What's happening

DeepSeek, the Chinese AI laboratory that has been valued at between $52 billion and $59 billion in a reported $7 billion fundraising round disclosed in June 2026, is developing its own AI chip, according to sources cited by Reuters on July 7, 2026. The chip development effort began approximately one year ago, around July 2025, and has involved active recruitment of chip-design engineers as well as discussions with foundries and memory companies about potential partnerships.

The move represents a significant strategic escalation for DeepSeek, which has until now sourced its AI accelerator hardware externally. The company used Nvidia's H800 chip to train its R1 model and subsequently turned to Huawei's Ascend chips for its V4 model, released in April 2026. By pursuing proprietary silicon, DeepSeek would reduce its dependence on both of those suppliers for training and running its AI models.

Why it matters for markets

Nvidia's position in the Chinese AI accelerator market has already been substantially constrained by U.S. export controls, which have restricted the sale of its most advanced chips to Chinese customers. Analyst Richard Windsor of Radio Free Mobile stated plainly: "Nvidia is at zero in China and staying there." The Chinese domestic AI chip market is estimated at approximately $50 billion, with Huawei having captured around half of that figure through its Ascend product line. DeepSeek's entry into chip design, if successful, would introduce a third significant competitor into that domestic market and could further erode the addressable opportunity for any foreign supplier.

For Nvidia, with a market capitalization of $4.74 trillion and annual revenue of $253.49 billion, the direct revenue impact of losing a single Chinese customer to in-house silicon is not the primary concern. The broader implication is that DeepSeek's chip initiative, if it gains traction, could serve as a template for other well-capitalized Chinese AI labs to vertically integrate into hardware — collectively compressing the long-term addressable market for third-party AI accelerators in China. Nvidia shares slipped approximately 1.6% in premarket trading on July 7, 2026, following the Reuters report.

DeepSeek's ambitions in silicon face meaningful structural constraints. Windsor noted that "DeepSeek has almost no chance of selling silicon outside of China unless it gets access to leading edge manufacturing," pointing to the foundry access limitations that continue to shape the competitive landscape for Chinese chip developers. The company's ability to close the gap with leading-edge process nodes will be a critical variable in determining how competitive any resulting chip could be relative to current-generation Nvidia or Huawei offerings.

Sectors and assets to watch

Nvidia (NVDA) is the most directly referenced public company in this development. Its H800 chip was used by DeepSeek to train the R1 model, and any successful in-house chip program at DeepSeek would reduce future demand for Nvidia hardware from that customer. Nvidia's current price sits at $195.55 within a 52-week range of $158.39 to $236.54, and the company generates $253.49 billion in annual revenue across its GPU and AI accelerator product lines, including the H100 and A100 data center chips.

Huawei, whose Ascend chips were used by DeepSeek for its V4 model and which holds approximately half of China's estimated $50 billion domestic AI chip market, is also a named party in this competitive dynamic, though it is not publicly traded. The foundry and high-bandwidth memory segments are implicated as well, given that DeepSeek has reportedly been in discussions with firms in both categories as it advances its chip design effort. SMIC, as China's leading domestic foundry, occupies a central role in any scenario where Chinese AI companies seek to manufacture custom silicon domestically.

What to watch next

Key developments to monitor include any formal announcements from DeepSeek regarding chip design partnerships with foundries or memory suppliers, the timeline and specifications of any prototype or production chip, and whether additional Chinese AI laboratories announce similar vertical integration efforts into custom silicon. The progress of DeepSeek's reported $7 billion funding round — and how proceeds are allocated between software research and hardware development — will also be relevant. On the regulatory side, any changes to U.S. export control policy affecting advanced chip manufacturing equipment or foundry access for Chinese firms could materially alter the feasibility of DeepSeek's hardware ambitions.