Chinese artificial intelligence startup DeepSeek is concluding its first major external financing round, raising more than 50 billion yuan (approximately $7.4 billion or £5.5 billion) at a valuation close to $60 billion. The new figure marks a dramatic six-fold increase from the company's $10 billion valuation just two months ago, reflecting intense investor enthusiasm for DeepSeek's cost-efficient AI models and its growing importance in China's technology landscape. The round is expected to draw commitments from some of the country's largest internet and industrial companies, alongside venture capital funds and state-backed investment vehicles, according to people familiar with the matter.
Tencent Holdings, China's most valuable publicly traded internet company and a major investor in gaming, social media, and cloud infrastructure, is expected to contribute around 10 billion yuan. NetEase, known for online games and music streaming, and JD.com, the e-commerce and logistics conglomerate, are each anticipated to invest approximately 3 billion yuan. CATL, the world's largest electric vehicle battery maker, is expected to inject about 5 billion yuan, a sign that DeepSeek's technology is seen as relevant beyond the software sector. Chinese venture capital firms, including IDG Capital, Monolith, Loyal Valley Capital, and Shixiang Tech, have also been named as prospective backers. DeepSeek's founder and chief executive, Liang Wenfeng, who previously built the artificial intelligence-oriented quant fund High-Flyer Quant, is likely to contribute around 20 billion yuan of his own capital to the round, underscoring his confidence in the company's future.
A strategic shift toward external capital
DeepSeek has long been known for its unusual independence. The company was founded in 2023 by Liang Wenfeng, a mathematician and entrepreneur who previously established High-Flyer Quant, one of China's most successful quantitative hedge funds. High-Flyer's extensive investments in powerful computer chips and data centers provided DeepSeek with an ample internal source of funding and technical infrastructure. For several years, DeepSeek deliberately avoided external investors, relying on the profitability of High-Flyer's trading operations to support its AI research and development. This self-funded approach gave the startup unusual freedom to pursue long-term, open-ended research goals without pressure to deliver short-term commercial returns.
However, reports from local media indicate that DeepSeek recently decided to open its capital base in order to secure a well-defined market valuation, which is considered essential for retaining top engineering talent in China's competitive AI sector. Stock options and equity-based compensation packages are difficult to offer meaningfully without a clear, externally validated share price. By accepting investment from prominent names such as Tencent, JD.com, and CATL, DeepSeek can establish a benchmark that helps recruit and retain highly sought-after researchers and developers, who are often courted by local rivals like Zhipu AI, Baidu, Alibaba, and by global players such as OpenAI and Anthropic.
The rise of DeepSeek and its disruptive AI models
DeepSeek first drew global attention in late 2024 with the release of DeepSeek-V3, a large language model that matched or exceeded the capabilities of leading American systems at a fraction of the reported training cost. Then, in January 2025, the company released DeepSeek-R1, a reasoning model that stunned the industry by performing at a level comparable to OpenAI's o1 series on complex mathematics, coding, and logic problems, while using far fewer computational resources. The release of R1 triggered sharp sell-offs in global technology stocks, particularly Nvidia and other AI infrastructure providers, as investors began to question whether the billions of dollars being spent on ever-larger clusters of graphics processing units (GPUs) were truly necessary to achieve frontier-level performance.
The company has continued to pursue what it calls a price-efficiency-first strategy. When DeepSeek released its V4 model in April of this year, it promised that prices would drop significantly in the second half of the year when mass shipments of Huawei's Ascend 950PR supernodes are expected. The Ascend 950PR is a high-performance AI computing platform built around Huawei's in-house designed Ascend processors, which are increasingly viewed as a viable domestic alternative to Nvidia chips in China, particularly under strict US export controls that limit access to advanced semiconductors. DeepSeek's close collaboration with Huawei and other domestic chipmakers positions the startup at the center of China's drive to build a self-reliant AI supply chain, one that is less vulnerable to disruptions from American sanctions.
Potential investors and strategic motivations
The participation of Tencent, NetEase, and JD.com is particularly notable because these companies are not just passive financial investors. Tencent has been aggressively integrating AI into its WeChat ecosystem, cloud services, and gaming products, and a stake in DeepSeek could provide access to cutting-edge models that can be deployed at scale. JD.com, which operates massive logistics and e-commerce networks, is likely seeking AI solutions for warehouse automation, demand forecasting, and customer service. CATL's interest may stem from the growing need for AI-powered battery research, manufacturing optimization, and intelligent energy management systems. The company produces batteries for electric vehicles from Tesla to BMW and invests heavily in advanced computing to improve battery chemistry and production efficiency.
Venture capital firms participating in the round, including IDG Capital, Monolith, Loyal Valley Capital, and Shixiang Tech, represent some of the most prominent names in Chinese tech investing. Their presence indicates that DeepSeek, despite its late entry into external fundraising, is being valued along the lines of established AI heavyweights. State-backed funds are also expected to take part, although final terms have reportedly not been settled, including which specific state vehicles will provide the capital. Government involvement is significant in China's strategic technology sectors, where national priorities such as AI self-sufficiency and industrial modernization shape investment patterns.
