The company built its own file system, container runtime, scheduler and networking, he wrote. Its capacity pool now spans 20-plus clouds. Functions, its first building block, lets developers write code that Modal runs, including workloads such as image generation, model training and language-model inference.
Sandboxes become a business of their own
Modal launched Sandboxes in 2023 after noticing that people were using language models to generate code. Bernhardsson said the product now has a $300 million run rate, with uses including background agents, reinforcement learning and coding platforms. He said Modal has created close to 3 billion sandboxes and that the service can scale to a million running concurrently.
The announcement also said virtual machines in Sandboxes are generally available. For training workloads, Bernhardsson highlighted a recently launched multi-node option that gives developers a connected compute cluster without a capacity commitment. On the inference side, he pointed to Modal servers for HTTP applications and features for serving language models.
Funding follows reported usage growth
Bernhardsson said usage had tripled since Modal announced its Series C five months earlier, and that the company had recently crossed $1 billion in annualized run rate.
Modal announced a $750 million Series D at a $15.75 billion post-money valuation, led by Accel. Kleiner Perkins joined the round, alongside existing investors including Menlo Ventures, General Catalyst, Redpoint, Lux Capital and Bain Capital Ventures.
