X · @teortaxesTex
· X / Twitter
At current V4 prices and speeds, I think DeepSeek would need on the order of 100K inference-only GPUs at max capacity to hit $8B ARR. Some 11.5 quadri…
At current V4 prices and speeds, I think DeepSeek would need on the order of 100K inference-only GPUs at max capacity to hit $8B ARR.Some 11.5 quadrillion output tokens, 230Q total (handwavy split). It's not a far-fetched scenario.myy: @teortaxesTex arr > $8bn…they need money badly for inference compute resources to ac