DeepSeek hits $1B revenue, eyes $74B raise

DeepSeek, the Chinese AI lab that shocked the industry with its open-weight models, has doubled its annualized revenue to $1 billion and is now raising roughly $7.45 billion at a $74 billion valuation — according to reporting from The Information on September 24, 2026 (UTC). The round, expected to close by the end of October, would be one of the largest AI funding rounds of the year and positions DeepSeek as the most valuable AI company outside the United States.

The numbers

The Hangzhou-based lab is targeting a 50 billion yuan (~$7.45B) Series B at a 500 billion yuan (~$74B) pre-money valuation, per multiple sources familiar with the plans. Existing backers and investors who missed the first round are expected to participate. CITIC Securities has been hired to lead a Shanghai STAR Market IPO, with a filing targeted for later this year.

What makes the round notable is the revenue trajectory. DeepSeek's annualized run rate has surged from under $500 million just a few months ago to $1 billion, driven by API price increases and sustained demand for its large language models. That puts its price-to-sales ratio at roughly 74x — rich by any standard, but not out of line with AI peers.

Sources: The Information, company disclosures, analyst estimates. Figures are approximate.

Why it matters

A $74 billion valuation for DeepSeek is a statement. It says investors believe a Chinese lab can compete at the frontier — not just as a cheap alternative, but as a genuine peer to OpenAI and Anthropic. DeepSeek's open-weight strategy has already forced the entire industry to lower prices; its R1 reasoning model set a benchmark that OpenAI and Google had to answer. Now it's converting that technical credibility into commercial momentum.

The revenue doubling is the more interesting number. Going from under $500M to $1B in a few months means DeepSeek's API is being adopted at a pace that rivals the early growth curve of OpenAI's platform. Part of that is price increases — DeepSeek raised API rates earlier this year after years of undercutting competitors — but part is genuine demand. Enterprises that started with DeepSeek as a cost-saving experiment are now building production workloads on it.

The 74x P/S ratio deserves a skeptical look. It's higher than OpenAI's and Anthropic's, which means investors are paying a premium for growth rather than current cash flow. That's a bet that DeepSeek can keep doubling revenue every few quarters. If API demand cools or competition from Qwen, Kimi, and domestic rivals intensifies, that multiple compresses fast.

There's also the geopolitical angle. DeepSeek is raising and listing in China at a moment when U.S. export controls on AI chips are tightening and the White House is restricting model sharing with allies. A $74B DeepSeek with a Shanghai IPO is a signal that China's AI ecosystem can fund itself without Western capital. Whether it can keep pace on compute — the real bottleneck — is the open question. DeepSeek has been creative with efficiency, but frontier training runs still demand enormous GPU clusters, and access to the latest NVIDIA hardware is constrained.

What to watch

The Series B close in October will tell us whether the $74B valuation holds or gets negotiated down. Watch for the investor list — if major Chinese tech firms (Tencent, Alibaba, ByteDance) participate, it validates the round; if it's mostly financial investors, that's a softer signal. The STAR Market filing, expected later this year, will reveal actual revenue and profit numbers for the first time, which will either confirm the $1B run rate or expose it as optimistic.

The bigger test is whether DeepSeek can maintain its technical edge while scaling commercially. Its next model release — expected before year-end — will show whether the lab can keep pace with GPT-6, Claude Opus 5.5, and Gemini 4 on capability, not just price. If it can, $74B starts to look cheap. If it slips, the round becomes a very expensive entry point.