User says a smooth DeepSeek workflow could justify spending $500K on GPUs

yacineMTB · x · 2026-08-04

The author says they need to make a bit more money, then adds that if a prompt they just sent to DeepSeek works smoothly after Codex’s “sandbagging,” they might spend as much as $500K on GPUs.

It’s a small but revealing note about practical model testing and how quickly serious users translate a good coding workflow into compute spend.

Related event: DeepSeek's Performance Sparks GPU Buying Banter(3 posts)→

Original post →

More from coding & agent

coding & agent channel →