DeepSeek Raises $7.4B at $59B Valuation, Six-Fold Jump in Two Months
DeepSeek closed its first-ever external funding round on June 3, taking $7.4 billion from Tencent, CATL, NetEase, JD.com and founder Liang Wenfeng at a valuation just under $59 billion. That is a six-fold mark-up from the $10 billion April figure. The lab spent years refusing outside capital; this round signals the end of that posture and the formal arrival of China's AI champion on the global capital map.
The cap-table reads like a Beijing strategy memo rather than a venture round. Tencent is in for 10 billion yuan, battery giant CATL for 5 billion, NetEase and JD.com for 3 billion each, with IDG Capital, Monolith, Loyal Valley and Shixiang filling the rest. Founder Liang Wenfeng is personally writing a 20 billion yuan check, roughly $2.9 billion, which means he is buying down dilution rather than cashing out. That is the inverse of the Silicon Valley founder-secondary pattern, and it is the tell: DeepSeek is being capitalized to keep operating, not to mint paper wealth.
The valuation gap is the part that should make Western investors uncomfortable. At $59 billion DeepSeek is being marked at roughly half of Anthropic's $30 billion Q1 round on revenue that is a fraction of Anthropic's, but the lab also spends a fraction of what Anthropic spends. The Chinese AI thesis has always been efficiency-first: R1 and V3 shipped on training budgets that frontier U.S. labs would consider a rounding error. If you believe the open-weights releases were honest about their compute, $59 billion is cheap. If you think they undercounted, it is still cheap, because the U.S. comparables priced in $500 billion infrastructure commitments that DeepSeek does not need to make.
The strategic read is that DeepSeek finally took the money because the cost of staying private had become higher than the cost of taking state-adjacent capital. Tencent and CATL are not passive LPs; they bring distribution, manufacturing supply chains, and political cover. That matters in a week when the White House signed a new advanced-AI executive order and Nvidia used Computex to position itself as the PC layer of the U.S. stack. For more on how the supply-chain side is consolidating, see our earlier coverage in Nvidia's RTX Spark Superchip push. DeepSeek raising at $59 billion the same week Microsoft shipped MAI-Code-1-Flash to lessen its OpenAI dependency is not a coincidence; it is the market pricing in a multi-pole AI industry where each region funds its own full stack.
What to watch next: whether any of this capital flows into a hosted inference business. DeepSeek's open-weights strategy has been generous to the point of self-harm commercially, and $7.4 billion does not make sense if the plan is to keep giving the models away. Expect a managed API tier, expect enterprise contracts inside the Tencent and JD ecosystems, and expect the next twelve months to test whether "efficient training" is a durable moat or a one-cycle advantage that the U.S. hyperscalers close once their next-generation silicon ships.
Bloomberg — DeepSeek Close to Sealing $7 Billion Funding → · South China Morning Post — DeepSeek nears US$7b haul in first-ever funding round → · China Money Network — DeepSeek Prepares to Raise $7.4 Billion at $59B Valuation →