Tag · 18 charts · newest first
AI Infrastructure
Our analysis and charts on the hardware behind AI: chips, data centres and supercomputers, and AI's energy, carbon and water footprint. The full set of charts is on the AI Infrastructure hub.
-
AI InfrastructureChina had 85 public or public-private AI supercomputers in 2025, as many as Europe and North America combined19 Oct 2026 · Chart · CSVView →
-
AI InfrastructureThe fastest AI chips now reach 2.5 petaFLOP/s at 16-bit precision19 Oct 2026 · Chart · CSVView →
-
AI Infrastructure84 notable AI models have been trained on Nvidia A100 chips, more than on any other accelerator19 Oct 2026 · Chart · CSVView →
-
AI InfrastructureAI chips worldwide reached 17.1 million H100-equivalents by end-2025; Nvidia supplied two-thirds19 Oct 2026 · Chart · CSVView →
-
AI InfrastructureAI data centre power capacity reached 29.6 GW by end-2025, close to New York state's peak demand19 Oct 2026 · Chart · CSVView →
-
AI InfrastructureThe US has 47% of the world's tracked data centres: 5,427 of 11,61419 Oct 2026 · Chart · CSVView →
-
AI InfrastructureThe US has 5,427 data centres, more than ten times Germany, the next largest19 Oct 2026 · Chart · CSVView →
-
AI InfrastructureThe most efficient AI chips now do 2.5 trillion operations per second per watt19 Oct 2026 · Chart · CSVView →
-
AI InfrastructureTraining Grok 3 drew an estimated 110 MW, the most of any model with published data19 Oct 2026 · Chart · CSVView →
-
AI InfrastructureTraining Grok 4 emitted an estimated 72,816 tonnes of CO₂, against 0.01 tonnes for AlexNet in 201219 Oct 2026 · Chart · CSVView →
-
AI InfrastructureTraining DeepSeek v3 emitted a fifteenth of the CO₂ of Llama 3.1 405B, despite more parameters19 Oct 2026 · Chart · CSVView →
-
AI InfrastructureThe most energy-hungry 2025 models used up to 23 Wh per medium-length prompt19 Oct 2026 · Chart · CSVView →