AI news ·
Applied Compute raises $350 million after reaching $100 million in annualized recurring revenue
AI infrastructure startup Applied Compute raised $350 million at a $3.25 billion valuation after hitting $100 million in annualized recurring revenue.
Applied Compute Inc., an AI infrastructure startup less than 18 months old, has raised $350 million in fresh funding at a $3.25 billion valuation after crossing $100 million in annualized recurring revenue. The round, led by prominent angel investor Elad Gil with participation from Nvidia, Shopify Ventures, and HubSpot Ventures, signals investor appetite for platforms that tackle the practical bottlenecks in training and running large language models.
The technical approach
Applied Compute operates a cloud platform for training LLMs and running inference workloads. Its software stack departs from the synchronous training methods common in the industry. Standard setups run data generation and weight adjustment components in lockstep, forcing each to wait for the other to finish. Applied Compute uses asynchronous reinforcement learning, which lets both components run simultaneously and cuts overall training time.
The platform also supports self-distillation, removing the need for separate teacher models that researchers typically use to supervise training runs. A built-in AI assistant handles repetitive tasks like summarizing what new skills an LLM acquired during a training run - work that normally requires manually reviewing lengthy reasoning logs.
Inference and routing features
On the inference side, an autoscaling engine adjusts hardware allocation based on user traffic. An online RL feature periodically fine-tunes model weights using the prompts the LLM receives, keeping outputs aligned with customer requirements. The company also offers a model router called Shift-1 that directs each user request to the most cost-efficient LLM for the job and can split complex tasks across multiple models.
Applied Compute counts Microsoft, DoorDash, and other major tech firms among its customers. The company said it will use the new capital to expand its market presence and buy more infrastructure.
The bigger stack
"Training, inference, and routing are the first pieces of a much larger stack," founders Rhythm Garg, Yash Patil, and Linden Li wrote in a blog post. "We're building the AI infrastructure every company will need to fully own their intelligence, from the models themselves to the systems that run, monitor and secure them."
The company also employs a team of machine learning researchers who work directly with customers on tasks such as creating quality evaluations for newly developed models - blending infrastructure-as-a-service with hands-on technical consulting.
Why this matters for executives and strategy leaders
Applied Compute's revenue trajectory - $100 million ARR inside 18 months - suggests enterprises are willing to pay for infrastructure that reduces LLM training time and inference costs. For executives evaluating build-versus-buy decisions in AI, platforms that combine training acceleration with model routing and autoscaling can shift budget conversations from raw compute spend to per-task efficiency. The involvement of Nvidia as an investor also signals where hardware ecosystems are placing bets beyond chip sales, which procurement and partnership teams should track.