NVIDIA Blackwell''s High Power Consumption Drives Cooling
High-performance AI server models like the B200 and GB200 are designed for maximum efficiency, with individual GPUs consuming over 1,000W. HGX models will house 8 GPUs each,
Maximum power consumption of AI server
Home / Maximum power consumption of AI server
AI servers consume significantly more power than traditional IT equipment, primarily due to the use of GPUs and high-performance accelerators. Typical ranges include: • Traditional servers: 300–800 W per server • GPU servers: 2–10 kW per server • AI racks: 20–100+ kW per rackWhere traditional server racks once operated at around 5–10 kW, modern AI environments are pushing far beyond that, often reaching 30 kW, 60 kW or even over 100 kW per rack. According to RAND Corporation research, AI data centers could require 68 gigawatts of power capacity globally by 2027, close to California's entire power grid. Today, a single NVIDIA GB200 NVL72 AI rack draws 132 kW — more than 16 times as much. It's a fundamental rewrite of how data centers provision, generate, store, and back up power. The IEA's latest report, Key Questions on Energy and AI (April 2026), puts the updated trajectory plainly: consumption will roughly double and reach almost 500 TWh in.
