DeepSeek’s Revenue Surged Tenfold, but Its Losses Remained Larger Than Sales

DeepSeek’s reported financial picture is a study in contrasts: revenue reached about 475 million yuan ($70.7 million) in the first seven months of 2026, yet the company remained deeply loss-making. At the same time, its API model-access business posted an 82.9% gross margin, suggesting that the economics of serving paid customers may be stronger than the company’s bottom line implies. 47

The key figures from January through July 2026

The Information reported that DeepSeek generated approximately 475 million yuan in revenue between January and July 2026. That was roughly 10 times the company’s full-year 2025 revenue—a striking comparison, although it compares seven months of 2026 with all of 2025 rather than equivalent reporting periods. 47

DeepSeek also recorded a net loss of about 715 million yuan during those seven months. The figure was below its 935 million yuan net loss for the whole of 2025, but the periods are not directly comparable: the 2026 result covers only seven months, while the 2025 figure covers an entire year. 17

Its reported margins were stronger than the loss figure might suggest:

• Overall gross margin: 44.6%
• API model-access services gross margin: 82.9% 47

Gross margin measures what remains after direct costs of delivering a product or service; it does not include every expense involved in running a company. A high API margin can therefore coexist with a large net loss if research, infrastructure, staffing and other operating costs are substantial.

Why strong API margins did not produce a profit

The gap between DeepSeek’s API margin and its net result highlights the capital intensity of frontier AI. The available reporting says DeepSeek spent about 11 billion yuan on AI infrastructure in the first seven months of 2026, compared with about 1.2 billion yuan in 2025. Those costs reportedly included servers equipped with AI chips, chips and other computing equipment. 32

That spending helps explain why an 82.9% gross margin on API services did not translate into company-wide profitability. API revenue may be attractive on a per-customer or per-call basis, but the business still has to finance model development, capacity expansion and the infrastructure needed to support growth.

The figures also suggest that DeepSeek’s commercial challenge is not simply whether it can sell access to its models. It is whether revenue can scale quickly enough to offset the enormous upfront and ongoing cost of building AI capacity.

DeepSeek versus OpenAI and Anthropic

DeepSeek’s revenue remained far smaller than that of the leading U.S. AI companies cited in the comparison. OpenAI reportedly generated $5.7 billion in first-quarter revenue at a 39% gross margin. Anthropic reportedly generated $11.5 billion in second-quarter revenue. 7

The comparison needs to be treated carefully because the companies’ reporting periods and margin definitions may differ. Still, the broad picture is clear: DeepSeek was growing rapidly, but its reported revenue was nowhere near the scale of OpenAI or Anthropic. Its advantage in the figures supplied by The Information was not revenue volume; it was the high margin attached specifically to API model access. 47

The available evidence does not provide a sufficiently comparable Anthropic gross-margin figure for the same period. That makes a precise numerical comparison between DeepSeek’s 44.6% overall margin and Anthropic’s business unreliable.

The $400 million–$500 million annualized run rate

The Information had separately estimated DeepSeek’s annualized revenue run rate at between $400 million and $500 million, with revenue coming primarily from enterprises and developers accessing its models through APIs. 3

An annualized run rate is a projection based on a recent revenue pace. It is not the same as revenue already recognized over a completed fiscal year, so it should not be added to the 475 million yuan reported for January through July. The two figures describe different things: one is a reported period total, while the other extrapolates recent performance across a full year.

Funding and potential IPO plans

The reported growth arrived as DeepSeek was pursuing another major capital raise. The company was reportedly seeking 50 billion yuan in a second funding round at a target valuation of 500 billion yuan, or approximately $74 billion. That followed a first round of more than 50 billion yuan completed in June, according to reporting from The Information. 135

The second round was described as pre-IPO financing, with a potential listing in Shanghai next year. 35 The fundraising plan reflects both sides of DeepSeek’s financial profile: fast-growing commercial demand and a continuing need for substantial capital to fund AI infrastructure.

What the numbers mean

DeepSeek’s reported results do not show a profitable AI company yet. They show a company with three more specific characteristics:

1. Rapid commercial expansion: 475 million yuan in revenue in seven months, roughly 10 times its full-year 2025 revenue. 47
2. Promising service-level economics: an 82.9% gross margin for API model-access services and a 44.6% overall gross margin. 47
3. Heavy capital requirements: a 715 million yuan net loss and reported infrastructure spending of about 11 billion yuan in the first seven months. 132

The central takeaway is that DeepSeek may have found a comparatively efficient way to monetize API access, but efficiency at the model-serving level has not yet overcome the broader cost of scaling an AI laboratory. Its planned fundraising and potential IPO therefore appear tied not only to growth ambitions, but also to the funding required to sustain that growth.