Key Takeaways
- The Chinese AI company is creating a proprietary chip specialized for inference operations rather than training
- This strategic shift aims to decrease dependency on both Nvidia and Huawei hardware
- Nvidia stock experienced a roughly 2% decline during premarket hours following the disclosure
- The project remains in nascent stages, with DeepSeek discreetly recruiting semiconductor design specialists
- Simultaneously, the company is securing $7 billion through its inaugural external investment round
A Chinese artificial intelligence company, DeepSeek, has initiated development of proprietary semiconductor technology, according to three individuals with direct knowledge of the initiative. The processor targets inference operations—the computational process that enables AI models to generate outputs—as opposed to the training phase where models initially learn.
DeepSeek has declined to issue any official statement regarding these reports.
Shares of Nvidia experienced approximately a 2% downturn during premarket trading after the information surfaced.
The Technical Focus of DeepSeek’s Chip
Inference processors differ significantly from the high-performance GPUs required for AI model training. These specialized chips generally cost less and consume considerably less power. With artificial intelligence applications proliferating across industries, the market for inference computation is experiencing explosive expansion.
DeepSeek’s processor would directly address this burgeoning market segment.
The initiative remains in preliminary development phases. DeepSeek has initiated conversations with semiconductor design houses, manufacturing foundries, and memory technology providers. Additionally, the firm has been recruiting chip engineering talent through private channels rather than public job postings, according to two individuals familiar with the hiring strategy.
Strategic Reasoning Behind the Development
DeepSeek has historically depended on hardware from both Nvidia and Huawei to develop and operate its artificial intelligence systems. American export restrictions prohibit Chinese enterprises from purchasing Nvidia’s cutting-edge processors. These limitations have compelled DeepSeek to increasingly utilize Huawei’s Ascend processor lineup in recent periods.
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 release.
However, Huawei’s dominance over China’s $50 billion AI semiconductor sector faces mounting challenges. Technology giants Alibaba and Baidu have both launched internal chip development programs and are capturing market territory.
DeepSeek’s semiconductor venture would position it alongside these competitors. OpenAI introduced its inaugural custom inference processor last month, dubbed Jalapeno, manufactured in partnership with Broadcom. Anthropic has similarly explored proprietary chip development possibilities.
For DeepSeek, additional obstacles exist. American regulations prevent Chinese semiconductor designers from accessing the most sophisticated international manufacturing facilities. Additional restrictions constrain availability of high-bandwidth memory—a critical component for inference chip functionality.
Liang Wenfeng, DeepSeek’s founder, acknowledged during a 2024 conversation that export regulations presented genuine operational difficulties for the organization.
The semiconductor initiative coincides with DeepSeek’s decision to accept external capital for the first time in its history. The enterprise was positioned to secure $7 billion in financing at a valuation ranging between $52 billion and $59 billion, Reuters disclosed in June. This represents a dramatic departure from years of rejecting outside investment.
DeepSeek captured worldwide attention during early 2025 when 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 closely observed AI companies, the organization has maintained a deliberately understated public presence. The ultimate success of its semiconductor endeavor will hinge on overcoming both engineering complexities and persistent American regulatory constraints.


