DeepSeek Restarts Second Funding Round With $8B Target

DeepSeek has restarted its second funding round, seeking nearly $8 billion as the Hangzhou-based artificial intelligence start-up pushes to finance a major expansion of its computing infrastructure, according to reports. The company is aiming for a valuation of nearly 500 billion yuan, equivalent to about $74 billion or £55 billion. The round had been paused in July after remarks made to investors were leaked, creating unwanted scrutiny around the closely watched AI developer. The decision to resume the process signals that DeepSeek and its backers believe investor demand remains strong despite the leak and a more cautious climate for late-stage technology financing.

Monolith Management, an investor in Beijing-based Moonshot AI, the developer of the Kimi line of models, is reportedly in talks to contribute to the round. The potential participation of an investor with experience in China’s competitive AI model market would add a strategic dimension to the fundraising. The talks are not yet completed, and the final size, valuation and investor list could still change. DeepSeek has not publicly confirmed the details of the round.

Key facts

• DeepSeek has resumed its second funding round after pausing it in July.
• The company is seeking nearly $8 billion, or about £6 billion.
• It is targeting a valuation of nearly 500 billion yuan, about $74 billion or £55 billion.
• Monolith Management is in talks to contribute, according to reports.
• DeepSeek plans a large data centre in Inner Mongolia as part of efforts to secure 1 gigawatt of computing power, potentially including leased capacity from third parties.
• Some of the new funding would reportedly go toward the data centre project.
• DeepSeek has told users it intends a substantial price increase, without specifying the new pricing.
• Its current API prices are $0.14 per million input tokens and $0.28 per million output tokens.
• Its first funding round closed earlier in the summer at a valuation of about 350 billion yuan.

The restart comes at a delicate moment for China’s AI sector. DeepSeek became one of the country’s most prominent model developers by showing that high-performance systems could be trained and operated at far lower cost than many Western observers had assumed. That message resonated with developers, enterprises and investors, and it challenged the idea that only companies spending hundreds of billions of dollars on infrastructure could compete at the frontier. The company’s low-cost API pricing made its models accessible to a broad range of users, from independent developers to larger software teams.

DeepSeek currently charges $0.14 per million input tokens and $0.28 per million output tokens. Those rates are dramatically lower than some competing offerings. Moonshot’s latest Kimi K3 model is priced at $3 per million input tokens and $15 per million output tokens, while Anthropic’s latest Fable 5 model is priced at $10 per million input tokens and $50 per million output tokens. The gap illustrates how aggressively DeepSeek has used price as a competitive weapon. It also explains why the company’s warning of a substantial price increase is being watched closely. DeepSeek told users last Thursday that it intended a significant rise in prices but did not specify the upcoming pricing plans. A price increase could help the company fund compute and infrastructure, but it could also test the loyalty of cost-sensitive developers who have built products around its low rates.

The fundraising is closely tied to infrastructure. DeepSeek is reportedly planning to construct a massive data centre in Inner Mongolia, part of a broader initiative to gain access to 1 gigawatt of computing power. That capacity may include leasing capacity from third parties as well as building its own facilities. A gigawatt of computing power is a substantial target for any AI developer, reflecting the enormous electricity and hardware requirements of training and serving advanced models. Data centres in Inner Mongolia can be attractive because of energy availability, cooler conditions and proximity to power sources, although the region also presents logistical and operational challenges. For DeepSeek, the project would represent a move beyond its reputation for software efficiency into the capital-intensive business of owning and securing infrastructure.

Some of the funding from the second round would reportedly be used to finance the Inner Mongolia data centre. That link helps explain why the company is seeking a large round despite its low-cost reputation. Training and deploying frontier models require not only chips but also land, power, cooling and networking. As demand for AI services grows, access to reliable compute has become a strategic bottleneck. DeepSeek’s plan to secure 1 gigawatt, whether through construction or leasing, suggests that it wants more control over its supply of computing power rather than relying entirely on external providers. The strategy mirrors a broader trend among AI developers, which are increasingly racing to lock in energy and data centre capacity as model sizes and user demand increase.

The funding round also provides a window into how investors are valuing China’s leading AI start-ups. DeepSeek’s first funding round closed earlier in the summer at a valuation of about 350 billion yuan. The new target of nearly 500 billion yuan would mark a substantial increase in a short period. That jump reflects both the company’s prominence and investor expectations for future growth, but it also raises questions about how quickly DeepSeek can convert technical attention into durable revenue. The company has not disclosed detailed financial results, and the reported valuation remains subject to negotiation and market conditions.

Monolith Management’s potential involvement is notable because of its existing exposure to Moonshot AI, another high-profile Chinese model developer. Moonshot’s Kimi models compete in the same broad market for AI assistants and developer tools. An investment in DeepSeek would give Monolith a stake in two of China’s better-known AI ventures, although the firms have different product strategies and pricing philosophies. For DeepSeek, backing from an investor familiar with the sector could provide more than capital. It could also bring operational insight, industry connections and a signal to other investors that the round is progressing.

The July pause shows how sensitive fundraising can be for AI companies in the public eye. DeepSeek’s remarks to investors were leaked, and the company chose to pause the process rather than proceed amid the resulting attention. Restarting the round suggests that the company has decided the benefits of fresh capital outweigh the risks of further scrutiny. It also indicates that the leak did not permanently derail its plans. Still, the company will need to manage information carefully as it negotiates with investors and prepares for a larger public profile.

The broader market backdrop is mixed. AI investment remains enormous, but investors have become more selective about which companies and business models they support. Companies with clear paths to revenue, strong technical teams and access to compute are attracting capital, while those without differentiation face tougher conditions. DeepSeek sits in a strong position on technical reputation and developer adoption, but it must now prove that it can scale its infrastructure and commercial model. The reported $8 billion target would give it substantial resources to do so, although the final amount may differ.

Pricing will be a central issue. DeepSeek’s low rates helped it win attention and users, but they also limit revenue per token. A substantial increase could improve unit economics and help fund the data centre build. However, competitors are also improving their models and adjusting prices. The company will need to balance the need for more revenue with the risk of pushing developers toward alternatives. The lack of specific pricing plans leaves the market guessing about how steep the increase will be and when it will take effect.

The data centre plan also carries execution risk. Building or securing 1 gigawatt of computing capacity is a multi-year undertaking that requires hardware, energy contracts, construction and operational expertise. Leasing capacity from third parties could speed access but may come at a higher cost and with less control. Building its own facility could offer long-term advantages but requires significant upfront investment. DeepSeek’s decision to pursue a mix of options, as reported, reflects the practical realities of the AI infrastructure race.

The first funding round closed earlier in the summer at a valuation of about 350 billion yuan. The new target of nearly 500 billion yuan would mark a substantial increase in a short period, underscoring how quickly investors have re-rated the company’s prospects even as it moves deeper into capital-intensive infrastructure.

Source:Silicon UK News