Big Tech’s infrastructure buildout has a price tag that goes beyond dollars. While Amazon, Microsoft, Alphabet, and Meta race toward a projected $600 billion in AI infrastructure spending in 2026, Oracle quietly cut roughly 21,000 jobs — about 13% of its entire workforce — as it scrambled to secure its own seat at the table. The AI race is the largest capital expenditure cycle in recent tech history, and someone is always paying for it.
The Spending Machine Nobody Can Stop
The numbers behind this buildout are staggering — and the cash strain is already showing.
Think of this capex cycle like streaming services circa 2019: everyone is burning cash on content libraries before a single subscriber has renewed. According to Reuters, the spending has strained cash flows and tested Wall Street’s patience. The money flows into data centers, specialized chips, networking gear, and the power systems needed to keep it all running.
Oracle’s numbers tell the story in stark detail:
- Projected AI infrastructure spend across four major hyperscalers in 2026: approximately $600 billion
- Oracle fiscal 2026 capital expenditures: $55.7 billion
- Oracle free cash flow: negative $23.7 billion
- Oracle contracted backlog: $638 billion, up from $138 billion the prior year
That backlog explosion traces back to a reported $300 billion contract with OpenAI, which catapulted Oracle from a legacy database company into a central Stargate Project-era AI infrastructure supplier. The contracted revenue is enormous. The cash to deliver it hasn’t arrived yet.
21,000 Jobs Is Not a Rounding Error
Oracle’s workforce shrank by roughly 13% in a single year, and the company says AI-driven cuts may continue.
Oracle’s headcount dropped from approximately 162,000 to 141,000 as of May 31, 2026. The company’s annual filing, cited by Bloomberg, stated plainly that AI adoption across its operations has led — and may continue to lead — to workforce reductions. No corporate euphemism softens what $1.8 billion in severance and restructuring expenses actually represents.
This is the winner-take-most dynamic playing out in real time. Companies are spending aggressively now to lock in future compute capacity. Bulls point to that $638 billion backlog as evidence the bet eventually pays off. Skeptics counter with negative free cash flow and a payback timeline that remains stubbornly theoretical. Both sides have receipts.
That $638 billion backlog suggests Oracle’s wager may pay out — eventually. But 21,000 people weren’t in a position to wait for “eventually.” If AI monetization lags behind the pace of infrastructure spending, the pressure on margins and headcounts only intensifies from here.
