Big Tech Needs $2T in AI Revenue by 2030 or They Wasted Their Capex (www.wheresyoured.at)

Tech giants have poured roughly three-quarters of a trillion dollars into AI infrastructure over the past few years — building data centers, buying GPUs (A100, H100 and now Blackwell) and signing large cloud commitments — but haven’t shown proportional revenue. The piece argues that Microsoft, Amazon, Google and Meta must collectively add about $2 trillion in AI-driven revenue by 2030 (roughly the next four years) to justify that capex, or face massive stranded-asset losses, write‑downs and investor scrutiny. Concrete red flags include Microsoft’s stop‑start disclosure of AI ARR ($833M/month in Oct 2024, $1.08B/month Jan 2025 then silence), OpenAI’s restructure giving Microsoft a 27% stake and a $250B Azure spend booked as “remaining performance obligations,” OpenAI’s heavily discounted compute payments that only cover Microsoft’s operating costs, and NVIDIA’s claim of >$500B in GPU bookings. Technically, the economics are brutal: GPUs are extremely costly to buy, operate and cool; their value decays quickly as new chips arrive yearly; reports even suggest negative gross margins on some Blackwell deployments. Meanwhile, most startups and hyperscalers remain unprofitable on AI services, and much of the cloud spend is effectively subsidized or deferred via RPO accounting. The implication for AI/ML: expect tougher cost discipline, pressure to monetize models (pricing for compute and services), potential consolidation, and a reckoning if expected revenue growth doesn’t materialize — otherwise years of capex could become sunk losses.