DeepSeek, the Chinese artificial intelligence start-up that has captured global attention with its powerful and cost-efficient models, is in talks to raise its first round of external funding at a valuation of $45 billion. The potential investment would be led by the China Integrated Circuit Industry Investment Fund, widely known as the "Big Fund," according to people familiar with the matter. The figure marks a dramatic escalation from earlier expectations, as DeepSeek initially sought a valuation of at least $10 billion when talks began only a few weeks ago.
The Big Fund is the country’s largest state-backed semiconductor investment vehicle, with a history of backing major domestic chip manufacturers such as Semiconductor Manufacturing International Corporation (SMIC) and memory chip producer Yangtze Memory Technologies Corp (YMTC). Its involvement in DeepSeek underscores the Chinese government’s growing commitment to building a self-sufficient artificial intelligence ecosystem. Earlier reports had also indicated that internet giants Tencent Holdings and Alibaba Group were in discussions to participate in the round, though their final commitments have not been confirmed.
A Rapidly Climbing Valuation
The valuation leap from $10 billion to $45 billion in a matter of weeks highlights the intense investor interest in DeepSeek’s technology and its potential strategic value. By late April, the reported valuation had already climbed to $20 billion, and the latest talks suggest the number has more than doubled since then. This rapid acceleration reflects not only the broader global enthusiasm for generative AI but also a specific belief in DeepSeek’s ability to compete at the frontier of model development.
Investors appear to be betting on DeepSeek’s long-term potential even though the company has so far focused on building cutting-edge frontier models rather than pursuing broad commercialisation. That distinguishes it from some domestic rivals such as Moonshot AI and Zhipu AI, the latter being a Hong Kong-listed company currently valued at approximately $52 billion. While Zhipu has moved aggressively to market its products and build enterprise revenue, DeepSeek has maintained a research-first ethos, releasing models that have repeatedly challenged Western competitors in performance benchmarks.
DeepSeek first made headlines in late 2024 with the release of its V3 model, followed by the R1 reasoning model in early 2025. These releases demonstrated that Chinese AI labs could match or approach the capabilities of leading American systems while using significantly fewer computational resources. The company’s open-weight approach, which allows researchers and developers to download and fine-tune its models, has made it a favourite in the global AI community. Its latest V4 model, announced in the spring, continues that trend with improvements in reasoning, coding, and multimodal understanding.
State-Backed Strategic Support
The involvement of the Big Fund is particularly significant because it marks a rare direct investment by the semiconductor vehicle into an AI model developer. Historically, the Big Fund has concentrated on chip design, manufacturing, and advanced packaging, with the aim of reducing China’s dependence on foreign semiconductors. By backing DeepSeek, the fund is expanding its mandate into the software and algorithmic layer of the AI stack.
This move aligns with a broader push by the Chinese government to encourage domestic tech companies to collaborate on an integrated AI ecosystem. Beijing has repeatedly called for greater coordination among firms across the AI supply chain, from data and compute to models and applications. The goal is to create a domestic alternative to the American-dominated AI infrastructure, especially in light of escalating export controls on advanced semiconductors.
Washington has implemented a series of restrictions that prevent Nvidia and other U.S. chipmakers from selling their most advanced AI accelerators to Chinese customers. In response, Chinese companies have had to rely on domestic alternatives or on chips that are slightly older and less capable. This has created a unique market dynamic in which Huawei, once known primarily as a telecom equipment maker, has emerged as the leading supplier of AI processors in China.
Huawei Optimisation and the Chip Race
DeepSeek has played a key part in this shift. At the release of its latest V4 model, the company stated that the software had been optimised to run inference workloads on Huawei’s Ascend 950PR chips. This is a major boost for Huawei’s ascend line of accelerators, as it demonstrates that a top-tier frontier model can operate effectively on domestic Chinese hardware. The optimisation also signals that DeepSeek is willing to work closely with local chip designers to ensure that software and hardware evolve together, similar to the tight integration seen between Nvidia and American AI labs.
Huawei’s market share in China has now overtaken that of Nvidia, according to industry estimates. That would have been unthinkable just a few years ago, when Nvidia commanded the vast majority of the Chinese AI chip market. The reversal was accelerated by U.S. export control rules that banned the sale of Nvidia’s H100 and A100 chips to China. Even the modified H800 and A800 versions were later restricted, forcing Nvidia to develop even less capable chips such as the H20, which still faced regulatory hurdles and lukewarm adoption.
The optimisation of DeepSeek’s V4 model for Huawei hardware is therefore both a technical and a political milestone. It demonstrates that China has reached a point where its leading AI models can run on domestically produced chips without a catastrophic loss of performance. While Huawei’s Ascend series is still considered less powerful than the top-end Nvidia accelerators in many respects, the gap has narrowed considerably. For Chinese enterprises that are unable to access Nvidia’s latest hardware, a combination of DeepSeek and Huawei can provide a viable substitute.
A New Phase for DeepSeek
Until now, DeepSeek has funded its operations primarily through the High-Flyer Group, a Chinese quantitative hedge fund founded by Liang Wenfeng, who also started DeepSeek. The move to accept external capital marks a new phase for the start-up, one that could bring not only additional resources but also pressure to generate returns. Many in the industry will be watching to see whether DeepSeek can maintain its research independence while taking money from state-backed funds and large internet companies.
The company’s decision to partner with the Big Fund may also be strategic in nature. Having a state-backed investor could provide DeepSeek with greater access to domestic compute resources and regulatory support, which are critical in an environment where data centres and advanced chips are increasingly subject to government scrutiny. It could also facilitate collaborations with other Big Fund portfolio companies, including chip designers and hardware manufacturers.
The funding round, if completed at a $45 billion valuation, would place DeepSeek among the most valuable AI start-ups in the world, trailing only a handful of companies such as OpenAI and Anthropic. It would also give the company a significant war chest to fund future model training runs, which require enormous amounts of computing power and data. Industry analysts estimate that training a frontier model can cost hundreds of millions of dollars, and costs are expected to rise further as models become more complex.
The involvement of Tencent and Alibaba adds another layer of complexity. Both companies are actively developing their own AI models and have large cloud computing businesses that compete with each other. Their participation in DeepSeek’s round could be seen as an attempt to hedge their bets by gaining exposure to an independent, highly capable model developer. It could also give DeepSeek access to distribution channels, user bases, and cloud infrastructure that would help bring its technology to a wider audience.
Implications for the Global AI Landscape
DeepSeek’s rapid rise has already shaken the global AI industry. When the R1 model was first released, it caused a sharp sell-off in technology stocks, as investors questioned whether U.S. firms could justify their massive capital expenditures on AI infrastructure if Chinese start-ups were able to achieve comparable results at a fraction of the cost. The concerns have not completely disappeared, but markets have since stabilised as American companies continue to push ahead with increasingly large models and new applications.
Nevertheless, DeepSeek’s success demonstrates that the AI race is no longer a two-player contest between the United States and China, but rather a highly competitive field with many participants. The company’s open-weight releases have accelerated research and development around the world, enabling smaller firms and even individual developers to build on state-of-the-art technology. This has also heightened concerns about the potential misuse of AI, as open-weight models can be fine-tuned for malicious purposes.
DeepSeek’s progression from a little-known offshoot of a hedge fund to a potential $45 billion company in the space of a few years is a testament to the rapid pace of change in the artificial intelligence sector. It also underscores the centrality of Chinese government support and the growing importance of domestic hardware. The final terms of the funding round have yet to be announced, but the market is already watching closely, as the outcome will have significant implications for the future of AI development in China and around the world.
Source:Silicon UK News