DeepSeek allegedly hit $1B ARR — and math says ~3,000 GPUs can carry it at off-peak prices
teortaxesTex · x · 2026-10-20
DeepSeek reportedly reached $1B ARR as of Sep 24. Back-of-envelope math using the lab's own production blog figures ($38 revenue per inference-card-hour, off-peak pricing): 42 GB200 NVL72 racks (3,000 Blackwell GPUs) could support $1B annualized revenue, serving 198 trillion total tokens/day including 1.83 trillion output tokens (2.29B tokens/sec instantaneous). Mixing peak billing would cut the requirement to 35 racks. At $18M per rack (including what the poster calls an absurd smuggling markup), payback takes about 10 months. The ledger highlights DeepSeek's extreme inference efficiency versus rivals running tens of thousands of GPUs.
Related event: DeepSeek Reportedly Hits $1B ARR, Needs Only ~3,000 GPUs(4 posts)→
More from Venture
- Dev who shipped 100+ apps lays out step-by-step path to building your first AI app — PrajwalTomar_ · 2026-10-20
- Two indie devs shipped 3 products in a year, totaling 36K users — cneuralnetwork · 2026-10-20
- TikTok was the fastest-growing top-10 website in Q3, growing 7.07% QoQ: Similarweb — gaganghotra_ · 2026-10-20
- AI virtual influencer hits 48M views and 53K followers in just two days — anthara_ai · 2026-10-20
- Character.AI Files for Bankruptcy, Sells Assets to Disney for $20M — menhguin · 2026-10-20
- AI chip startup Tenstorrent raises $1.5B at $6.2B valuation after takeover interest — pstAsiatech · 2026-10-20