• Chinese AI startup Moonshot developed its Kimi K3 model using a cluster of around 20,000 Nvidia GPUs provided by Alibaba, according to people familiar with the matter.
  • The partnership underscores Alibaba's growing role as a key supplier of AI infrastructure in China, even as US export controls limit access to cutting-edge chips.
  • Moonshot's rapid scaling highlights the intensifying race among Chinese AI developers to achieve frontier-level capabilities.

A Compute-Powered Leap

Moonshot AI, an Alibaba-backed startup, has trained its latest Kimi K3 model on a massive computing cluster supplied by Alibaba, according to people familiar with the matter. The cluster, comprising roughly 20,000 Nvidia GPUs, provided the computational firepower needed to train a model that the company claims rivals top US systems.

This collaboration illustrates a pivotal shift in China's AI landscape: rather than relying solely on sanctioned advanced chips, domestic players are increasingly leveraging cloud providers like Alibaba to access large-scale compute. Alibaba's investment in Moonshot, including a 36% stake from prior funding rounds, positions its cloud arm as a critical enabler of frontier AI development.

Export Controls and Workarounds

Moonshot's achievement comes amid stringent US export controls on advanced GPUs, including Nvidia's Blackwell series. While the Kimi K3 was trained on older-generation chips, reports indicate Moonshot is seeking access to Blackwell processors for future models. Alibaba's ability to procure and deploy high-performance GPUs—albeit within regulatory limits—has become a strategic advantage.

"Without access to Alibaba's cluster, Moonshot would have faced significant hurdles in scaling its model," said an industry analyst who asked not to be named. "This partnership showcases how domestic cloud providers are stepping up to fill the compute gap."

A Bellwether for China's AI Ambitions

Moonshot's rise mirrors broader efforts in China to close the AI capability gap with the US. The company's rapid iteration cycle—launching Kimi K3 just months after its predecessor—signals an aggressive push for leadership. Industry observers note that while Chinese firms face hardware constraints, they are compensating with software innovations and efficient training techniques.

"The collaboration between Moonshot and Alibaba is a testament to the growing maturity of China's AI ecosystem," said a spokesperson for Alibaba Cloud, who did not provide specific details about the GPU cluster.

Implications for Cloud and AI Competition

As Moonshot scales, Alibaba Cloud stands to benefit from increased demand for AI infrastructure, both from Moonshot and other startups. This trend could reshape the competitive dynamics among Chinese cloud providers, with Alibaba positioning itself as the go-to platform for large-scale AI training.

However, experts caution that reliance on foreign-derived chips remains a vulnerability. Any tightening of export controls could disrupt Moonshot's roadmap. "The geopolitical environment adds a layer of uncertainty," noted the analyst. "But for now, Alibaba's cluster provides a lifeline."

Moonshot has not publicly commented on the specifics of its compute resources, but the company's rapid trajectory suggests it remains undeterred by the challenges. As the AI race accelerates, all eyes will be on how Moonshot and its partners navigate the complex interplay of technology, policy, and commerce.

This article was updated to clarify the number of GPUs in the cluster and to include comments from Alibaba Cloud.