The AI company that challenged Silicon Valley with comparatively little outside capital has brought Tencent, CATL and some of China's largest state-backed institutions onto its investor list.
For much of its short history, DeepSeek stood apart from the rest of the AI industry. While American rivals raised billions to build larger models and buy computing power, the Chinese startup relied heavily on the resources of its founder's quantitative trading business.
That era appears to be over.
DeepSeek has closed a funding round worth approximately 105.5 billion yuan, or $15.8 billion, according to The Information, citing people with direct knowledge of the transaction.
Nearly 30 investors participated, including Tencent, battery manufacturer CATL, semiconductor companies and several of China's largest state-backed financial institutions.
The financing values DeepSeek at approximately 615.5 billion yuan, or around $87 billion, according to the report.
It is the company's second external funding round and the largest venture capital raise reported for a Chinese AI company.
The scale of the deal is striking. DeepSeek's first fundraising, completed earlier this year, raised more than $7 billion. The company had previously been an outlier among major AI developers, financing its research largely through founder Liang Wenfeng's hedge fund, High-Flyer.
The latest transaction also changes the profile of its shareholders.
Tencent and CATL were among DeepSeek's earliest external backers and remain two of its largest investors.
The second round brought in a much broader group, including memory chipmaker ChangXin Memory Technologies, semiconductor manufacturer SMIC, smartphone maker Vivo, appliance manufacturer Midea and automaker Geely.
Meituan, Luxshare, Trip.com and social media platform Xiaohongshu also participated, according to The Information.
These businesses operate across industries that depend increasingly on AI, from consumer applications and electric vehicles to semiconductor manufacturing.
For them, an investment in DeepSeek provides exposure to one of China's most prominent foundation-model developers.
For DeepSeek, the capital brings relationships with companies that collectively control substantial industrial, computing and distribution resources.
The participation of SMIC and ChangXin is particularly notable.
China's AI developers face restrictions on access to some of Nvidia's most advanced processors, while domestic semiconductor manufacturers are working to close the performance gap with overseas suppliers.
Closer commercial relationships between model developers and Chinese chipmakers could become strategically important as the country attempts to reduce its reliance on imported AI hardware.
The composition of the round also reveals a significant shift in DeepSeek's ownership.
State-backed investors supplied approximately 27% of the new financing, compared with 2% in the company's first external round, according to The Information.
Participants included China Mobile, China Telecom, China Life, CICC and Citic Securities, alongside China's national AI investment fund.
The increased participation of government-linked institutions could give DeepSeek access to additional financial and industrial resources.
It may also intensify scrutiny from Washington, where policymakers have already raised questions about the relationship between China's AI industry, national security priorities and state-backed technology development.
State investment does not necessarily mean that the Chinese government controls DeepSeek. The company's ownership arrangements and investor voting rights remain important distinctions.
In its first round, most investors reportedly invested through an entity controlled by Liang, with limited voting rights and restrictions on selling their holdings.
The full governance terms of the latest financing have not been publicly disclosed.
DeepSeek became internationally prominent in January 2025 when its R1 reasoning model challenged assumptions about the cost of developing competitive AI.
Its use of efficient model architectures, reinforcement learning and open-weight releases made it an important competitor to American AI laboratories.
But efficiency does not eliminate the need for computing infrastructure.
Training larger models, serving growing numbers of users and developing new architectures still require substantial investment in processors, networking and electricity.
The Information reported that DeepSeek expanded its latest fundraising target as the price of advanced Nvidia systems rose sharply in China.
Access to Blackwell-generation hardware has become particularly expensive, reflecting both supply constraints and US export restrictions.
The company's investors are therefore financing more than research salaries. They are helping fund an increasingly capital-intensive computing operation.
DeepSeek is also working to improve its use of domestic hardware, including Huawei accelerators, as China builds alternatives to American chip suppliers.
That creates a difficult engineering challenge.
Models trained on Nvidia's hardware and software ecosystem cannot necessarily be moved to alternative accelerators without extensive optimization.
For a company competing at the frontier of AI, every improvement in hardware utilization can affect both model performance and operating economics.
Liang founded DeepSeek in 2023 after building High-Flyer into one of China's prominent quantitative investment firms.
The hedge fund had accumulated substantial computing resources before US restrictions tightened, giving the research team an advantage that few independent startups could replicate.
DeepSeek initially concentrated on developing models rather than building an expensive consumer subscription business.
Its open-weight strategy helped establish its reputation among developers and encouraged businesses to build applications around its technology.
The company's rapid fundraising expansion suggests that maintaining that position will now require access to capital on a scale closer to its biggest international competitors.
According to The Information, DeepSeek is also preparing for a potential listing on Shanghai's STAR Market.
No completed IPO or definitive public listing timetable has been announced.
The new financing marks a turning point for the company that became famous for challenging the idea that frontier AI required unlimited spending.
DeepSeek demonstrated that technical efficiency could change the competitive landscape. Its latest funding shows that even highly efficient AI laboratories may eventually need enormous amounts of capital to remain at the frontier.
Sources: The Information on investors and structure · The Information on the close · Reuters