DeepSeek Hits $1B ARR, Targets $70B Raise and STAR Market IPO

According to The Information, DeepSeek founder Liang Wenfeng told investors that the company's annualized revenue run rate has reached $1 billion, more than doubling in a few months (previously under $500 million). It is finalizing a new funding round of RMB 50 billion (about $7.5 billion) to be completed by the end of October at the latest, targeting a valuation of around RMB 500 billion, and is preparing to list on the STAR Market. This is the first time DeepSeek has been reported to have commercialization and capital plans of this scale, making it worth watching.

Confirmed

Annualized revenue run rate reached $1 billion, up from under $500 million a few months ago; Liang Wenfeng attributed the growth to last month's API price increase (2.3x to 4.5x the original price) and said it did not cause customer churn
A second funding round of RMB 50 billion (about $7.5 billion) is in progress, targeted to close by the end of October at the latest, with a target valuation of around RMB 500 billion, alongside preparations for a STAR Market IPO
Of the roughly $1 billion compute budget, 70% goes to training and 30% to inference
To ease compute shortages, DeepSeek is testing aggressive model optimizations to shift part of the inference workload of smaller models onto much cheaper NVIDIA gaming GPUs; Liang Wenfeng told investors that internal tests show the small models run well on gaming cards

Why it matters

$1 billion ARR places DeepSeek among the fastest-monetizing foundation model companies, and if its claim of no customer churn after the price hike holds, it signals strong demand inelasticity
Running small-model inference on gaming GPUs, if validated, could substantially cut deployment costs and offer a new approach to scaling inference under compute constraints; several posters (e.g., teortaxesTex, rohanpaulai) see it as a pragmatic response to compute shortages
The RMB 50 billion raise and STAR Market IPO plan mean DeepSeek is evolving from a High-Flyer-affiliated project into an independent capitalized entity, and its valuation trajectory will be a bellwether for domestic foundation model companies

Not yet confirmed

All of the above comes from The Information's reporting and Liang Wenfeng's statements to investors; the revenue, valuation, and IPO timeline lack corroboration from official public documents

2026-09-24 ~ 2026-09-24 · 7 related posts

Primary sources

DeepSeek hits $1B annualized revenue, raises another $7B at ~$70B valuation — 智东西 · 2026-09-24 16:43 (UTC+08:00)
DeepSeek tests running smaller models on cheap gaming GPUs, per The Information — rohanpaul_ai · 2026-09-24 17:38 (UTC+08:00)
DeepSeek hits $1B ARR, splits compute 70/30 and runs inference on gaming GPUs — teortaxesTex · 2026-09-24 21:26 (UTC+08:00)

[source] DeepSeek hits $1B annualized revenue, raises another $7B at ~$70B valuation — 智东西 · 2026-09-24
DeepSeek testing gaming GPUs to run smaller models' inference, per The Information — rohanpaul_ai · 2026-09-24
DeepSeek hits $1B annualized revenue, targets $7.5B raise at $70B valuation — 快鲤鱼 · 2026-09-24
[source] DeepSeek hits $1B ARR, splits compute 70/30 and runs inference on gaming GPUs — teortaxesTex · 2026-09-24
DeepSeek reportedly runs inference on NVIDIA gaming GPUs, 70% of $1B budget goes to training — teortaxesTex · 2026-09-24
Industry token usage tops 30Q/month; $1B ARR implies Anthropic-tier throughput — teortaxesTex · 2026-09-24

1 near-duplicate retellings: rohanpaul_ai