power consumption

Power consumption is the amount of electrical energy used by computing systems, particularly large language models and the data centers that run them. It scales with factors like model size, training runs, and inference frequency.

Why it’s in the news: It is in the news due to the massive and unpredictable energy demands of large AI models, raising concerns about grid capacity and environmental impact.

Latest on power consumption

• nobody is talking about the power consumption required to run them at scale
• Large language models don't consume power on a predictable curve — they scale with context length, inference frequency, and training runs that haven't been built yet
• Son's bet is fundamentally terrestrial — more data centers, more power consumption, more grid dependency
• Every large language model training run consumes megawatts.
• Nvidia goes green to keep grid capacity from zapping its revenues

Connections

4 entities linked to power consumption across the news graph.