Moonshot AI, a leading Chinese model builder, has been exploring ways to access Nvidia's advanced Blackwell chips through European data centers, according to sources. This move comes amid U.S. export controls that restrict the direct shipment of these chips to China, but allow remote access via overseas facilities. The White House previously accused Moonshot of accessing Blackwell chips through a facility in Thailand, highlighting the growing challenge for Washington to enforce its AI chip export policies.
Nvidia's dominance in the AI chip market has made its products a critical asset for AI development. While U.S. labs like OpenAI and Anthropic face competition from Chinese developers, Moonshot's interest in European-based GPU clusters underscores the global nature of the AI race. The company is reportedly seeking to lease capacity for both inference and training tasks, with discussions involving clusters of 8,000 or more Blackwell GPUs. These arrangements could provide Moonshot with significant computational power, potentially narrowing the gap with its American counterparts.
The U.S. government's efforts to restrict access to advanced chips have not fully curbed the flow of computing resources to Chinese firms. Remote access through international data centers offers a workaround, and Moonshot's aggressive pursuit of this path reflects its strategic intent to maintain competitive advantage. However, the U.S. is pushing for legislative action, such as the Remote Access Security Act (RASA), to expand export controls to include cloud-based access. Despite these efforts, the global AI landscape continues to evolve, with Chinese models gaining traction due to their lower costs and capabilities.
The financial implications of this situation are significant. Leasing AI compute is a lucrative business, with Chinese firms often willing to pay higher upfront fees to secure access. This has led to discussions about pricing structures that could benefit neocloud providers and financing firms. Meanwhile, Nvidia's stance on the importance of global AI infrastructure underscores the interconnected nature of the industry. As the U.S. and China continue to compete in the AI race, the role of international data centers and cloud services will remain pivotal in shaping the future of artificial intelligence.