For years, Liang Wenfeng was the high priest of Chinese AI’s monastic movement. His company, DeepSeek, operated more like a secluded research institute than a Silicon Valley disruptor, eschewing venture capital, ignoring KPIs, and letting researchers leave by 6 p.m. This 'romantic path' was funded by the windfall profits of his quantitative hedge fund, High-Flyer Quant. But as the global AI race accelerates into a war of attrition, Liang is finally trading his lab coat for a suit of armor.
DeepSeek is now reportedly seeking a staggering 50 billion yuan ($7.4 billion) in its first-ever external funding round, a move that would value the firm at upwards of $59 billion. In a move that underscores both his conviction and his desire for control, Liang has pledged 200 billion yuan of his personal wealth to the round. Industrial giants like Tencent and CATL are expected to lead the external investment, marking the end of DeepSeek’s era as a privately funded outlier in a landscape dominated by tech titans.
The shift comes as DeepSeek’s 'efficiency miracle' faces its harshest test yet. While the firm famously built world-class models like R1 and V3 at a fraction of the cost of OpenAI’s GPT series, it has become a prime target for poaching. Chinese giants like Xiaomi and ByteDance have reportedly lured away key researchers with compensation packages reaching nine figures. Without a formal equity structure and the liquidity of a massive valuation, Liang found himself unable to stem the talent drain that threatened to hollow out his research-first culture.
Beyond human capital, the physical infrastructure of the AI race has changed. DeepSeek is currently navigating a high-stakes migration from NVIDIA’s CUDA ecosystem to Huawei’s Ascend architecture. This 'sovereign AI' pivot is technically grueling and capital-intensive, with a single training run for the upcoming V4 model estimated to cost over $500 million. The move is no longer just about clever algorithms; it is about building a domestic full-stack ecosystem that can survive US export restrictions.
The competitive landscape has also grown more crowded. While DeepSeek was once the 'Price Butcher' of the industry, rivals like ByteDance’s Doubao and Alibaba’s Qwen have engaged in a brutal price war, matching DeepSeek’s low margins with superior ecosystem reach. As of mid-2026, Doubao’s monthly active users have surpassed DeepSeek’s, forcing Liang to move beyond the laboratory and into the commercial arena, deploying sales teams and exploring consumer-facing products to justify his new astronomical valuation.
Liang’s decision to personally bankroll nearly half the funding round is a 'firewall' intended to protect his vision from being diluted by corporate interests. He famously compared proprietary technology to building walls, while calling open-source technology a bridge. By keeping over 80% of voting control, he hopes to remain a 'contributor' rather than just a beneficiary of the AI revolution, even as he invites the very giants he once avoided into his cap table.
The romanticism of a small, focused team outmaneuvering the world’s largest corporations has collided with the reality of 'Big AI.' Liang Wenfeng’s massive bet suggests that in the next phase of the intelligence revolution, efficiency and brilliance are merely the table stakes. To survive the 'Great Game' of AI, even the most independent minds must eventually align themselves with the scale of the state and the deep pockets of the market.