Unverified50% confidenceBenchmarkExact time
From NVIDIA's A100 to H100 to B200, each generation provides approximately 1.5 to 3 times the AI training compute at the same power consumption.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
9/24/2026
First Seen
Valid until: 12/23/2026
Sources
Related Entities
Related Claims
UnverifiedResearch data from Epoch AI shows that the computing power used for AI training has grown by more than 300,000 times from 2012 to the present.75% similarVerifiedResearch data from Epoch AI shows that the training compute for frontier AI models roughly doubles every 6–10 months, far exceeding the growth rate of Moore's Law.72% similarUnverifiedThe total energy consumption of the AI inference phase has gradually surpassed that of the training phase, because inference runs continuously 24×7.71% similarVerifiedThe AI industry is transitioning from the training side to the inference side, with the inference-side competition centered on tokens per second throughput, energy consumption per token, and time to first token as core metrics.70% similarUnverified随着AI模型参数量持续膨胀，单个训练作业跨越多个NVL72机架将成为常态70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/948290API
curl https://kongchang.com/api/v1/knowledge/claims/948290MCP
get_claim(id=948290)