DeepSeek Seeks Second Funding Round at $71B Valuation as Infrastructure Push Accelerates
DeepSeek, the Chinese artificial intelligence startup that rattled global AI markets earlier this year, has begun preliminary discussions with new investors for a second funding round at a pre-money valuation of approximately $71 billion — a 37% premium over the post-money valuation established in its first external raise just weeks ago, according to two people familiar with the matter cited by the Financial Times on July 14, 2026.
The new round comes roughly one month after DeepSeek closed its debut external financing around late May 2026, which raised approximately $7 billion (RMB 47.46 billion yuan) at a post-money valuation of about $52 billion. The terms of the current round have yet to be finalized.
The rapid succession of fundraising rounds signals a sharp escalation in DeepSeek's capital requirements, driven primarily by an aggressive push to build out its own computing infrastructure. In June 2026, the company published a recruitment plan on its official website for an in-house IDC data center team, with job descriptions indicating plans to develop gigawatt-scale computing campuses — covering everything from planning and design to operations — managed entirely by DeepSeek's own personnel.
The capital demands of such an undertaking are substantial. As a reference point, a 1GW data center project signed in April 2026 by domestic clean energy company Jinko Technology with the Zhongwei municipal government in Ningxia carried a planned total investment of approximately RMB 24.5 billion yuan (US$3.4 billion).
Beyond physical infrastructure, DeepSeek is also reported to be developing proprietary AI inference chips. Reuters reported on July 7 that the company initiated in-house chip development work roughly a year ago and has been engaging external partners, including chip design firms, foundries, and memory manufacturers. Recruitment of chip design engineers has increased in recent months, though hiring has been conducted privately rather than through public platforms, according to people familiar with the matter.
The infrastructure build-out is closely tied to DeepSeek's expanding product ambitions, particularly in the area of AI agents. Cui Tianyi, head of the DeepSeek Harness team, posted publicly on X in June that the team was severely understaffed despite conducting high-frequency interviews and broad recruiting campaigns. Agent-based products — which require continuous multi-round reasoning, planning, tool invocation, and reflection — can consume tens to hundreds of times more compute per task than standard conversational queries, meaning compute demand will likely intensify further once such products are launched.
The company is also in the midst of a broad hiring drive aimed at at least doubling headcount across all departments. DeepSeek has announced that the official release of its DeepSeek-V4 model is scheduled for mid-July — this week — marking the first major model launch since the completion of its first funding round and the opening move in what appears to be a new phase of accelerated product iteration.
For investors, the central question is whether DeepSeek can convert capital at this pace into durable technological and product competitiveness in an increasingly crowded global AI landscape.
Related Coverage:
DeepSeek's $7.1 Billion Pivot: From Frugal Lab to Capital-Intensive AI Contender