Skip to content
NewA calmer, faster ZeroHour

Research claims banned Nvidia chips trained China’s leading open models

Original title: Export-banned Nvidia chips are behind China’s most powerful AI models—uh oh, Jensen!

TechCrunch · AIby Tim Fernholz ()4 min read1
Part of a story covered by 4 sourcesWeight study finds Blackwell fingerprints in six Chinese modelsMerged summary & timeline
AI summary · x-ai/grok-4.7Summaries by Vani AI

A former OpenAI researcher says Kimi K3 and DeepSeek V4 weights show Blackwell fingerprints; Jensen Huang left Nvidia this week.

Unemployment Labs’ Will DePue says weights from DeepSeek, Moonshot AI, MiniMax and Ant Group reflect training on export-banned Blackwell GPUs. Nvidia CEO Jensen Huang left this week after admitting an affair with Megaspeed executive Alice Huang, whose firm is implicated in chip smuggling. US rules ban Blackwell exports to China but allow H200 shipments under a 25% tariff. Epoch AI estimates about two million advanced chips entered China in 2024 and 2025.

Unconfirmed: The fingerprint analysis and smuggling claims are not confirmed; chips may have been used in offshore data centers.

Key facts
  • DePue cited Moonshot’s Kimi K3 and DeepSeek’s V4 as showing Blackwell fingerprints.
  • White House adviser Michael Kratsios earlier said Moonshot used banned chips, citing US intelligence.
  • Epoch estimates 275,000 Nvidia AI accelerators were imported to Malaysia in 2024 and 2025.
  • About 600,000 H200 chips are reportedly stockpiled in Hong Kong awaiting import clearance.
Explain this AI news with Vani AIWhat was announced, what is new and why it matters, in a new chat

From the original article

4 min read at techcrunch.com

New research shows the most advanced AI models from China were trained on banned Nvidia chips.

TechCrunch · AI

By Tim Fernholz · techcrunch.com

Read the full article at TechCrunch · AI

A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.