I watched the after-hours tape as Oracle’s shares tumbled and a roomful of investors went quiet. You could feel the math closing in: more borrowing, more data centers, and a thinner safety net. The company had just said it would raise another $40 billion (€37 billion) to fund AI infrastructure, and that landed like a physical thud.
I’ll walk you through what matters, why this rattled the market, and where the real danger sits. Think of me as the colleague who points out the cracks before you sign the lease.
Trading screens flashed red after Oracle’s earnings release.
The headline number — $40 billion (€37 billion) of new debt and equity planned next fiscal year on top of $43 billion (€40 billion) raised last year — is not subtle. With well above $100 billion (€92 billion+) of exposure tied to AI projects, Oracle sits among the largest debt issuers in the tech boom.
Oracle says the cash covers a buildout that included $55.7 billion (€51 billion) in capital expenditures last fiscal year, roughly $5 billion (€5 billion) more than early estimates. That level of spending turned free cash flow negative, and analysts now model that weakness into 2030. Investors reacted the way they often do to surprise leverage: they sold first and asked questions later.
How much debt did Oracle raise for AI?
In the last fiscal year Oracle raised $43 billion (€40 billion) and now plans an additional $40 billion (€37 billion). Those totals are explicitly tied to data-center construction and related AI infrastructure, which is why debt markets are watching this as a sector-level signal.
At a public pitch last year, Stargate was framed as a national-scale buildout.
Oracle’s Stargate program promised 7 gigawatts of planned capacity — roughly the power to run more than 5 million homes — and that ambition pushed the stock higher, briefly making Larry Ellison one of the richest people on the planet. At the time, the market rewarded outsized bets on AI infrastructure.
But timelines slipped and projects stalled. When construction delays met rising borrowing, the arithmetic shifted from optional growth to balance-sheet pressure. The company’s cost discipline turned into layoffs — Oracle cut thousands of roles earlier this year — in part to slow the burn. That spending pattern, paired with slower project delivery, leaves Oracle liable for interest and principal long before many of the revenue streams are fully realized. It’s like a pressure cooker: heat and pressure build until someone lifts the lid.
Reports surfaced that OpenAI’s finance team flagged revenue concerns.
OpenAI is Oracle’s single-largest reported customer for certain infrastructure contracts, a relationship that elevated Oracle’s risk profile by tying its cash flow to another company’s ability to pay. Public reports said OpenAI CFO Sarah Friar worried about revenue growth and the firm’s capacity to meet computing bills.
There are other dynamics at play: Anthropic and Google are aggressively competing, and OpenAI is still burning cash without a clear path to sustained profit. If OpenAI scales back purchases or delays contracts — remember the cited $300 billion deal figure in some reports — that would leave Oracle servicing debt at scale without matching customer receipts. That scenario turns a large bet into a fragile one, a house of cards that could collapse faster than new capacity comes online.
Can OpenAI afford its computing contracts with Oracle?
Short answer: uncertain. OpenAI has significant traffic and growth potential, but it is also under competitive pressure and high cash burn. Public concern from its CFO about revenue growth is a red flag for counterparties like Oracle who are counting on multi-year compute spend. If OpenAI trims commitments, Oracle’s forecasted revenues tied to those contracts would be at risk.
On the tape and behind closed doors, investors are recalibrating expectations.
Oracle still reported record revenues, which is why the reaction was so stark: strong top-line performance matched with outsized capital commitments and rising leverage. When I speak with investors, they frame the issue simply — returns on these AI infrastructure bets must arrive before debt costs compound and investor patience runs out.
Yes, Oracle is financing scale to win share in the AI market. But financing scale at the pace the company announced puts its balance sheet and the broader market in a sensitive position. If customer payments slow or projects underperform, you’re not just watching a single company wobble; you’re watching a potential stress point for lenders and suppliers tied to the buildout.
So where does that leave you as someone tracking AI and markets? Watch three things closely: Oracle’s quarterly cash flow updates, milestone progress on Stargate data centers, and any public signs of strain from OpenAI or other large buyers. Tools you probably already use — SEC filings, Bloomberg terminal screens, and earnings call transcripts — will be the fastest way to separate noise from credible risk.
Oracle has bet big on a future where companies pay for vast amounts of compute. The market is asking whether those customers will pay on time, and whether the scale of borrowing was wise. If the answer leans negative, the shock won’t be limited to one balance sheet; it could rattle banks, vendors, and even tech hiring plans — and that raises the question everyone should be asking now: who absorbs the bill if the AI buildout stops paying back?