DeepSeek Raised $7.4 Billion Last Month at $50 Billion. Now It Wants $1.5 Billion More and a 2027 IPO
Weeks after closing one of the largest private AI rounds ever, DeepSeek is back - plotting more capital at a $71 billion valuation and setting its sights on going public.
DeepSeek funding rounds are arriving faster than its model releases. Bloomberg reported Monday that the Chinese AI lab is in preliminary talks to raise around $1.5 billion in new capital at roughly a $71 billion pre-money valuation - this, just weeks after closing a $7.4 billion round in June at a $50 billion valuation that ranked among the largest private AI financings ever recorded. A 2027 IPO is also under consideration, Bloomberg said, with some sources suggesting an earlier debut at the end of this year.
No deal has been signed. Talks are early stage. But the speed of the follow-on signals something more pressing than standard capital recycling - DeepSeek is moving fast toward building its own data center infrastructure and acquiring more AI chips, a need that existing cash may not fully cover.
The June Round Came With Terms Investors Rarely Accept
DeepSeek's first-ever outside funding closed on terms that would have stopped most Western investors cold. Rather than taking equity directly in the company, investors placed their funds into a limited partnership managed by DeepSeek founder and CEO Liang Wenfeng - the same founder who runs High-Flyer, one of China's leading quantitative hedge funds. Liang personally committed around $2.94 billion to the round. Tencent led with approximately $1.5 billion; battery giant CATL put in about $735 million.
Investors accepted a five-year lock-up on their capital and agreed to receive no voting rights. One exception: China's National Artificial Intelligence Industry Investment Fund, the $8 billion state-backed vehicle, kept direct equity and voting privileges. Alibaba also participated. For a company that built its model from scratch on Huawei chips - without NVIDIA hardware, without external funding, without the cloud subsidies OpenAI and Anthropic relied on early - accepting outside capital at all was already a departure from the prior playbook. Doing it on these terms was a statement about who controls the company.
DeepSeek Is Processing 23% of Enterprise AI Traffic - and Still Growing
Investors took those terms anyway because the underlying numbers are hard to argue with. In June 2026, DeepSeek accounted for 23% of all tokens processed through Vercel's AI gateway - a platform used by enterprises to route requests across AI providers. Anthropic held 32%. OpenAI, which once dominated this category, was behind both. DeepSeek hit that position running open-weight models on Huawei infrastructure, in a country under US chip export controls that were designed specifically to limit Chinese AI compute capacity.
Open-source AI inference at scale is a capital-intensive business - Together AI raised $800 million at $8.3 billion just to run other companies' open models at enterprise grade. DeepSeek runs its own models, serves its own traffic, and wants to own its own hardware. The $71 billion valuation being discussed for the new round reflects that expanded ambition, not just the model quality.
A 2027 IPO Would Put DeepSeek on the Same Timeline as OpenAI
DeepSeek isn't alone in targeting 2027 for a public market debut. OpenAI pushed its own IPO to 2027 after SoftBank's bridge loan structure made an earlier listing complicated. A DeepSeek IPO on the same timeline would create a direct comparison investors cannot avoid: two frontier AI labs with open-ended valuations, opposing chip strategies, and competing model philosophies hitting public markets in the same window.
No listing venue, no underwriter, and no formal filing have been announced. Bloomberg's sourcing describes the IPO as a direction, not a decision. Still, the pattern is consistent - a company that operated without external funding for three years, then raised $7.4 billion in its first round, is now plotting a second raise within weeks and considering public markets within 18 months. DeepSeek funding is moving on a timeline that matches the urgency of the compute buildout it needs to execute, not the leisure of a company that already has everything in place.