China's AI supply chain and chip ecosystem
DeepSeek's relationship with domestic chipmakers extends beyond Huawei. Cambricon and MetaX, two Chinese semiconductor companies focused on AI processing, have said they are optimizing their chips for DeepSeek's V4 model. Cambricon, founded by researchers from the Chinese Academy of Sciences, specializes in chips designed specifically for neural network inference and training. MetaX, also known as Shanghai MetaX Integrated Circuit, is a relatively younger startup backed by Chinese state investment funds, and its processors are being deployed in some of China's AI data centers.
This collaboration between a leading AI model developer and domestic chipmakers is a key part of China's broader industrial policy. The government has repeatedly emphasized the need to reduce reliance on imported chips, particularly after the United States placed stringent export controls on advanced Nvidia GPUs such as the A100, H100, and H200. Chinese companies have responded by accelerating their own chip development programs, but one of the long-standing challenges has been a lack of mature software stacks and proven model compatibility. DeepSeek, by actively optimizing its algorithms to work well on Huawei, Cambricon, and MetaX hardware, helps demonstrate that these domestic alternatives can be used in real-world, cutting-edge AI applications.
Industry analysts have noted that DeepSeek's models are particularly well suited to the memory and compute constraints of domestic chips because of their efficient architecture and innovative training techniques. For example, DeepSeek's use of mixture-of-experts layers and multi-head latent attention reduces the number of parameters activated per token, lowering the amount of memory bandwidth required. These design choices deliver performance gains that are less dependent on having the largest, most advanced GPU clusters. As a result, a startup running on Chinese-built accelerators can achieve competitive performance without breaking the bank, a challenge to the conventional assumption that the US holds an insurmountable lead in artificial intelligence.
Retaining talent in an ultra-competitive market
Beyond the financial and strategic dimensions, the funding round is ultimately about people. The intense battle for AI talent has driven up compensation packages across China and the United States, and senior AI researchers can command annual salaries in the millions of dollars. For a startup like DeepSeek, which has historically used the profits of High-Flyer Quant to fund generous research budgets, offering equity in a formally valued company is an additional tool to keep key team members motivated. Liang Wenfeng's decision to invest his own money in the current round is meant to signal aligned interests with outside investors and employees.
The startup's academic approach has attracted some of the brightest minds in Chinese machine learning, many of whom value the freedom to publish research and influence the field in ways that big companies sometimes restrict. DeepSeek has released many of its model weights publicly, a practice that has both earned it widespread acclaim and caused concern among rival firms that see open models as a serious commercial threat. The decision to accept external funding is not necessarily a sign that DeepSeek is pivoting away from open research. Instead, it may allow the company to fund even larger experiments, such as training next-generation models on hundreds of thousands of domestic accelerators, without relying solely on High-Flyer's trading profits.
Timeline and next steps
Reports indicate that the funding round is still being finalized, with some details yet to be settled. The lineup of investors may change, and the exact amount attributable to state-backed funds remains unclear. However, the broad parameters appear to have been set: more than 50 billion yuan raised at a valuation of just under $60 billion. That valuation places DeepSeek among the most valuable private AI companies in the world, trailing only the likes of OpenAI and Anthropic when measured by recent secondary market transactions.
The completion of this round is expected to fuel DeepSeek's next phase of growth, which includes expanded computational infrastructure, deeper push into enterprise and consumer applications, and continued collaboration with Huawei's Ascend ecosystem. The company has already indicated that pricing for V4 model usage will fall dramatically once Ascend 950PR supernodes enter widespread deployment. If those systems deliver the expected improvements in energy efficiency and raw processing power, DeepSeek could further consolidate its position as a provider of low-cost, high-performance AI to developers and businesses throughout China and beyond.
The development also represents a meaningful moment for China's technological self-reliance strategy. As US restrictions on advanced semiconductor exports remain tight, the combination of DeepSeek's efficient model architecture and Huawei's accelerating chip roadmap could provide a template for how China achieves world-class AI capabilities under challenging conditions. The fact that Tencent, a company with global reach, and CATL, the dominant player in EV batteries, are willing to invest sizable sums into DeepSeek shows that Chinese industry leaders believe the startup's technology is strategically critical, not merely a laboratory curiosity.
In recent weeks, DeepSeek's V4 model has been benchmarked against several open-weights competitors, and early results suggest it continues to improve on reasoning, code generation, and multilingual understanding. DeepSeek has also made available several distilled smaller models that are inexpensive to host on edge devices or mobile phones, broadening the potential user base. The company is reportedly working on an official API platform with more reliability and uptime guarantees, a move that could attract enterprise clients who require service-level agreements for production workloads.
The first external funding round for DeepSeek is therefore much more than a financial milestone; it is an acknowledgment that the startup is transitioning from an ambitious research project into a scaled, commercially relevant AI company. The investors involved are not simply betting on one product release. They are supporting a broader infrastructure play that connects advances in chip design, algorithmic efficiency, and cloud deployment. With the expected price drops in the second half of the year and the optimization of chips from Huawei, Cambricon, and MetaX, DeepSeek is preparing to make its disruptive models accessible to a far wider audience. As the terms of the final agreement are settled and the capital begins to flow, the global AI community will be watching closely to see how DeepSeek invests this new war chest and whether it can maintain its position as the world's most exciting and unconventional AI startup.
Source:Silicon UK News