Unverified50% confidenceFactExact time
Converted according to official figures, DeepInfer generates $38 in revenue per card per hour, while B200 market rental is about $8, a difference of more than four times, but this is revenue versus rental and does not deduct electricity, depreciation, storage and other costs.
1
Sources
50%
Confidence
Short-term (~14 days)
Relevance
10/31/2026
First Seen
Valid until: 11/14/2026
Sources
DeepSeek Open-Sources DeepInfer: The Cost-Saving Tricks Are There, the Scheduling Isn't
bilibili硬哥拆解10/31/2026
Related Entities
Related Claims
UnverifiedOutsiders speculate that DeepInfer's average revenue per GPU per hour may be around $46, corresponding to annual revenue of about $400,000.78% similarUnverifiedCalculated at off-peak pricing, the DeepSeek cluster earns an average of $38 per card per hour.76% similarUnverifiedDeepSeek responded that the $38 per GPU per hour is a theoretical revenue ceiling calculated based on running at full load 24 hours a day at the stock price, rather than an actual profit margin.68% similarUnverifiedIn the inference system description DeepSeek published in February 2025, the cost was calculated as follows: an average of 226.75 nodes running, rented at 2 US dollars per card per hour, for a cost of 87,000 US dollars per day; billed at R1's price at the time, revenue was 562,000 US dollars per day, a cost profit margin of 545%.68% similarUnverifiedDeepSeek open-sourced an inference framework on October 20, and the accompanying documentation gives a guaranteed minimum price of 38 US dollars per hour for a single card.66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1120181API
curl https://kongchang.com/api/v1/knowledge/claims/1120181MCP
get_claim(id=1120181)