Overview
The Chinese AI lab DeepSeek has chosen not to share its upcoming flagship model, expected around the Lunar New Year, with the usual US hardware partners Nvidia and AMD. Instead, it granted early access to domestic Chinese chipmakers, notably Huawei. This deviation from industry practice raises questions about compliance with US export controls and the broader geopolitical contest over advanced AI chips.
Key Developments
- Early access to V4 model was provided to Chinese chipmakers weeks before the public launch, while US firms were excluded.
- The model’s release was slated for the Lunar New Year holiday of 2026.
- US chipmakers Nvidia and AMD declined to comment on the decision.
- According to a senior Trump‑administration official, the previous DeepSeek model was trained on Nvidia’s Blackwell chip in a mainland‑China data centre, potentially breaching US export rules.
- DeepSeek may attempt to mask the use of US chips and claim training on Huawei’s processors.
Important Facts
Since its launch in January 2025, DeepSeek’s models have been downloaded over 75 million times from the open‑source repository Hugging Face. Chinese open‑source models now lead global download statistics, intensifying US debates on the export of advanced AI chips. Last year, the US allowed limited shipments of Nvidia’s H20 and AMD’s MI308, while keeping newer, more powerful processors under restriction.
UPSC Relevance
This episode touches upon several GS topics: technology policy and strategic autonomy (GS3), the impact of **export controls** on international trade and security (GS3), and the **self‑reliance drive** of the Indian government in AI and semiconductor sectors (GS3). Understanding the dynamics of US‑China tech rivalry helps aspirants analyse future policy directions, such as India’s push for indigenous chip design and the possible replication of export‑control frameworks.
Way Forward
Policymakers may consider:
- Strengthening domestic AI‑chip ecosystems to reduce dependence on US hardware.
- Formulating clear guidelines for open‑source AI model sharing that balance innovation with national security.
- Monitoring compliance of foreign AI firms with export‑control regimes, possibly extending licensing mechanisms similar to those applied to Nvidia’s H20 and AMD’s MI308.
For UPSC candidates, tracking such developments offers insight into how technology, trade, and geopolitics intersect in contemporary governance.