On the morning of March 31, Oracle employees across five countries opened their inboxes to find a termination email from “Oracle Leadership.” No call from a manager. No HR conversation. Just a DocuSign link and an access revocation. TD Cowen estimates the cuts will reach 20,000 to 30,000 people — roughly 18% of Oracle’s 162,000-person workforce.
The usual explanation for layoffs this size is that the company is struggling. Oracle is not. Its Q3 FY2026 earnings showed net income of $6.13 billion, up 95% year over year. Remaining performance obligations — contracted future revenue — hit $523 billion, up 433%. This is not a company in revenue distress. It is a company harvesting its own SaaS workforce to fund a different business entirely.
The capital reallocation machine
The financial logic is blunt. Oracle has committed to an estimated $156 billion in AI infrastructure capex, anchored by a $300 billion, five-year cloud computing contract with OpenAI announced in mid-2025. Building 4.5 gigawatts of data centre capacity requires cash Oracle doesn’t have on hand. Since formalising the OpenAI deal, Oracle has taken on over $100 billion in total debt and swung to deeply negative free cash flow, not expected to recover until around 2030.
The workforce reductions are projected by TD Cowen to free up $8–10 billion in cash flow. Oracle’s own March 2026 10-Q SEC filing disclosed a $2.1 billion restructuring plan, with $982 million already booked in the first nine months of the fiscal year.
Where did the cuts land? Reports from TNW, Bloomberg, and employee posts on Reddit and Blind confirm that Revenue and Health Sciences (RHS) and SaaS and Virtual Operations Services (SVOS) — the operational spine of Oracle’s cloud applications business — saw reductions of at least 30%. Oracle didn’t cut its data centre teams. It cut the humans who build, sell, and support its SaaS products.
The SaaSPocalypse — but from inside
Larry Ellison gave the game away on Oracle’s Q3 earnings call when he said the “SaaSpocalypse applies to others but not to us.” Oracle CEO Mike Sicilia framed the company’s AI adoption as a defence: “We are building brand new SaaS products using AI and also embedding AI agents right into our existing application suites.”
The contradiction runs deeper than the usual “AI replaces workers” story. Oracle is not saying AI made 30,000 of its people redundant. It’s saying it needs the money those people cost to build something else. The SaaS business is being treated as a cash cow to be milked for AI infrastructure — the exact trajectory the hypothesis predicts, but applied reflexively. The software vendor is commoditizing its own product division’s workforce to fund the technology that will commoditize software.
This is the capital-allocation version of seat compression. When we’ve tracked Atlassian’s declining seat counts or Workday’s headcount-linked license pressure, the mechanism was external: AI reduces the customer’s need for human workers, which reduces the customer’s need for per-seat software. Oracle adds an internal vector. The vendor itself decides its SaaS operations require fewer people, because the margin is better deployed building AI infrastructure. The SaaS division doesn’t shrink because customers leave — it shrinks because the parent company redirects its operating budget.
The market isn’t buying the pivot
Wall Street initially loved the OpenAI deal. Oracle’s stock surged 43% in a single day in September 2025. Since then, shares have fallen 54% from that peak. Multiple US banks have pulled back from financing Oracle-linked data centre projects. And in early March, Bloomberg reported that Oracle and OpenAI had scrapped plans to expand the flagship Stargate data centre in Abilene, Texas — OpenAI wanted newer Nvidia chips available elsewhere, not the Blackwell processors Oracle had already ordered.
The Stargate dispute reveals a second-order risk for software companies betting their balance sheets on AI infrastructure: the customer’s requirements move faster than capital cycles. Oracle committed billions to construction and hardware before OpenAI’s demand forecasts stabilised. That timing mismatch — infrastructure capex measured in years, AI model requirements measured in months — creates a structural fragility that pure software businesses never had to manage.
What this tells us
The hypothesis says software loses value because AI can create it cheaply. Oracle’s restructuring suggests a darker corollary: software vendors agree with the hypothesis, at least implicitly, and are acting on it with their own capital allocation. When a company with $6 billion in quarterly net income decides its SaaS workforce is a less productive use of capital than GPU clusters, the revealed preference is clear. The SaaS business funds the AI transition; it doesn’t survive it unchanged.
The falsification check: is Oracle an outlier? Not entirely. The pattern of AI capex crowding out SaaS operational spending echoes what we tracked in the broader IT budget data — AI budgets growing 100%+ while overall IT budgets grow 8%, starving everything else. Oracle is just doing it inside a single balance sheet instead of across a customer’s vendor portfolio.
What to watch: whether Oracle’s SaaS revenue holds through fiscal Q4 with 30% fewer people in its operational divisions. If it does, that’s evidence that AI-augmented operations really can maintain output with radically fewer humans — precisely the mechanism that makes seat-based pricing obsolete. If customer satisfaction scores or renewal rates crack, it’s evidence that the “SaaSpocalypse applies to others but not us” story was premature cost-cutting dressed up as strategy.
Sources: Oracle Q3 FY2026 earnings release, TNW, CX Today, Bloomberg, Times of India, TD Cowen via DatacenterDynamics