Global AI Compute Projected to Grow Sixfold by 2028, Led by Two Labs

Altimeter projects global installed AI compute will expand from 18 GW in 2025 to 115 GW by 2028, with the industry potentially adding 43 GW in 2027 alone. OpenAI and Anthropic's combined share of global AI data center compute may rise from 20% to 35% over that period. Dylan Patel estimates each lab will grow from roughly 2 GW entering 2026 to above 5 GW by year end, reaching north of 50 GW each by end of 2028, with the two combined capturing 70-80% of incremental compute. Anthropic committed to up to 1 million Google TPUs through a roughly $200 billion financing structure involving Blackstone, Apollo, Morgan Stanley, and Broadcom, with Fluidstack building the sites. Patel cites Anthropic revenue reaching as high as $50 million per megawatt, with the lab projecting $190-200 billion in 2028 revenue. Jensen Huang stated 400,000 GPUs are planned to come online at Stargate Texas in the coming year. On valuations, Gavin Baker argued that memory makers at 3-5x PE alongside Nvidia at a comparatively low PE cannot both be pricing the same data center spending boom consistently. Europe holds less than 5% of global AI compute capacity versus 75% for the United States, and its data-center capacity shortfall could grow from roughly 3 GW today to 14 GW by 2030.