DeepSeek Nears $45 Billion Valuation as China's State Chip Fund Leads Its First-Ever Investment Round

In January 2026, when DeepSeek’s R1 model stunned Western AI developers by matching frontier performance at a fraction of the cost, the company’s unusual financial structure was part of the mystique. Unlike every major Western AI lab — which had raised billions from venture capitalists, Big Tech, and sovereign wealth funds — DeepSeek had been funded entirely from the balance sheet of High-Flyer Capital Management, the quantitative hedge fund co-founded by its 40-year-old creator, Liang Wenfeng.

That era of radical self-reliance is now ending. According to reporting by the Financial Times, confirmed by Bloomberg and multiple other outlets, China’s state-backed National Integrated Circuit Industry Investment Fund — colloquially known as the “Big Fund” — is in advanced talks to lead DeepSeek’s first-ever external financing round at a valuation of approximately $45 billion. Cloud giants Tencent and Alibaba are reportedly in parallel discussions to participate as co-investors.

A Valuation That Moved Fast

The speed of the valuation shift is itself newsworthy. In mid-April 2026, discussions had reportedly centered on a much more modest $300 million raise at a $10 billion valuation. Within three weeks, the figure had ballooned to a $3–4 billion raise at $45 billion — a 4.5x jump in the headline valuation in less than a month, reflecting both the intensity of investor demand and China’s strategic urgency around securing influence over its most prominent AI lab.

The $45 billion figure would make DeepSeek one of the most valuable AI companies in the world, placing it in the same bracket as some of the largest private technology companies globally. Neither DeepSeek nor the Big Fund has publicly confirmed the talks, and the terms remain subject to change, but the FT’s reporting is based on multiple sources familiar with the discussions.

The proceeds — estimated at $3 billion to $4 billion — are intended primarily to expand DeepSeek’s computing infrastructure and increase compensation for its engineering and research staff. That second priority reflects a growing reality in China’s AI ecosystem: talent retention is intensifying as both domestic companies and, to a lesser extent, international labs compete for the same narrow pool of researchers capable of working at the frontier.

Who Is the Big Fund?

The China Integrated Circuit Industry Investment Fund, established in 2014, was created explicitly to reduce China’s dependence on foreign semiconductor technology. It has invested in virtually every significant player in China’s chip industry, from SMIC — China’s largest foundry — to memory chipmakers and equipment developers. The fund has historically focused on hardware, making its lead role in a software and AI research company a notable strategic pivot.

That pivot reflects a broader shift in China’s industrial policy calculus. After years of US export controls on advanced semiconductors — most recently tightened in late 2025 — Beijing has increasingly concluded that the path to AI self-sufficiency runs as much through model development and software efficiency as through hardware. DeepSeek’s ability to deliver frontier-class performance while using older, export-controlled-compliant chips made it a national asset almost overnight.

By putting the Big Fund in front of DeepSeek’s cap table, Beijing is doing more than providing capital. It is signaling — to domestic companies, foreign governments, and the investment community — that DeepSeek’s continued development is a matter of state priority, not just commercial ambition.

DeepSeek’s Path to This Moment

Liang Wenfeng founded DeepSeek in July 2023, spinning it out of High-Flyer, the quant fund he co-founded in 2015 that became one of China’s most successful algorithmic trading shops. The move was described at the time as a bet that mathematical rigor and computational efficiency — core competencies in quantitative finance — could be transferred to frontier AI development.

That bet has paid off spectacularly. DeepSeek’s V3 model, released in December 2025, reportedly cost approximately $5.6 million to train, a figure that sent shockwaves through an industry where comparable runs were assumed to cost hundreds of millions. The January 2026 release of R1 — a reasoning model benchmarking alongside the best Western counterparts — triggered a selloff in US AI infrastructure stocks and prompted urgent reassessments across Silicon Valley of the economics of model development.

The efficiency advantage appears to stem from a combination of factors: novel architecture choices, aggressive use of techniques like mixture-of-experts and distillation, and a culture that prioritizes engineering efficiency as a first principle rather than a cost-saving measure. Whether those advantages will persist as both Western and Chinese labs absorb and replicate DeepSeek’s published techniques remains the central open question in the global AI race.

Geopolitical Dimensions

The structure of the fundraising carries significance beyond the balance sheet. A state-guided investment fund taking a meaningful stake in DeepSeek changes the company’s character, at least in the eyes of foreign governments. What had been an unusually independent Chinese AI lab — funded by a private hedge fund, staffed largely by academic researchers, and seemingly insulated from direct government influence — will now have a more legible relationship with the state.

That matters for several reasons. For governments and corporations considering whether to use or collaborate with DeepSeek’s technology, the Big Fund’s involvement will accelerate existing national security reviews. The US, EU, and others have already expressed concerns about Chinese AI products in sensitive applications; a state-fund investment will sharpen those concerns and potentially trigger new restrictions.

For China domestically, however, the signal runs in the opposite direction: legitimacy, stability, and a commitment that the government sees DeepSeek’s success as a national strategic objective rather than a private commercial achievement.

What Comes Next

The fundraising, if completed at the reported terms, will significantly expand DeepSeek’s capacity to compete at the frontier. More compute means larger training runs, which in turn enable more capable models — assuming the efficiency gains that made DeepSeek’s earlier work remarkable can be maintained at greater scale.

The participation of Tencent and Alibaba as co-investors would also embed DeepSeek deeper into China’s domestic technology ecosystem, potentially accelerating distribution through their cloud platforms and consumer applications. That distribution advantage could matter enormously as AI model competition shifts from benchmark performance to real-world deployment.

For Western AI companies watching these developments, the message is unambiguous: the assumption that China’s AI sector would remain perpetually behind due to chip access restrictions has not survived contact with reality. DeepSeek’s technical achievements have already demonstrated that efficiency can substitute for raw compute up to a point — and now it appears poised to acquire substantially more of that too.