Key Takeaways
- Chinese AI firm DeepSeek is creating a proprietary AI chip optimized for inference operations
- This strategic shift aims to decrease dependency on Nvidia and Huawei silicon
- Nvidia stock dropped approximately 2% during premarket hours following the announcement
- The semiconductor project remains in nascent stages, with DeepSeek discreetly recruiting chip designers
- Simultaneously, DeepSeek is securing $7 billion through its inaugural external investment round
Chinese artificial intelligence company DeepSeek has initiated development of a proprietary semiconductor, multiple informed sources confirm. The processor targets inference operations — the computational process that enables AI models to deliver outputs — as opposed to the initial training phase for new models.
DeepSeek has maintained silence regarding the development publicly.
Shares of Nvidia experienced a roughly 2% decline in premarket sessions after the information surfaced.
Understanding the Chip’s Purpose
Inference-focused processors differ substantially from the high-performance GPUs utilized in AI model training. These chips generally cost less and consume reduced power. With artificial intelligence applications proliferating globally, computational requirements for inference are expanding rapidly.
DeepSeek’s semiconductor initiative would address this expanding market segment.
The development remains preliminary. DeepSeek has established contact with chip-design specialists, manufacturing facilities, and memory providers. Additionally, the organization has been recruiting semiconductor engineers through private channels rather than public job postings, according to two informed sources.
Strategic Reasoning Behind the Initiative
DeepSeek has historically depended on semiconductors from both Nvidia and Huawei for constructing and operating its AI systems. American export restrictions prohibit Chinese entities from acquiring Nvidia’s cutting-edge processors. These limitations have compelled DeepSeek toward greater utilization of Huawei’s Ascend processor lineup recently.
This past April, DeepSeek unveiled its V4 model optimized specifically for Huawei’s Ascend infrastructure. Demand for Huawei’s Ascend 950 processors skyrocketed following that product introduction.
Yet Huawei’s dominance over China’s $50 billion artificial intelligence chip sector faces mounting challenges. Both Alibaba and Baidu are engineering proprietary chips and capturing market portions.
DeepSeek’s semiconductor ambitions align it with similar industry players. OpenAI introduced its inaugural custom inference processor last month, dubbed Jalapeno, manufactured through partnership with Broadcom. Anthropic has similarly explored proprietary chip possibilities.
For DeepSeek, additional obstacles exist. American regulations prevent Chinese chip architects from utilizing cutting-edge international manufacturing facilities. Additional constraints restrict availability of high-bandwidth memory components, essential for inference chip functionality.
Liang Wenfeng, DeepSeek’s founder, acknowledged in a 2024 conversation that export limitations posed genuine obstacles for the organization.
The semiconductor initiative coincides with DeepSeek’s first acceptance of external investment. The firm was positioned to secure $7 billion through a financing round establishing valuation between $52 billion and $59 billion, Reuters disclosed in June. This marks a reversal from years of declining outside capital.
DeepSeek captured international spotlight in early 2025 after its R1 reasoning model achieved viral status. That model underwent training using Nvidia’s H800 chip, a China-specific variant that Washington subsequently prohibited.
Despite emerging as one of China’s most-scrutinized AI companies, the organization maintains a reserved public presence. Success of its semiconductor venture will hinge on overcoming both engineering complexities and persistent American trade limitations.





