Deepseek prices their models based on a 10 month hardware payback period. They report ~$38 revenue/GPU-hour on V4.1-Flash clusters.

Inference is clearly profitable. The question is whether it's profitable enough to justify the training research costs.

▲

scheme271 37 minutes ago | parent | next [-]

Are those statements audited or fact checked? Inference profitability is great but how much of that profit gets eaten up paying back training costs, R&D, and operating expenses.

▲

javier2 5 hours ago | parent | prev | next [-]

At best, this tells you v4.1 flash is profitable, but can they really get $38 per gpu hour? Sounds like a lot.

	▲	riknos314 an hour ago \| parent \| next [-]
		From their own inference framework docs: "A single cluster currently serves 155 trillion tokens per day, with overall revenue of US$38 per GPU-hour (at valley pricing)." https://github.com/deepseek-ai/DeepInfer/blob/main/doc/infer... So this implies they actually hit about double that during peak pricing.
	▲	mustyoshi 4 hours ago \| parent \| prev [-]
		If they keep it running at near constant usage, you'd only need 80-120 million tokens an hour to hit 38$ of revenue (at off peak pricing, half as much for peak pricing).

▲

margorczynski 4 hours ago | parent | prev | next [-]

V4.1-Flash is probably quite smaller and much more resource effective than the frontier models. Add to that much cheaper energy in China.

▲

calmworm 4 hours ago | parent | prev | next [-]

Clearly potentially profitable. Not clearly profitable.

▲

parineum 5 hours ago | parent | prev [-]

How much did it cost to train Deepseek v4.1 flash?

	▲	riknos314 an hour ago \| parent [-]
		Great question. Nobody knows. The cool thing about models designed for cheaper inference is that post-training (which often requires inference as a part of the feedback loop) becomes cheaper as well, so almost certainly cheaper than models that are less optimized.