Databricks CEO Ali Ghodsi said the company is running out of GPU capacity while hosting open-source AI models such as Kimi [1].

This shortage highlights the immense computational pressure facing cloud and data platforms as the industry shifts toward open-source artificial intelligence. The struggle to secure enough hardware suggests that demand for flexible, open models is outpacing the current global supply of high-end processors.

Ghodsi said these details during an interview on CNBC's program “The Exchange” [1]. He said that the company is actively expanding its AI services to accommodate a growing variety of models. The surge in demand is specifically linked to the hosting of open-source options, which allow developers more customization than proprietary systems.

"We're hosting open source models like Kimi and running out of GPUs," Ghodsi said [1].

The hardware bottleneck comes at a time when enterprises are increasingly moving away from a single-vendor approach to AI. By supporting models like Kimi, Databricks aims to provide a more diverse ecosystem for its users, though this strategy requires significant physical infrastructure.

GPU shortages have become a recurring theme for AI companies attempting to scale. As Databricks integrates more open-source capabilities into its platform, the need for massive amounts of compute power continues to grow. The current capacity strain underscores the ongoing race for hardware dominance in the AI sector [1].

“"We're hosting open source models like Kimi and running out of GPUs,"”

The capacity constraints at Databricks signal a broader industry trend where the democratization of AI through open-source models is creating a hardware crisis. As more companies move away from closed-source ecosystems, the reliance on GPUs creates a critical dependency on hardware manufacturers, potentially limiting the speed of AI deployment for enterprise clients.