Chinese AI startup DeepSeek is working on its own semiconductor, according to three sources familiar with the matter. The chip is designed for inference — the process of running an AI model to generate responses — rather than training new models from scratch.
DeepSeek is reportedly developing its own AI chip, per Reuters.
The chip is being designed for inference, not training, as DeepSeek looks to reduce reliance on Nvidia and Huawei chips.
DeepSeek is reportedly in talks with foundries, memory suppliers and chip-design firms, while… pic.twitter.com/xi3kFertIC
— Wall St Engine (@wallstengine) July 7, 2026
The company has not commented publicly on the report.
Nvidia shares fell about 2% in premarket trading following the news.
Inference chips are different from the powerful GPUs used to train AI models. They are typically cheaper and use less energy. As AI apps become more widespread, demand for inference computing is growing fast.
DeepSeek’s chip would target that demand directly.
The effort is still in its early stages. DeepSeek has been reaching out to chip-design firms, foundries, and memory companies. The company has also been quietly hiring chip engineers without posting jobs publicly, two sources said.
DeepSeek has relied on both Nvidia and Huawei chips to build and run its models. US export controls ban Chinese companies from buying Nvidia’s most advanced processors. That has pushed DeepSeek to lean more on Huawei’s Ascend chips in recent months.
In April, DeepSeek released its V4 model adapted for Huawei’s Ascend hardware. Orders for Huawei’s Ascend 950 chips surged after that launch.
But Huawei’s grip on China’s $50 billion AI chip market is already being tested. Alibaba and Baidu are both developing their own chips and taking market share.
DeepSeek building its own chip would put it in similar company. OpenAI last month launched its first custom inference chip, called Jalapeno, made with Broadcom. Anthropic has also been weighing its own chip development.
For DeepSeek, there are added hurdles. US rules block Chinese chip designers from using the most advanced overseas chip factories. Separate restrictions limit access to high-bandwidth memory, a key component for inference chips.
DeepSeek’s founder, Liang Wenfeng, said in a 2024 interview that export controls were a real challenge for the company.
The chip push comes as DeepSeek also opens up to outside investment for the first time. The company was set to raise $7 billion in a funding round valuing it at between $52 billion and $59 billion, Reuters reported in June. That reverses years of refusing external capital.
DeepSeek gained global attention in early 2025 when its R1 reasoning model went viral. The model was trained on Nvidia’s H800 chip, a version designed for China that Washington later banned.
The company has kept a low profile despite becoming one of China’s most-watched AI firms. Whether its chip effort succeeds will depend on navigating both technical challenges and ongoing US restrictions.
The post China’s DeepSeek Is Making Its Own AI Chip — Nvidia Shares Dropped 2% appeared first on CoinCentral.
