GMI Cloud Raises $668 Million to Boost AI Inference Platform

GMI Cloud has closed a $668 million funding round, split between $223 million in Series B equity and $445 million in debt facilities, according to ChainCatcher. The round was led by ARCHIV with participation from Nvidia.

The company provides a platform dedicated to running AI models—the stage where a model responds to user queries after completing its training. The company says its platform processes nearly four trillion tokens weekly, a usage metric underscoring the massive demand for compute power.

Its client list includes Fireworks, Nous Research, OpenRouter, and Trend Micro—entities that rely on external platforms to run their models instead of building their own infrastructure. This list reflects growing demand for turnkey inference services among companies seeking to avoid the cost of building private data centers.

The financing ranks among the largest rounds in the sector, and Nvidia's participation serves as an indicator of the inference layer's importance in the AI value chain. The main challenge for these companies remains competing for compute capacity and electricity, alongside pricing pressure as major players enter the market.

What do these terms mean?

Inference: The stage where a model is used after training to generate answers to direct requests, usually the phase users pay for. Series B: A funding stage following Series A, typically financing company expansion rather than its founding. Debt facilities: Loans or credit lines secured by the company to be repaid with interest, differing from equity which gives investors an ownership stake. Token: The smallest unit of text processed by a model, used weekly as an indicator of demand volume.