Moonshot AI seeks Nvidia Blackwell access via European neoclouds amid US export curbs
Moonshot AI, a Chinese AI lab accused by the White House of accessing Nvidia’s Blackwell chips remotely in Thailand, is reportedly negotiating to lease Nvidia-powered compute capacity from European neocloud providers. Four sources familiar with the matter told CNBC that Moonshot discussed renting clusters of 8,000 or more Blackwell GPUs for inference and training, with talks beginning after the…
Key points
- Moonshot AI is negotiating to lease Nvidia Blackwell GPU clusters in Europe for inference and training, per four sources
- Talks involve clusters of 8,000+ GPUs, with upfront payments 30–50% higher than non-Chinese clients’ rates
- U.S. export controls block Nvidia’s Blackwell chips from China but allow remote access via overseas data centers
Moonshot, which builds the Kimi model series, has gained traction as a cheaper alternative to U.S. AI labs like OpenAI and Anthropic. Its July release of Kimi K3 reportedly narrowed the gap with leading Western models, though the company paused new subscriptions due to capacity constraints. Meanwhile, U.S. lawmakers have proposed the Remote Access Security Act (RASA) to close the cloud-access loophole, though it stalled in the Senate. Nvidia has urged cloud providers to prioritize non-Chinese customers, citing geopolitical risks to AI infrastructure dominance.
Chinese AI lab Moonshot turned to Europe for Nvidia Blackwell access amid U.S. China curbs, sources tell CNBC
CNBC Technology · 5 November 2026
Loading the full article…
This text was published by CNBC Technology and written by Kai Nicol-Schwarz. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Hardware & Compute
All →- HomeView will support third-party AI chatbots, suggests code · 1 src
- NVIDIA releases CUDA 13.5 with Vera Rubin support and GPU improvements · 1 src
- University of Tokyo researchers develop optical AI framework to cut computation needs · 1 src
- Nvidia’s 600,000 H200 chips stuck in Hong Kong due to China’s import freeze · 5 src
- Nvidia reports Q3 earnings on Nov 17 amid AI market uncertainty · 3 src
Comments
via GitHub Discussions