Chinese artificial intelligence start-up DeepSeek is finalising its first external funding round at a valuation of just under $60 billion. The round is expected to raise more than 50 billion yuan, roughly $7.4 billion or £5.5 billion, according to unnamed people familiar with the matter. This valuation is a major leap from the company’s reported $10 billion valuation from two months ago and reflects the fast-moving competition in advanced AI development.

A sixfold valuation leap

DeepSeek began as a model-making unit supported by High-Flyer Quant, the Chinese quantitative investment manager founded by Liang Wenfeng. In the past year, the start-up went from relative obscurity to a central figure in the global AI conversation. Its open-weight releases and efficient model designs showed that large-scale AI research could be done with less capital than many industry leaders had assumed.

The latest funding talks imply a valuation close to the $60 billion level. That is more than six times the $10 billion figure that was attached to the company just over two months ago. Such a sharp rise has attracted attention because private AI valuations are usually tied to revenue, user growth, and long-term technical advantage. DeepSeek has all three, although its revenue base is still developing.

Major investors line up

Tencent Holdings is expected to invest 10 billion yuan, making it one of the largest corporate backers in the round. NetEase and JD.com are each expected to contribute about 3 billion yuan. CATL, the world’s largest electric vehicle battery maker, is expected to invest 5 billion yuan. Several venture capital firms, including IDG Capital, Monolith, Loyal Valley Capital and Shixiang Tech, are also said to be involved.

The mix of investors shows the wide reach of DeepSeek’s technology. Tencent brings cloud, social and gaming capabilities. NetEase is focused on interactive media and AI-driven entertainment. JD.com represents e-commerce, logistics, and enterprise supply chain services. CATL adds expertise in intelligent manufacturing and EV batteries. This breadth gives DeepSeek potential distribution channels and data-rich application environments across several major industries.

Founder capital and state-backed funds

DeepSeek founder and chief executive Liang Wenfeng is expected to invest about 20 billion yuan of his own money, according to reports. That would make his personal commitment larger than any single external corporate investor in the round. The presence of founder capital is often interpreted as a confidence signal, and it may help bridge any disagreement over valuation or control.

State-backed funds are also expected to participate, though final terms have not been settled. There is still uncertainty about which investment vehicles will supply the state capital and in what form. Some Chinese technology financing rounds include state guidance funds that are designed to encourage domestic innovation in critical technologies such as semiconductors, AI, and advanced software.

The decision to raise external capital

DeepSeek had previously resisted bringing in outside investors. Because High-Flyer Quant was profitable and able to provide infrastructure funding, the start-up enjoyed unusual independence. That structure allowed researchers to focus on breakthroughs without quarterly revenue or user growth targets.

The change of approach appears to be related to talent retention. DeepSeek is reportedly seeking a well-defined market value in order to offer equity packages that can compete with other major AI companies in China. The fierce competition for experienced engineers, researchers, and technical leaders has forced model developers to be more transparent about their financial standing.

An external funding round also helps institutionalise corporate governance. A board of investors may bring operational discipline, while still allowing Liang Wenfeng to maintain a strong technical vision. For the wider market, a formal valuation increases confidence among enterprise customers and potential business partners that DeepSeek is stable enough for long-term commitments.

DeepSeek’s cost-efficient breakthrough

DeepSeek disrupted the AI world last year by releasing a series of capable models that combined strong performance with lower training and inference costs. Developers around the world were surprised by the low prices and open access that DeepSeek offered compared with some Western alternatives.

The company’s large language models and reasoning systems found particular traction among developers who wanted to experiment with advanced AI without committing to the budgets demanded by other frontier labs. DeepSeek’s engineering choices, including efficient attention mechanisms and optimised training strategies, reduced the number of GPU hours needed to reach state-of-the-art results.

V4 model and falling prices

DeepSeek has continued that price-focused strategy with newer releases. In April, after the release of its V4 model, the company said prices would drop significantly in the second half of the year. The expected price cuts are linked to the availability of Huawei’s Ascend 950PR supernodes, which are scheduled to ship at scale during that period.

These supernodes are large parallel computing clusters based on Huawei-designed AI processors. Successfully deploying them for model training and inference could lower DeepSeek’s cost base and reduce dependence on imported accelerators. The statement about lower prices is also a message to the broader market: DeepSeek plans to use hardware advances to make AI more affordable.

Domestic chipmakers optimise for DeepSeek

Chinese chipmakers including Huawei, Cambricon and MetaX are already responding to DeepSeek’s model architecture. They say they are optimising their AI processing chips for DeepSeek’s V4 model. These efforts are important for China’s goal of building an independent AI supply chain.

Domestic chip companies have faced challenges in making their software stacks compatible with the most popular open-source AI frameworks. DeepSeek’s cooperation can help close that gap. By tailoring low-level kernels and programming interfaces to DeepSeek’s workloads, Huawei and its rivals can make their hardware more attractive to Chinese AI developers.

The collaboration also gives chipmakers a concrete customer path. Instead of selling chips to corporate data centres with uncertain software support, they can point to a frontier AI company that is actively using their processors to train and serve advanced models.

Value for strategic investors

The round’s strategic importance goes beyond financial returns. For Tencent, investing in DeepSeek may accelerate the adoption of GenAI features across its WeChat, cloud and gaming operations. For NetEase, AI-generated content can lower the cost of producing interactive entertainment and improve personalised user experiences. JD.com can apply DeepSeek’s natural-language models to search, customer support, and warehouse automation, deepening its AI integration throughout the retail supply chain.

CATL’s projected investment represents the industrial side of the AI ecosystem. Battery manufacturing relies heavily on optimising complex supply chains and manufacturing processes. AI models that understand video from factory floors can improve quality control, while language models can assist in maintenance and production planning. A partnership with DeepSeek may also open up future applications in battery research.

Implications for the global AI race

DeepSeek’s rising valuation comes at a time when nations and companies are reassessing their positions in the artificial intelligence market. Export controls have made advanced Nvidia chips harder to obtain in China. DeepSeek’s ability to develop efficient models under those constraints has turned a potential weakness into a competitive narrative.

If Huawei’s Ascend platforms become reliable alternatives, Chinese AI companies will be less exposed to US-led restrictions. DeepSeek is positioned to prove that a combination of domestic chips and optimised algorithms can work at the frontier level. A successful deployment would encourage more Chinese start-ups to build on domestic hardware.

At the same time, DeepSeek’s rise intensifies competition for global model builders. The promise of lower prices and open weights will pressure Western providers to justify their pricing and architecture decisions. Enterprises may begin to benchmark popular models against DeepSeek’s performance-per-dollar ratio.

What could come next

Several details of the funding round remain unresolved. The final number of investors, the exact amount of state-backed participation, and the legal structure of the transaction are still under negotiation. But the general outline is clear: DeepSeek is moving into a new phase with external capital and a publicly recognised valuation.

The company will likely use the new resources to expand compute availability, develop its enterprise services, and hire across research and engineering functions. It may also invest in new data centres, which are critical for training next-generation models.

DeepSeek’s relationship with Huawei, Cambricon and MetaX will be watched closely. The success of those chip optimisation programmes could set the pace for the wider Chinese AI industry. For now, DeepSeek remains one of the most closely followed AI start-ups in the world, balancing technical innovation with market-building efforts and a growing roster of powerful corporate allies.

Source: Silicon UK News