DeepSeek Reportedly Develops In-House AI Inference Chip in Secret

News Editor

2026-07-07 10:42:08

DeepSeek, a Chinese AI foundation model company, is reportedly working on a proprietary AI inference chip, according to Reuters citing people familiar with the matter. The project was said to have started around a year ago and is focused on inference workloads rather than model training. Reuters also reported that DeepSeek has been recruiting chip design engineers in recent months to move the effort forward. At present, the company relies mainly on chips from Nvidia and Huawei for both model training and inference. If the in-house inference chip progresses successfully, it could help DeepSeek reduce dependence on external suppliers and improve control over hardware costs. However, the initiative remains at an early stage. The report noted that the project may also face constraints tied to U.S. export controls, particularly in chip manufacturing and access to memory components.

DeepSeekAI chipsInference chipNvidiaHuaweiExport controlsLarge language models

Project said to focus on inference rather than training

DeepSeek, a Chinese AI large-model company, is secretly developing its own AI inference chip, according to Reuters, which cited people familiar with the matter. The project reportedly began about a year ago and is aimed primarily at inference workloads instead of model training. That positioning suggests the effort is designed more for deployment efficiency and real-world model serving than for the heavy compute requirements of large-scale pretraining.

Company has reportedly been hiring chip design engineers

The report said DeepSeek has continued recruiting chip design engineers over the past several months to support the development program. The hiring activity indicates that the company is allocating ongoing resources to build more internal hardware capabilities. For major AI model developers, a successful in-house inference chip can potentially improve deployment efficiency, strengthen supply-chain resilience, and provide better control over per-unit computing costs.

DeepSeek still relies on Nvidia and Huawei chips

At present, DeepSeek mainly depends on chips from Nvidia and Huawei for both model training and inference. If its self-developed inference chip is eventually brought into production and deployment, the company could lower its reliance on outside suppliers while gaining more flexibility in hardware cost management. Even so, the project remains in an early stage, and the Reuters report noted that it may face practical obstacles created by U.S. export controls.

Manufacturing and memory access remain key constraints

According to the report, one of the main external challenges involves restrictions related to chip manufacturing and memory procurement under the current U.S. export control framework. Those constraints could affect both production planning and the broader feasibility of scaling an in-house inference chip program. As a result, while the project points to DeepSeek’s push toward greater hardware independence, its eventual outcome remains uncertain at this stage.

This article was originally published by Bit.Fan. For more cryptocurrency news and market insights, visit www.bit.fan.

300

Disclaimer:

The market information, project data, and third-party content displayed on this platform are for industry information sharing only and do not constitute any form of investment advice or return commitment.

Cryptocurrency trading carries high risks. Users should fully assess their risk tolerance and make independent decisions. All profits, losses, and legal responsibilities are borne by the users themselves.