THOUSANDS OF ORACLE employees woke up on Tuesday to find their careers had ended while they slept. At 6 a.m. Eastern, termination notices landed in inboxes across the United States, India, Canada, and Mexico — no call from a manager, no heads-up from HR, just a form email signed "Oracle Leadership" and a link to submit a personal address before their accounts went dark. The layoffs, estimated by TD Cowen at between 20,000 and 30,000 people, represent roughly 18% of Oracle's 162,000-person workforce and would rank among the largest in the company's 48-year history.
The arithmetic behind the cuts is blunt. TD Cowen reckons the reductions will free up $8 billion to $10 billion in annual cash flow — money Oracle urgently needs to fund a debt-fueled expansion into AI data centers that has left it carrying some $148 billion in long-term debt, up nearly 50% from the end of 2025. Capital expenditures are expected to hit $50 billion in the current fiscal year, roughly double what Oracle spent a year prior. The company disclosed a $2.1 billion restructuring charge in its most recent SEC filing, with nearly $1 billion already booked. Its stock has lost more than half its value since peaking in September, and its credit-default swap spreads — a measure of perceived risk — recently hit levels not seen since 2009. One Oracle investor told Yahoo Finance the company had become the "poster child for fears of an AI bubble."
Yesterday's chips, tomorrow's debt
Yet the layoffs are not the story of a company in decline. They are the story of a company making an enormous, leveraged bet on a future that keeps moving. Oracle's remaining performance obligations — contracted revenue not yet recognized — stood at $553 billion at last count, up more than fivefold year-over-year, the bulk of it tied to a deal worth more than $300 billion with OpenAI. Net income jumped 95% last quarter to $6.13 billion. Revenue is expected to reach $67 billion this fiscal year and, if management is to be believed, $225 billion by 2030. This is not a business running out of customers. It is a business running out of cash.
But the deeper complication is not financial — it is physical. Oracle is building data centers that may be stocked with outdated hardware before the power is even connected. In early March, CNBC reported that OpenAI had walked away from plans to expand the Stargate data center in Abilene, Texas — the crown jewel of its partnership with Oracle — because it wants facilities equipped with Nvidia's next-generation Vera Rubin chips rather than the Blackwell processors the Abilene site is designed to house. Nvidia now ships a new generation of data center processors every year, each offering a leap in capability; Vera Rubin, unveiled at CES in January, delivers roughly five times the inference performance of its predecessor. For companies training frontier models, where marginal performance gaps translate directly into market share, last year's silicon is a competitive liability.
This creates an agonizing mismatch. Securing a site, connecting power, and standing up a facility takes 12 to 24 months at minimum. But Nvidia's roadmap moves on a 12-month cycle. Oracle — the only major hyperscaler funding the AI buildout primarily with debt rather than cash flow, and the only one carrying a BBB-tier credit rating while its peers sit comfortably at AA or A — is uniquely exposed to this timing risk. Amazon, Microsoft, and Google can absorb a generation of chip depreciation as a cost of doing business. Oracle is borrowing at a pace that requires its bets to pay off on schedule. It has taken on $58 billion in new debt in just the past two months; some of its banking partners have reportedly stepped back from financing certain projects; and its bondholders have sued, alleging misleading statements in an $18 billion debt sale last September.
The company's co-CEOs, Clay Magouyrk and Mike Sicilia — who replaced longtime chief executive Safra Catz in October — have argued that demand for AI infrastructure continues to exceed supply. That is true, for now. But demand for yesterday's AI infrastructure is not the same as demand for AI infrastructure in general, and Oracle is building at a scale where the distinction matters enormously. The question facing investors is not whether the AI boom is real; the remaining performance obligations suggest it is. The question is whether a company can borrow its way to a $225 billion revenue target when its anchor customer is already shopping for newer hardware elsewhere — and when 30,000 employees just learned, at dawn, that they were the first line item to be optimized. ■
For more, join 75,000 subscribers getting tech's favorite brief here
