Chinese AI lab DeepSeek raised $7.4 billion at a $50B+ valuation, routing investor capital through a founder-run limited partnership with no voting rights attached.
DeepSeek's first external funding round has closed at more than $7.40 billion, according to a report by The Information published Tuesday. The figure values the Hangzhou-based lab above $50 billion and instantly places this among the largest single fundraises in artificial intelligence history.
The deal structure is as striking as the number. Emirates247 reported that investors did not place capital directly into DeepSeek itself. Instead, money flowed into a limited partnership controlled by CEO and founder Liang Wenfeng. Most participants accepted a five-year lock-up and received zero voting rights in the operating company.
Reuters could not immediately verify the terms, and DeepSeek did not respond to requests for comment.
The deal participants
Liang himself has pledged 20 billion yuan of his personal wealth to the round, Reuters reported earlier this month. Tech conglomerate Tencent is considering a 10 billion yuan contribution; battery and materials giant CATL is weighing 5 billion yuan. Both would become the largest external investors if their participation is confirmed.
China's National Artificial Intelligence Industry Investment Fund received sharply different terms. It invested directly into DeepSeek rather than through the limited partnership, retaining voting rights and freedom from the lock-up that applies to everyone else. That single carve-out for a state vehicle tells you something about how Beijing positions itself relative to its national artificial intelligence champion.
Why the structure matters
Founder-control mechanisms have become routine in Silicon Valley. Dual-class share arrangements at Alphabet and Meta allowed their founders to retain decision-making authority while taking on outside capital. Liang's approach takes that logic further. Most investors here are not receiving equity in DeepSeek at all. They hold interests in a vehicle Liang manages at his discretion, with no board representation and no exit window for half a decade.
Private markets have grown accustomed to aggressive founder-friendly terms when the underlying asset is hot enough. But the combination of conditions in this round — no voting rights, multi-year lock-up, no direct equity — begins to resemble a structured debt instrument more than a conventional venture investment. The $50 billion valuation is what sophisticated capital agreed to pay under those constraints, not a figure validated by a public market or competitive bidding between investors with full information.
The state fund's exemption from every one of those restrictions deserves a separate read. While Liang retains operational control, Beijing has preserved the ability to exit and to vote on decisions that no private investor can influence. As the United States debates a federal AI preemption bill that would override state-level rules — a fight The Verge described this week as Big Tech's last-ditch attempt before potentially hostile midterm elections shift the balance in Congress — China has already settled the question of government involvement by embedding it directly into the cap table.
How DeepSeek got here
The lab earned its reputation in early 2025, when its V3 and R1 models drew genuine admiration from engineers across Silicon Valley and unsettled assumptions about China's artificial intelligence capabilities. Working under U.S. chip export restrictions, DeepSeek's team produced results that demonstrated compute efficiency could partially substitute for raw hardware access. The episode prompted a broad reassessment of what frontier AI development actually requires.
DeepSeek's round sits inside a wider acceleration of AI capital deployment. Arcade AI, an agent authorization platform, closed a $60 million Series A round led by SYN Ventures, SiliconAngle reported Monday. Sarvam AI, an Indian language model startup, crossed unicorn status after HCLTech acquired a 10.5 percent stake for $150.7 million, Goodreturns reported Tuesday. The appetite for AI exposure spans geographies and deal sizes, with no obvious ceiling in sight.
Investors who accepted a five-year lock-up and no vote are betting Liang delivers the next leap in model performance. They have no mechanism to revisit that bet if he does not.
Frequently asked questions
What is DeepSeek's current valuation?
DeepSeek's first outside funding round, reported in June 2026, values the company at more than $50 billion based on a 50 billion yuan raise. No prior external valuation existed.
Who are DeepSeek's investors?
Founder Liang Wenfeng committed 20 billion yuan personally. Tencent and CATL are considering stakes of 10 billion yuan and 5 billion yuan respectively. China's National AI Industry Investment Fund holds a direct equity stake with voting rights.
What is a limited partnership structure in venture capital?
A limited partnership pools investor capital under a general partner who makes all investment decisions. Limited partners provide money but have no management authority. In DeepSeek's case, Liang Wenfeng acts as general partner, meaning outside investors have no formal say in how the company operates.
How did DeepSeek become globally known?
DeepSeek released its V3 and R1 models in early 2025. Both achieved competitive benchmark results despite being built under U.S. semiconductor export restrictions, which surprised many Western AI researchers and triggered a reassessment of the U.S.-China gap in artificial intelligence development.
