In One Night, 30,000 Laid Off as AI Infra Firms Recast Work — Four Programmers Speak Out

A mass purge and personal stories

It has been reported that roughly 30,000 workers were let go in a single wave as infrastructure-focused tech firms rationalize costs amid an AI arms race. Reportedly, four programmers affected by the cuts described disrupted lives — sudden unemployment, stalled projects and an uncertain job market where their skills are being re‑priced. Short, brutal and very public: that is what the transition looks like for many engineers who built the plumbing of the old software world.

Money, machines and a changing value chain

Why now? It has been reported that Oracle (甲骨文) plans to raise annual capital expenditures to around $50 billion to build data centres and AI infrastructure — a shift that, according to TheStreet, has already pushed the company’s free cash flow from roughly $11.8 billion in 2024 into negative territory and contributed to a roughly 25% slide in its stock this year. AI infrastructure is heavy‑asset: high‑end GPUs cost tens of thousands apiece and full data centres can require tens of billions to build. As value concentrates around token generation (models) and token consumption (large user platforms), traditional infrastructure increasingly looks like electricity or bandwidth — necessary, but low‑margin. Who keeps the pricing power? The answer helps explain why headcount is suddenly the easiest cost to trim.

The human and geopolitical angle

Automation and better tooling are replacing many support roles — system maintenance, data processing, ops — even as some firms double down on hiring to seize opportunity. WHOOP, for example, is reportedly planning to hire about 600 people and its CEO framed today’s layoffs as a chance to recruit top talent. Geopolitics complicate the picture: U.S. export controls and broader trade frictions that restrict access to cutting‑edge chips have raised costs and supply uncertainty for global AI builds, making large capital outlays even riskier for firms that do not control models or customer gateways. The result is a bifurcated market: winners who own models or massive user funnels, and everyone else squeezed toward cost efficiency. Can engineers caught in the middle pivot quickly enough? For now, many are left to answer that question themselves.