Applied Compute raises $350 million to expand cloud platform

Applied Compute has raised $350 million at a $3.25 billion valuation, with plans to expand its market presence and buy more infrastructure. Elad Gil led the round, joined by Nvidia, Shopify Ventures, HubSpot Ventures and other investors. The company’s annualized recurring revenue has topped $100 million less than 18 months after launch. Customers include Microsoft and DoorDash.

The startup provides a cloud platform for training large language models and running inference, backed by machine learning researchers who help customers with tasks such as model quality evaluations. Its asynchronous reinforcement learning approach allows training data generation and model weight adjustments to run simultaneously, speeding development. The platform also supports self-distillation, removing the need for teacher models to supervise training, while a built-in AI assistant automates repetitive work such as summarizing skills learned during training runs.

For inference, an autoscaling engine adjusts hardware allocation to match traffic, while online reinforcement learning periodically fine-tunes models using incoming prompts. Its Shift-1 router directs coding-agent requests to the most cost-efficient model capable of handling them and can divide complex tasks among multiple models.