DeepSeek Targets $59 Billion in First-Ever Raise: What China’s Cheapest AI Still Cannot Match

DeepSeek, the Hangzhou artificial intelligence company that rattled global markets in early 2025, is preparing to raise about $7.4 billion in its first external funding round — a deal that could value the firm at as much as $59 billion and rank among the largest private technology financings ever arranged in China. According to reporting aggregated by Yahoo Finance, Tencent is weighing roughly 10 billion yuan and battery maker CATL about 5 billion yuan, positioning the two as the round's largest outside backers. For anyone who uses or pays for AI, the raise signals that the industry's most aggressive price-cutter now has the capital to keep cutting — though its models still trail the best Western systems in ways a buyer should weigh.

The financing pulls in a notable roster of co-investors. TechStartups reported that NetEase, JD.com, Hong Kong-based IDG Capital and Monolith Capital, and several state-backed Chinese AI funds are among the prospective participants. Founder Liang Wenfeng is reportedly contributing 20 billion yuan of his own capital — a striking commitment from a founder who has financed the company largely through his quantitative hedge fund, High-Flyer.

A company built on giving models away now wants a war chest

DeepSeek earned its name on open-weight releases that matched far more expensive Western systems. Its R1 reasoning model, released in January 2025, triggered a sharp sell-off in U.S. AI stocks after it delivered frontier-level results at a fraction of the reported training cost. That is what makes the new raise pointed: a firm famous for doing more with less is now assembling one of the larger balance sheets in the field.

The money appears aimed at the one input efficiency cannot conjure — compute. Serving large models at scale demands high-end accelerators, and U.S. export controls have throttled the flow of the most advanced chips into China. Billions in fresh capital lets DeepSeek secure domestic silicon, stockpile what it can obtain, and fund its next models without leaning on Liang's personal fortune.

How does DeepSeek's valuation compare to OpenAI and Anthropic?

The round also exposes how differently the two markets price the same capability. As Proactive Investors noted, DeepSeek's $59 billion target is a fraction of the figures attached to U.S. peers — Anthropic recently closed near a $965 billion valuation and OpenAI has been valued in the hundreds of billions — despite DeepSeek's competitiveness on many public benchmarks. The contrast is less a verdict on quality than a measure of two financing cultures.

Independent testers still flag a real performance gap

DeepSeek's published benchmark numbers are strong, but they are not uniformly confirmed by outside evaluators. Western researchers running their own test suites have generally found the models capable on reasoning and coding while noting that real-world reliability, tool use, and long-context stability lag the best closed systems from OpenAI, Anthropic, and Google. Any company's self-reported benchmarks deserve skepticism until a named third party reproduces them, and that caution applies with particular force where outside auditors cannot fully replicate the conditions.

The ecosystem gap compounds the picture. International developers who adopt DeepSeek's open weights inherit integration work, thinner documentation, and a smaller third-party tooling community than the dominant U.S. platforms offer. For a hobbyist those are minor frictions; for an enterprise they become engineering time and support risk that erode the savings.

China's National Intelligence Law is a fixed condition, not a footnote

For anyone weighing DeepSeek's hosted services, one factor holds regardless of price or performance. Under China's National Intelligence Law, Chinese companies are legally required to cooperate with state intelligence requests. That obligation attaches to data handled by a China-headquartered firm no matter its privacy policy or where individual servers sit. Running the open weights locally keeps data off DeepSeek's systems; using its cloud API does not. The distinction is decisive for sensitive or regulated information.

What the raise changes for the rest of the market

If it closes, the round hardens DeepSeek's standing as China's leading independent AI lab and prolongs the price pressure its cheap, capable models already apply across the industry. Low-cost open-weight alternatives give developers leverage against incumbent subscription and API pricing, and a better-capitalized DeepSeek can sustain that leverage longer. The terms remain subject to change, and as of early June 2026 DeepSeek, Tencent, and CATL had not publicly confirmed the details. People familiar with the talks say the round could close within weeks.

The honest takeaway is not that a cheaper AI is simply getting bigger. It is that the cheapest credible AI now has the balance sheet to keep competing — while still carrying measurable performance gaps, a thinner ecosystem, and a state data-sharing obligation that any buyer must weigh against whatever the lower price saves.

Frequently Asked Questions

How much is DeepSeek raising and at what valuation?

DeepSeek is preparing to raise about $7.4 billion in its first external round, which people familiar with the talks say could value the company at as much as $59 billion. The terms are not final and had not been publicly confirmed by the company as of early June 2026.

Who is investing in the DeepSeek funding round?

Tencent is reportedly weighing roughly 10 billion yuan and CATL about 5 billion yuan, with NetEase, JD.com, IDG Capital, Monolith Capital, and state-backed Chinese AI funds also named as prospective backers. Founder Liang Wenfeng is said to be adding 20 billion yuan of his own money.

Is it safe to use DeepSeek for sensitive data?

Using DeepSeek's hosted cloud service sends your data to a China-headquartered company that is subject to China's National Intelligence Law, which compels cooperation with state intelligence requests. Running the open-weight models locally avoids transmitting data to the company, which is the safer path for regulated or confidential information.