Oracle’s mass layoffs reveal a hard truth: betting big on AI means someone always pays
Larry Ellison has never been a man given to understatement. The Oracle chairman, who once compared himself favourably to Napoleon and has spent decades building one of the world’s most powerful software empires, is now staking that empire on a single, audacious wager: that artificial intelligence is the greatest human achievement will consume computing power at a scale the world has never seen, and that Oracle will be the one supplying it.
The cost of that wager is now landing on Oracle’s workforce. Per Bloomberg, the company is planning to cut thousands of jobs. The estimates for this range from a few thousand to as many as 30,000 in what promises to be its most sweeping round of redundancies in years.
The culprit, if one can call it that, is ambition. Oracle has committed to a $300 billion partnership with OpenAI, the Sam Altman-led artificial intelligence laboratory that has become Silicon Valley’s most prized (and most expensive) client. Building the data centres capable of honouring that commitment is proving ruinously capital-intensive.
The numbers tell an uncomfortable story. Oracle’s capital expenditure for fiscal 2026 is now expected to run $15 billion higher than the $35 billion originally projected, a revision so dramatic it sent analysts scrambling to revise their models. The company burned through roughly $10 billion of cash in just the first half of its fiscal year.
Wall Street, rarely short of optimism for a technology giant riding the AI wave, is nonetheless projecting Oracle’s cash flows to remain in the red until at least 2030. To plug the gap, Oracle said it would raise between $45 billion and $50 billion in 2026 through a mix of debt and equity sales. That’s a fundraising drive of almost wartime proportions.
And yet, the money is still not enough. Investment bank TD Cowen estimates that the planned job cuts could generate between $8 billion and $10 billion in free cash flow. That is no doubt a meaningful sum, but one that offers a sobering illustration of just how wide the financing gap has grown. American banks, which might once have been expected to step up with loans, have reportedly begun pulling back from funding Oracle’s data centre plans, adding urgency to the workforce reductions.
There is a certain dark irony in the situation; a company building the infrastructure to automate labour is itself automating labour to pay for that infrastructure. It is a loop that would delight a satirist. More prosaically, Oracle has also begun asking customers to contribute to the cost of data centre construction and is exploring a “bring your own chip” model, whereby new clients would supply their own hardware. The bill, in other words, is being distributed as widely as possible.
Oracle’s predicament reflects a broader dynamic beginning to surface across the technology industry. The AI gold rush has been characterised by colossal spending commitments made at extraordinary speed, with the expectation that demand would justify the outlay. For Oracle, that demand is real; its cloud infrastructure customers include not only OpenAI but also Meta, Nvidia, TikTok, and Elon Musk’s xAI. The contracted revenues are substantial. The problem is that the data centres must be built before the revenues arrive, and building them requires money the company does not yet have.
Oracle is also considering the sale of Cerner, its healthcare software division acquired for $28.3 billion in 2022. Whether it finds a willing buyer at a suitable price will say much about the market’s confidence in Oracle’s strategic direction. Cerner was purchased at the peak of optimism about healthcare technology; selling it now would represent a meaningful retreat from one of Ellison’s previous grand ambitions.
What the episode ultimately reveals is that the infrastructure layer of the AI economy requires a tolerance for financial pain that not every company can sustain. It is an unglamorous business to be building and running the data centres that underpin everything from chatbots to drug discovery, albeit essential, so one can understand Oracle’s big bet. But Amazon, Microsoft, and Google, Oracle’s principal rivals in cloud computing, are better capitalised and have been building at scale for longer. Oracle is the upstart in this race, and upstarts, however bold, must sometimes sacrifice comfort for competitive position.
Whether that position ultimately proves worth the sacrifice depends on whether the AI boom continues to accelerate, whether OpenAI and its peers remain committed customers, and whether Oracle can execute at the scale it has promised. The workers receiving redundancy notices this month might reasonably question whether the gamble was theirs to take. The answer, unfortunately for them, tends to be the same in every era of industrial transformation: it never is.
