Silicon Atlas(@Silicon_Atlas):An AI energy forecast has to connect workload demand to the whole system that serves it. NVIDIA's DSX framing brings compute, power, and cooling into that picture. Start with the model, input/output lengths, concurrent requests, and response-time targets. Those conditions shape GPU energy per output token. Then account for how the fleet is used over time, the rest of the IT equipment, and cooling. A site's power cap sets its ceiling; actual energy consumption depends on how that system runs. The useful forecasting question is: how much energy will this workload require across the system that serves it?

2026.09.17 07:06

An AI energy forecast has to connect workload demand to the whole system that serves it. NVIDIA's DSX framing brings compute, power, and cooling into that picture. Start with the model, input/output lengths, concurrent requests, and response-time targets. Those conditions shape GPU energy per output token. Then account for how the fleet is used over time, the rest of the IT equipment, and cooling. A site's power cap sets its ceiling; actual energy consumption depends on how that system runs. The useful forecasting question is: how much energy will this workload require across the system that serves it?

NVIDIA AI Infrastructure@NVIDIAAIInfra

2026.09.16 19:50

AI factory efficiency takes a full-system approach. The NVIDIA DSX AI factory platform brings compute, power, cooling, and grid flexibility together to make the entire AI factory more efficient and generate greater AI output from available energy.