DeepSeek Bets Big on Huawei Chips to Bypass U.S. Export Controls
Source: The Information (opens in a new tab)
Intel Summary
DeepSeek CEO Liang Wenfeng informed investors that transitioning to domestic silicon for AI model training is a top priority, with Huawei Technologies expected to begin delivering training chips as early as the fourth quarter, according to reporting by The Information. The move aims to circumvent restrictions imposed by U.S. export controls.
Why It Matters
Transitioning compute-heavy model training workloads to domestic hardware tests whether non-U.S. silicon can support frontier-class AI development. Successful deployment would validate alternative AI hardware ecosystems and reduce reliance on restricted foreign semiconductor supply chains.
Part of an ongoing development
Additional reportingDeepSeek transitions to Huawei chips for AI model training
DeepSeek CEO Liang Wenfeng informed investors that transitioning to domestic silicon for AI model training is a top priority, with Huawei Technologies expected to begin delivering training chips as early as the fourth quarter, according to reporting by The Information. The move aims to circumvent restrictions imposed by U.S. export controls. Claims are as reported; this summary makes no determination about accuracy or significance.
- Confidence
- Low confidence
- Corroboration
- Unconfirmed
Organizations & Entities
Related Intelligence
- DevelopmentDevelopingAlso involving DeepSeek
Anthropic details distillation campaigns from Alibaba, Moonshot AI, and DeepSeek
TechCrunch reports that Anthropic has published a report alleging persistent model distillation campaigns conducted by China-based AI companies, specifically citing Alibaba, Moonshot AI, and DeepSeek. Claims are as reported; this summary makes no determination about accuracy or significance.
2 independent sources - DevelopmentDevelopingAlso involving DeepSeek
DeepSeek released V4.1-Flash
DeepSeek has released V4.1-Flash, a multimodal model containing 552 billion total parameters with 16 billion active parameters per token. According to benchmark results reported from DeepSeek, V4.1-Flash narrowly outperforms Opus 5 and GPT-5.6 Sol on the DeepSWE coding benchmark and is published under an open MIT license. Claims are as reported; this summary makes no determination about accuracy or significance.
2 independent sources - DevelopmentNewAlso involving DeepSeek
China investigates DeepSeek and Moonshot AI over data leaks to Anthropic
China's internet regulator has opened an investigation into AI developers DeepSeek and Moonshot AI regarding potential unauthorized data routing to Anthropic's Claude models, according to The Information. The probe follows a 154-page report published by Anthropic alleging widespread illicit Claude usage by seven Chinese firms, including an instance where DeepSeek reportedly forwarded prompts from engineers building a police surveillance system. Claims are as reported; this summary makes no determination about accuracy or significance.
- ReportAlso involving DeepSeek
Intelligence agencies warn of China’s large-scale AI model distillation efforts
Three intelligence agencies issued a joint warning stating that Chinese companies—including DeepSeek, Moonshot AI, Alibaba, MiniMax, StepFun, and Z.AI—have used model distillation tactics to extract billions of tokens from U.S. AI models.
Nextgov/FCW (AI)