DeepSeek’s reported $1bn run rate comes with a sharper price and compute story

DeepSeek has reportedly passed a $1bn annualised revenue run rate after raising API prices, while putting more than 70% of its computing capacity into training new models. The account suggests a shift from bargain-priced challenger to commercial operator, with a sizeable bet still being placed on what comes next.

DeepSeek Watch analysis

What happened

The Information, in a report carried by 36Kr on 28 September, says DeepSeek’s annualised revenue run rate has more than doubled from below $500m a few months ago. It attributes the figures to chief executive Liang Wenfeng speaking at an investor meeting. The report says DeepSeek raised model API prices by about 2.3 to 4.5 times in mid-August, and that Liang said demand remained strong without customer churn.

The report also says more than 70% of DeepSeek’s computing capacity is being used to train new models, leaving less than 30% for running existing services. The $1bn figure is a reported run rate, not audited revenue, and the detailed claims come through 36Kr’s account of The Information’s reporting.

Why it matters

The reported pricing change puts a useful question behind the headline growth: can DeepSeek charge more without losing customers who came for inexpensive models? If the account is right, demand has so far held up, while the compute split suggests the company is prioritising future model development over maximising present-day inference capacity.

That is a consequential mix for customers and competitors alike. Higher API prices may change the economics of building on DeepSeek, while a training-heavy allocation could mean more capability later, but less headroom for serving current demand now.

Our read

This is more revealing than a revenue milestone on its own: it links reported growth to both pricing power and a deliberate resource choice. The figures are not audited, so treat them as an attributed commercial snapshot, not settled accounts. Still, if customers really absorbed a steep price rise without leaving, DeepSeek’s low-cost reputation may be evolving into something more complicated.

What to watch

• Whether DeepSeek’s API prices and customer demand hold at the reported levels.
• Whether the training-heavy compute allocation produces new models customers want to use.
• Whether later filings or reporting substantiate the run-rate and investor-meeting figures.

Discussion spark: If DeepSeek’s reported figures are accurate, is raising API prices while reserving most compute for training a sensible long-term bet, or a risk to the customers who helped it grow?

not affiliated with or endorsed by DeepSeek