizon.
Microsoft’s exposure is deeper than any other company’s
Microsoft sits at the center of this tension. It is OpenAI’s largest investor, biggest partner, and main cloud provider. Its $250 billion cloud agreement with OpenAI is one of the most consequential pieces of the entire arrangement, because the cost of compute expansion directly flows through Azure’s infrastructure. Microsoft’s AI strategy is now intertwined with OpenAI’s financial stability and compute demand in a way that analysts compare to a shared balance-sheet dependency.
If OpenAI stumbles, Microsoft absorbs both operational and strategic shock. Azure’s data-center buildouts are anchored to OpenAI’s growth path. Office, Windows, GitHub, and Bing integrations rely on OpenAI’s underlying models. And Microsoft’s own market valuation has been propped up in part by expectations that generative AI will drive the next decade of revenue growth. Any slowdown in OpenAI’s trajectory could hit the world’s most valuable company at multiple points simultaneously.
At the same time, Microsoft provides stability that OpenAI cannot easily replicate elsewhere. The company has the cash flow to sustain multi-hundred-billion-dollar infrastructure expansion in a way few corporations on earth can match. For now, that makes Microsoft one of OpenAI’s key lifelines, even as it carries significant exposure on its own books.
The wider productivity debate has returned with new intensity
The enormous financial strain has also revived a debate that economists have circled for years: whether the promised productivity gains from AI will ever show up in national statistics. HSBC echoed the well-known remark by Nobel laureate Robert Solow that modern economies seem able to generate computers and software everywhere except in their productivity numbers.
Some economists believe this time will be different. Harvard’s Jason Furman, quoted by Fortune, calculated recently that without data centers, U.S. GDP growth in the first half of 2025 would have been just 0.1%.
Bank of America’s Savita Subramanian told Fortune in August that she sees genuine structural productivity improvements emerging out of the 2020s economy, though not purely because of AI. Instead, she said that companies have been forced by post-pandemic wage inflation to redesign operations to “do more with fewer people,” replacing manual processes with scalable systems. Still, she noted that the most innovative tech firms have shifted from asset-light models to enormous capital-heavy ones, especially in data-center construction, which carries considerable financial risk.
That tension sits at the center of OpenAI’s story. The company dominates the consumer AI landscape and shapes global economic expectations, yet the productivity gains from generative AI at an industrial scale remain difficult to quantify.
The unresolved question for markets
The stakes in the next five years are unusually high. OpenAI has become the most visible avatar of an AI revolution that has swept the global economy, yet it faces one of the most daunting financial challenges any tech company has ever confronted. HSBC’s verdict—that OpenAI will still not be profitable by 2030—lands with force because it captures the contradiction at the heart of this moment: extraordinary technological progress built on a foundation of negative cash flow and soaring infrastructure costs.
The company is now asking global markets to keep funding a buildout that requires trillions of dollars, in anticipation of productivity and revenue gains that are still not proven. And the broader market, while enthusiastic, is starting to show signs of strain.
The next chapter will turn on whether OpenAI can convert its dominance into durable profit fast enough to justify a compute bill that resembles the electrical needs of a small country, financed in part by debt markets that are becoming more anxious, and stitched into the strategic core of a company—Microsoft—whose own valuation is tied to the very same gamble.