Modal gives AI developers a serverless place to run models, training jobs and AI agents without managing their own computing clusters. Axios reports that Accel is set to lead a new financing at a roughly $15 billion valuation, though the transaction remains in talks and no round size was disclosed.

Developers send Modal code; its service packages the work in containers, places it on pooled cloud capacity and automatically adds or removes containers as demand changes. The same platform handles low-latency inference, model training, highly parallel batch jobs and isolated sandboxes where AI agents can execute generated code.

The product combines compute access with an orchestration layer. Customers do not need to reserve a fixed block of GPUs or maintain Kubernetes themselves; Modal charges by the second while a workload runs. The company says users span generative media, computational biology and software development, with Cognition using the platform for reinforcement-learning infrastructure and production inference.

The proposed valuation would mark another sharp reset. Modal raised $355 million in May at a $4.65 billion post-money valuation in a round led by General Catalyst and Redpoint; Accel joined that financing. At roughly $15 billion, the new talks place the company at more than three times that value about four months later.

The business sits between AI developers and the underlying clouds. Modal says it pools GPU capacity across multiple providers, while its own file system, container runtime and scheduler decide how customers' code starts and scales. Teams get one programmable layer for workloads that can jump from no activity to thousands of processors.

The talks would deepen an existing investor relationship rather than add an entirely new backer: Accel participated in the May round. A valuation under discussion is not a closed financing, and the eventual amount and terms could still differ.