← GPU Economics

Sub-$2/hour H100s mean the market is in oversupply

Why Decentralized AI Needs Cosmos: Greg Osuri of Akash Explains (The Interop)

“If one wants to amortize their chip within two years, I think the price is around $2 an hour, roughly, including the energy cost… so if anything less than $2 an hour, you’re losing money. We have chips for [a] dollar an hour too… that tells you the market is — there’s oversupply of chips now.” — 00:21:57

Context: Post-2023 GPU crunch: opportunists financed small fragmented GPU fleets, demand didn’t catch up, and loose (non-interconnected) GPUs crashed in price while large clusters stayed expensive. He notes Lambda went from refusing crypto buyers to “begging us to take the GPUs.”

See it among all GPU Economics predictions →