Chinese AI startup DeepSeek has begun preliminary talks with new investors about a fresh funding round just six weeks after closing its first-ever financing in late May. The Financial Times first reported the development on July 14, with Bloomberg separately confirming that the company has entered discussions with new backers about a second round.
The latest round would value DeepSeek at around $71 billion before the transaction is completed, up 37% from the $52 billion post-money valuation at which it raised its debut round of about $7 billion. DeepSeek did not respond to requests for comment from Reuters or the Financial Times. The details of the new round have yet to be determined, and it is still in its early stages.
Along with the fundraising conversations, Bloomberg reported that DeepSeek has began preparing for an initial public offering, which may take place as early as this year – an extremely ambitious pace for a business that only finished its first external investment round in May.
The combination of a rapid second fundraise and IPO preparations signals that DeepSeek is operating on a capital deployment schedule that has surprised even those watching the Chinese AI sector closely. Three sources with knowledge of the matter told the FT that DeepSeek’s founder Liang Wenfeng was the largest single investor in the first round, personally contributing approximately $3 billion of the $7 billion raised. Proceeds from that round are currently being used to strengthen the company’s computing infrastructure and accelerate hiring of AI researchers.
“Chinese AI startup DeepSeek is weighing raising new funds one month after closing its first round and has begun preparations for an IPO, according to media reports. A new round would value the company at about $71 billion before the deal.”~Reuters
Data Centres, AI Chips and Agents Are Driving the Capital Demand:
The simple explanation for DeepSeek’s exceptionally quick return to capital markets is that the company’s computational requirements are expanding more quickly than its current resources can handle. The FT’s sources claim that the accelerated funding timeline is motivated by the anticipated amount of money needed to build DeepSeek’s own data center and purchase more AI chips at scale. The demand for computational power is significantly higher than what its previous model development work required due to its push into AI agents, which are systems that can act independently in the actual world instead of just answering questions. Running AI agents at scale is far more resource-intensive than serving a conversational AI, and DeepSeek’s rapid user growth since its V3 and R1 models made global headlines has only added to the infrastructure pressure.
The chip issue is especially problematic for DeepSeek considering the geopolitical climate in which it works. Reuters reported earlier this month that DeepSeek is actively developing its own proprietary AI chip in order to reduce its reliance on both Nvidia, whose most advanced GPUs are subject to US export controls that prevent them from being sold to Chinese entities, and Huawei, DeepSeek’s primary alternative supplier. Building domestic chip capability at the scale needed to train and serve frontier models would require significant capital over an extended period, which provides additional context for why a second fundraise is happening this quickly after the first.
“DeepSeek is in preliminary talks about a new funding round valuing it at about $71bn, weeks after raising $7bn in its first external fundraise. DeepSeek founder Liang Wenfeng personally invested roughly $3bn in the first round. The company is also preparing for a possible IPO as early as this year.”~Financial Times
From Open-Source Disruptor to $71 Billion Juggernaut in Under Two Years:
DeepSeek’s rise from relative obscurity to one of the most discussed AI companies in the world happened with extraordinary speed. The company originally founded as a research division of Chinese quantitative hedge fund High-Flyer Capital released its V3 and R1 models in late 2024 and early 2025, drawing immediate and widespread praise from AI researchers globally, including in Silicon Valley.
The R1 reasoning model in particular garnered attention for performing at the level of leading Western systems while having been trained using methods widely described as significantly more compute-efficient, a result that challenged the prevailing assumption that China was several years behind the US in frontier AI capability.
The models’ open-source release and strong benchmark performance triggered what some described as a market shock in US AI stocks, briefly denting valuations at Nvidia and other AI infrastructure companies as investors absorbed what DeepSeek’s efficiency meant for compute demand. In the months since, DeepSeek has continued releasing new model versions and has become China’s most internationally recognised AI brand. The $52 billion first-round valuation it secured in May already put it ahead of many well-established Western AI companies. A $71 billion pre-money valuation for the second round would make it one of the most valuable private AI companies in the world, behind only OpenAI, Anthropic, and xAI.
“DeepSeek has begun talks with new backers about a fresh funding round and may file for an IPO as early as this year, Bloomberg reports. The new round would follow a $7 billion raise at a $52 billion valuation just six weeks ago.”~Bloomberg Technology
IPO Preparations Add a New Dimension to China’s AI Capital Race:
The IPO preparations reported by Bloomberg add a dimension beyond simple fundraising. A public listing would give DeepSeek access to a far larger pool of capital over a longer period, provide a public currency for acquisitions and talent retention, and establish a market-set valuation that goes beyond private investor negotiations. The timing potentially this calendar year would be ambitious given the regulatory processes involved in a Chinese company listing, whether domestically or in Hong Kong. It would also make DeepSeek the first major Chinese AI frontier lab to go public, setting a benchmark that other Chinese AI companies including Zhipu AI, Moonshot AI, and Manus would be measured against.
The pace of DeepSeek’s capital activity, a $7 billion first round in May, a potential $71 billion second round in July, and IPO preparations simultaneously reflects the broader intensity of the global AI infrastructure race in 2026. US hyperscalers are spending hundreds of billions of dollars building out compute capacity. Chinese AI companies face a separate but equally urgent pressure, compounded by chip access restrictions that make domestic investment in alternative chip development and data centre construction not just commercially sensible but strategically necessary.
“DeepSeek is considering a second financing round, weeks after concluding its first. The company raised $7 billion at a $52 billion valuation in May. A new round could value it at $71 billion — a 37% jump. The fast schedule is driven by data centre construction and AI agent development needs.”~PYMNTS
