DeepSeek, China's most prominent artificial intelligence startup, is developing its own AI chip designed for inference workloads, according to three people familiar with the matter. The move represents a significant strategic pivot for the company and underscores how U.S. export restrictions are pushing Chinese tech firms to build domestic semiconductor alternatives.

The inference chip project, which began roughly a year ago, targets the fastest-growing segment of AI computing demand. Inference is the stage where a trained AI model generates responses for users, as opposed to the computationally heavier process of training new models from scratch. An inference-focused chip can be cheaper and less power-hungry than general-purpose graphics processing units (GPUs), making it an attractive target for companies looking to optimize their operations.

Why Is DeepSeek Building Its Own Chip?

DeepSeek's push into semiconductor development carries clear strategic importance. U.S. export controls bar Chinese companies from purchasing Nvidia's most advanced chips, a constraint that has directly affected DeepSeek's operations. The company previously trained its R1 reasoning model, which triggered significant volatility in U.S. tech stocks in January 2025, on Nvidia's H800 chip, a processor designed for the Chinese market that Washington banned in late 2023.

Since those restrictions tightened, DeepSeek has increasingly relied on Huawei chips. In April, the company released its V4 model adapted specifically for Huawei's Ascend processors, and Huawei confirmed its chips were used in part of the training for V4-Flash, a lighter version of the model. Orders for Huawei's Ascend 950 chips from Chinese tech companies surged following the launch.

DeepSeek founder Liang Wenfeng acknowledged the challenge in a rare 2024 interview, stating that chip export controls posed a significant obstacle for the company. Building its own silicon would provide greater autonomy and reduce vulnerability to future U.S. restrictions.

How Is DeepSeek Approaching This Challenge?

• Early-Stage Development: DeepSeek's chip effort remains in its infancy, with the company reaching out to external partners and holding discussions with chip-design firms, foundries, and memory companies to build its semiconductor ecosystem.
• Quiet Hiring Push: The company has increased recruitment of chip-design engineers in recent months, but notably conducted this hiring privately without posting positions on public job platforms, suggesting an effort to keep the project under the radar.
• Strategic Partnerships: DeepSeek is engaging with specialized partners across the chip supply chain, from design to manufacturing to memory components, rather than attempting to build everything in-house.

The company's effort to develop custom silicon mirrors a broader industry trend. OpenAI unveiled Jalapeno, its first custom inference chip developed with Broadcom, last month. Anthropic has also been weighing whether to build its own AI chips, according to reporting from April.

What Are the Obstacles Ahead?

Despite the strategic rationale, DeepSeek faces formidable technical and regulatory hurdles. Designing a competitive AI chip typically requires years of development and substantial capital investment. Manufacturing presents an even steeper challenge, as U.S. restrictions prevent Chinese chip designers from accessing the most advanced overseas foundries. Additionally, separate U.S. curbs have limited China's access to high-bandwidth memory, a critical component for AI inference chips.

Industry analysts remain skeptical about DeepSeek's near-term prospects. Richard Windsor, an analyst at Radio Free Mobile, stated that while DeepSeek's chip development is noteworthy, the company faces a fundamental constraint: "Nvidia is at zero in China and staying there. DeepSeek has almost no chance of selling silicon outside of China unless it gets access to leading edge manufacturing." Windsor added that the development does not meaningfully affect Nvidia's competitive position.

Huawei's dominance in China's domestic AI chip market, where it controls roughly half of the $50 billion market, is already being challenged by competitors. Alibaba and Baidu are developing their own AI chips and gaining market share, fragmenting what was once a more consolidated landscape.

What Does This Mean for the Broader AI Industry?

DeepSeek's chip initiative coincides with the company's first embrace of outside capital. In June, Reuters reported that DeepSeek was set to raise $7 billion in a maiden funding round that would value the company between $52 billion and $59 billion, reversing its long-standing strategy of rejecting external investment.

The timing is significant. DeepSeek rose to global prominence more than a year ago after releasing two highly efficient AI models that went viral worldwide, surprising many observers in Silicon Valley and Washington. Despite becoming a standard-bearer of China's AI ambitions, the company has maintained a notably low profile, declining to comment on its chip development efforts.

The broader context reveals how U.S. export controls are reshaping the global AI hardware landscape. Rather than slowing Chinese AI development, restrictions on advanced chip access are incentivizing domestic alternatives. This dynamic mirrors historical patterns in technology competition, where supply constraints often accelerate local innovation and reduce dependence on foreign suppliers. Whether DeepSeek succeeds or fails, its effort signals that Chinese AI companies view semiconductor independence as essential to long-term viability in an increasingly fragmented global tech ecosystem.