Chinese artificial intelligence startup DeepSeek closed its first external funding round in mid-June. The haul topped $7.4 billion. That sum catapulted its valuation past $50 billion. Suddenly the company stood as China’s most valuable AI venture. Founder Liang Wenfeng poured in roughly $3 billion of his own capital. He now ranks among the world’s richest pure-play AI entrepreneurs.
But the deal carried an unusual twist. Investors did not purchase shares directly in DeepSeek. Their money flowed instead into a limited partnership under Liang’s control. A five-year lock-up applied. Voting rights stayed at zero. One exception existed. The state-backed National Artificial Intelligence Industry Investment Fund secured direct equity along with actual voting power and no lock-up. Forbes laid out these terms clearly.
The structure handed Liang ironclad command. Backers gained exposure yet surrendered influence and liquidity. Some observers question what that means for governance. Others see it as a pragmatic response to geopolitical pressure. Foreign capital stayed away. Domestic players filled the gap. Tencent weighed a 10 billion yuan commitment. CATL eyed 5 billion yuan. NetEase, JD.com, IDG Capital and Monolith Management also joined discussions. Reuters reported those figures in early June.
DeepSeek had shunned outside money for years. Its parent, the quant hedge fund High-Flyer Capital Management, funded early work. Liang built the operation on that foundation. Then the models took off. V3 and R1 arrived in early 2025. They delivered performance that rivaled top American systems. Costs ran far lower. Praise poured in from Silicon Valley. U.S. assumptions about Chinese AI talent cracked. The company earned its status as national champion.
April brought V4. DeepSeek called it a redefinition of the state of the art for open-source agentic systems. Independent benchmarks tempered the claim. The model trailed certain U.S. and Chinese rivals on specific tasks. Still, adoption surged. Developers downloaded the weights. Enterprises tested them in production. Efficiency became the selling point. So did the price. Many organizations now route at least some workloads to DeepSeek. One Bay Area CTO told contacts the shift had moved beyond experiments.
Just weeks after closing that record round, DeepSeek eyed more capital. Talks opened with fresh investors. The pre-money target sat near $71 billion. Preparations for an initial public offering accelerated. Filing could come before year-end. A listing on Shanghai’s A-share market might follow in 2027. Yahoo Finance, citing the Financial Times and Bloomberg News, broke those details on July 14.
The speed surprises even hardened analysts. One month. Two capital raises. An IPO path sketched out. DeepSeek appears determined to amass resources before export controls tighten further or domestic competition intensifies. ByteDance’s Doubao and Alibaba’s Qwen press from above. Smaller labs chase from below. Scale now decides survival.
Hardware forms the next battleground. DeepSeek runs on Nvidia and Huawei chips today. Both face constraints. U.S. sanctions limit Nvidia’s newest offerings in China. Huawei’s Ascend lineup carries its own shortages and performance gaps. So the startup has begun designing its own silicon. The focus sits on inference. That stage, where trained models answer user queries, consumes growing compute demand as applications proliferate.
Recruiters have pulled in chip engineers at a quickened pace. Partnerships with foundries and memory suppliers sit under discussion. Success would shrink dependence on outside vendors. It would also hand DeepSeek tighter control over cost and performance. Reuters revealed the project on July 7. Bloomberg News echoed the account the same day.
Vertical integration marks a strategic pivot. Models alone no longer suffice. Data centers, power infrastructure, and custom silicon now command equal attention. CATL, one of DeepSeek’s new backers, already expands into AI energy storage and data-center equipment. Tencent seeks to bolster its own Hunyuan model while trailing market leaders. The investor list reads like a who’s who of China’s drive toward technological self-reliance.
Liang’s personal wealth swelled to roughly $36 billion after the June close. That figure placed him 63rd on the Bloomberg Billionaires Index. His stake reportedly approaches 84 percent when indirect holdings factor in. The founder now outranks several prominent American AI leaders in net worth. Yet questions linger about how state ties could affect global operations. Certain countries already restrict access to DeepSeek models over data-security worries.
Comparisons with U.S. counterparts highlight the gap. Anthropic closed a $65 billion round in late May at a $965 billion valuation. OpenAI raised $122 billion in March at $852 billion. DeepSeek’s $50 billion-plus post-money figure looks modest beside those sums. Still, the Chinese firm achieved its mark with far less prior capital and in a constrained funding environment. Efficiency, once again, defines its edge.
Market reaction on X reflected the momentum. Posts celebrated Liang as the world’s richest AI founder in the large-language-model category. Others noted production traffic moving to DeepSeek models inside American companies. A few flagged the $71 billion talks and IPO timeline as signals of infrastructure competition overtaking pure model development. The conversation moves fast. So does the company.
Executives have stayed silent on the latest overtures. DeepSeek, Tencent, CATL and the state fund declined comment to multiple outlets. That reticence fits a pattern. The firm rarely courts publicity. Its products speak instead. Open-source releases continue. Performance benchmarks appear. Enterprises integrate the technology. The capital simply accelerates what was already underway.
Challenges remain. Export controls could tighten. Talent wars rage across China’s tech sector. Training runs demand ever-larger clusters. If the inference chip succeeds, DeepSeek might carve out cost advantages that prove difficult to match. If it stumbles, reliance on Huawei or smuggled Nvidia parts could cap growth. The stakes sit high.
For now the trajectory points upward. From hedge-fund-backed research outfit to $50 billion AI leader in barely three years. A potential $71 billion round on deck. Silicon design in progress. An IPO horizon visible. DeepSeek has compressed timelines that once seemed impossible. Observers will watch whether the unusual ownership structure delivers the freedom Liang seeks or introduces frictions down the road. The answer could shape not only one company but the broader contours of China’s bid for AI supremacy.
WebProNews is an iEntry Publication