Written by Admin Alex · Fact-Checked by M.Ali · Info Verified September 2026
We review and update this article regularly as new information becomes available.
TL;DR: DeepSeek’s annualized revenue run rate has crossed $1 billion, more than double its pace from a few months ago, as the Chinese AI lab tries to close a $7.5 billion funding round at a $74 billion valuation by the end of October. Some of that growth came from raising its own API prices, and the company is still losing money.
One billion dollars a year. That’s where DeepSeek’s annualized revenue run rate landed as of late September 2026, and it got there fast. A few months ago, the Chinese AI lab was running under $500 million annualized. Now it’s more than double that, right as the company works to close a funding round that would value it at $74 billion.
The underlying numbers back up the headline figure, even if they need some unpacking. DeepSeek pulled in roughly $70.7 million (475 million yuan) in revenue across the first seven months of 2026. For comparison, its entire 2025 revenue was about $7 million. That’s a tenfold jump packed into a matter of months, not years.
Where the growth actually came from
Some of this is genuine demand. Developers are building on DeepSeek’s models in real volume, and the API business is profitable enough at the gross margin line to post roughly 82.9%. That’s a strong number for anyone selling inference at scale.
But a meaningful piece of the jump has a much simpler explanation. In August 2026, DeepSeek raised its own API prices, by 2.3x on the low end and up to 4.5x depending on the model. Charge two to four times more for the same requests and revenue climbs even if usage barely moves. Nobody has published a clean split of how much of the run rate increase is new developers versus the same developers paying more for what they were already using. That distinction matters a lot if you’re trying to judge whether this growth is durable or a one-time repricing bump working its way through the numbers.
Still losing money
Despite the jump, DeepSeek reported a net loss for the first seven months of 2026. Training frontier models and serving inference at scale is expensive, and a nine-figure run rate doesn’t automatically cover that cost. Worth noting: the consumer chatbot most people associate with the DeepSeek name is still free. Every dollar of this revenue is coming from developers paying for API access, not from consumer subscriptions or ads.
The bigger story is the funding round
These revenue figures are surfacing now for a reason. DeepSeek is in the middle of raising money, targeting $7.5 billion (50 billion yuan) at a $74 billion valuation (500 billion yuan), and it’s aiming to close the round by the end of October 2026. If it lands anywhere close to that number, it marks a sharp step up from where things stood just months earlier. Tencent and CATL led a $7.4 billion round back in June 2026, and this new target is meaningfully larger and richer on valuation.
DeepSeek hasn’t always courted institutional capital at this scale. Founder Liang Wenfeng built the company off the back of High-Flyer, the quantitative trading hedge fund he ran before turning to AI research. For a long stretch, High-Flyer’s own money was effectively DeepSeek’s funding source. A $74 billion valuation, if the round actually closes at that level, would mark a very different chapter for a company that started as an offshoot of a hedge fund’s compute budget.
Also today: Lovable crossed $600 million in annualized revenue on the AI coding side, and Oura’s $2.2 billion IPO shows the same pattern of massive valuations chasing AI-era growth.
Bottom Line: A $1 billion run rate is a real milestone for a company that made $7 million total last year, but anyone writing a check at a $74 billion valuation should look past the headline number. A meaningful slice of that growth came from price hikes rather than new demand, and DeepSeek is still losing money getting there.
