GPU demand is (~1Mx) distorted by efficiency problems which are being solved (opens in new tab)

Originally published on Andrew’s SubStack. GPU demand forecasts don’t distinguish waste from fundamental compute requirements. Mid-2024, Andrej Karpathy trained GPT-2 for $20. Six months later, Andreessen Horowitz reported LLM costs falling 10x annually. Two months after that, DeepSeek shocked markets with radical reductions in training and inference requirements. For AI researchers, this is all good news. For executives, policymakers, […]