Moonshot AI Targets European Data Centers to Access Nvidia Blackwell Chips

Moonshot AI, the Chinese lab behind the Kimi AI model, is reportedly turning to European data centers to secure access to Nvidia's latest Blackwell AI GPUs. The company is allegedly in talks to lease clusters containing at least 8,000 GPUs to support model inference and training.

• Moonshot AI is looking to lease 8,000+ Nvidia Blackwell GPUs from European data center providers.
• The strategy aims to bypass U.S. chip export bans by utilizing cloud-based remote access.
• The U.S. House has passed the Remote Access Security Act to potentially close this loophole.
• Nvidia has advised its clients to limit service to Chinese outfits to under 49% of capacity.

The 'Cloud Loophole' and U.S. Regulations

Current U.S. export controls strictly prohibit the sale of high-performance AI chips to China. However, these bans do not prevent Chinese labs from renting compute capacity via foreign data centers. This exploit has allowed companies like Moonshot AI to maintain their competitive edge in the global AI race by leveraging hardware they cannot legally acquire themselves.

The proposed Remote Access Security Act is a direct response to this exploitation. While it has passed the House, it still requires Senate approval and presidential signing. Until then, Chinese firms are racing to lock in as much compute as possible, knowing their access window is rapidly closing due to both legislative pressure and industry compliance.

Nvidia's Strategic Compliance

Nvidia has taken steps to mitigate regulatory risks by instructing its data center partners to ensure that no more than 49% of their capacity is allocated to Chinese entities. This proactive measure highlights the growing tension between global tech supply chains and geopolitical restrictions, signaling that the era of easy cloud access for restricted nations is coming to an end.

Implications for the AI Ecosystem

For global AI developers and tech businesses, this situation underscores the fragility of relying on cloud infrastructure subject to geopolitical shifts. Organizations must now account for extreme regulatory risks when planning long-term AI model training. The tightening of these rules suggests that localized compute sovereignty will become a critical differentiator in the next phase of the AI development cycle.