Oracle Shed 21,000 Jobs as Its AI Backlog Hit $638 Billion

The database company turned AI landlord cut 13% of its staff in twelve months while pouring a record $55.7 billion into the data centers it is building for OpenAI and other customers.

Oracle eliminated roughly 21,000 jobs over the past year, a figure the company tucked into its annual report and one that turns out to be far larger than anything it had confirmed publicly.

The filing, submitted to regulators on Monday, put Oracle's full-time headcount at about 141,000 as of May 31, down from around 162,000 twelve months earlier. That is a 13% reduction, and Oracle pointed straight at the technology it has staked its future on. Deploying AI across its operations, the company wrote, "have resulted, and may continue to result, in reductions to our workforce." It added that more cuts could follow.

A fivefold jump in restructuring costs

The bill for that shrinkage was steep.

Oracle spent about $1.8 billion on severance and other restructuring costs, a category that also covers office closures and contract terminations. A year earlier the comparable figure was $374 million. The roughly fivefold jump is the clearest signal in the report of how much reorganizing the company has done, and how fast.

The database giant now sells AI compute

To understand why a profitable software company would pay nearly two billion dollars to push out a seventh of its staff, look at what Oracle is becoming. The firm that spent four decades selling database licenses now rents out gigawatts of GPU computing to the companies training the world's largest AI models. That shift shows up most starkly in a single number. Remaining performance obligations, the contracted revenue Oracle has booked but not yet delivered, reached $638 billion at the close of its fiscal year. The backlog grew 363% in twelve months.

Cloud is the engine. In the quarter ending May 31, Oracle's cloud revenue climbed 47% to $9.9 billion, and its Oracle Cloud Infrastructure unit, the piece that sells raw AI compute, jumped 93%. Full-year revenue reached $67.4 billion.

The spending that drove the cuts

None of this is cheap. Capital spending hit $55.7 billion for the year, up 162% and well past the $50 billion Oracle had guided. Building at that pace dragged free cash flow to negative $23.7 billion. To cover the gap, the company raised about $48 billion through debt and equity during the year and told investors it expects to raise roughly $40 billion more in the next one.

Here is where the layoffs and the buildout connect. Analysts who have studied this year's tech cuts argue that the firings are less about software replacing workers and more about freeing cash for infrastructure. Payroll is one of the few large costs a company can slash fast enough to offset a capital program of this scale. At several of Oracle's larger peers, projected AI spending now runs four to five times the entire compensation bill, which means firing every employee still would not pay for the GPUs. The cuts are a financing decision dressed in the language of automation. Some of that language is what skeptics call "AI washing," pinning on artificial intelligence the reductions a company would have made for other reasons.

A $300 billion bet on a single customer

A second, sharper risk hides inside the $638 billion backlog. An estimated $300 billion of it, close to half, is tied to a single customer: OpenAI, through the Stargate data center program Oracle is helping construct. If OpenAI's funding or revenue stumbles, Oracle is left holding purpose-built infrastructure that cannot easily be re-rented to anyone else.

The rest of the industry is cutting too

Oracle is not alone in any of this. Amazon has cut around 30,000 roles in recent months while committing $200 billion to AI this year. Meta laid off 8,000 people in May, about a tenth of its workforce, with Mark Zuckerberg telling staff the reductions followed directly from the company's AI budget. Microsoft pushed voluntary buyouts on roughly 7% of its US employees. By some counts, more than 80,000 tech workers lost their jobs in the first quarter of 2026 alone, with AI named as a driver in close to half of those events. The four biggest spenders, Amazon and Microsoft along with Alphabet and Meta, plan to put as much as $725 billion into AI infrastructure this year, a 77% jump.

The human cost and the gamble

The people on the other side of these decisions do not transfer neatly into the new economy. India, where Oracle runs large engineering operations in cities such as Bengaluru, Hyderabad, Pune and Chennai, saw anxiety ripple through its IT workforce after reports of rescinded campus offers, though the company has not disclosed how many Indian roles were affected. The broader mismatch is stark. Industry trackers count around 275,000 open AI-specific positions at the same moment that laid-off database administrators and support engineers are struggling to qualify for them. Oracle conceded the strain in its own filing, warning that the reorganization could leave it short of skilled people in certain areas and dent productivity and earnings.

For now the company is betting that demand for AI compute holds long enough to justify the debt it has taken on and the payroll it has shed. The next test arrives with its first-quarter results, when investors will get to see whether revenue is finally catching up to the capital Oracle keeps pouring into the ground.