DeepSeek’s annualized revenue run rate has reached $1 billion, more than double the under $500 million pace the Chinese AI lab was posting a few months ago, The Information reported on September 24. Two people familiar with the lab’s finances gave the outlet the figure, and DeepSeek has not commented.
The jump follows a steep increase to DeepSeek’s API prices. The Hangzhou based lab, known for open weight models priced well below Western rivals, warned developers in August that fees were going up, a move the South China Morning Post covered at the time. Depending on the model, The Information reports, the increase ran as low as 2.3 times the old price and as high as 4.5 times it. In a pitch to investors, DeepSeek’s founder Liang Wenfeng said customers stayed anyway.
That is the part worth sitting with. A price increase of that size usually pushes some users toward cheaper alternatives, especially when open weight competitors exist. DeepSeek helped create that competitive pressure in the first place by pricing aggressively when it launched. Developers stuck around even at the top end of that range, which suggests its models have become more of a default choice than a discount option.
A run rate is a projection: it takes a recent stretch of earnings and stretches that pace out over twelve months, so it tracks momentum rather than cash already collected. DeepSeek pulled in about 475 million yuan, or $70.7 million, over January through July of 2026, per an earlier Information report, which put the company at roughly ten times what it earned in all of 2025. Developers paying to reach DeepSeek’s models through its API supply nearly all of that money. The consumer chatbot remains free.
DeepSeek’s second funding round is moving toward a decision at the same time this revenue number surfaced. The Information says the company is finalizing a $7.5 billion raise, which Bloomberg has put at a valuation near 500 billion yuan, or $74 billion. The round has not gone smoothly: DeepSeek froze it in July once private comments from Liang spread widely across Chinese social platforms, and only picked talks back up in August. Its first outside round, roughly $7 billion, closed in June. Liang himself covered a large share of that money.
DeepSeek only began taking outside capital this year, after building its reputation entirely on aggressively priced open weight models. A valuation near $74 billion would put the lab alongside Western rivals that have spent years raising outside money to fund training compute. DeepSeek is reportedly doing it on revenue that only recently crossed $1 billion annualized, and largely without the subscription tiers that OpenAI and Anthropic depend on.
For any company competing with DeepSeek on API pricing, the read is uncomfortable. The lab tested how much its developer base would tolerate, and the ceiling turned out higher than a straightforward pricing model would have predicted. Rivals banking on price sensitivity to win developers away from DeepSeek now have less evidence that it works.
Reported by The Next Web, citing The Information, on 24 September 2026.
