Verified70% confidenceBenchmarkExact time
Research shows that a single training run of GPT-3 consumes approximately 700,000 liters of fresh water, though this figure is subject to considerable controversy due to varying system boundary definitions.
4
Sources
70%
Confidence
Long-term
Relevance
7/6/2026
First Seen
Sources
AI Data Center Water Consumption: The Hidden Environmental Cost We've Been Underestimating
hackernewshackernews7/3/2026
Related Claims
UnverifiedTraining GPT-3 consumed approximately 3.14×10²³ FLOPs.71% similarVerifiedIt is estimated that training a large language model of GPT-3's scale requires approximately 1,287 megawatt-hours of electricity.69% similarVerifiedIt is estimated that training GPT-3 once emits approximately 552 metric tons of CO2 equivalent.69% similarUnverifiedApproximately 60% of GPT-3's training data came from filtered Common Crawl.67% similarPartially VerifiedFew-shot learning capability was almost unusable in the era of GPT-2 (1.5 billion parameters) and suddenly became practical with GPT-3 (175 billion parameters).67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/115191API
curl https://kongchang.com/api/v1/knowledge/claims/115191MCP
get_claim(id=115191)