Applied Compute raises $350M at $3.25B valuation after hitting $100M ARR

Funding round details

Applied Compute Inc., an AI cloud infrastructure startup, has secured $350 million in funding at a $3.25 billion valuation. The round was led by angel investor Elad Gil, with participation from Nvidia Corp., Shopify Ventures, HubSpot Ventures, and other undisclosed backers. SiliconANGLE first reported the deal on Monday.

How the platform works

The startup’s platform provides cloud-based infrastructure for training large language models (LLMs) and running inference workloads. Applied Compute also offers machine learning research support to help customers optimize model development and evaluation.

Its software stack introduces two efficiency improvements: asynchronous reinforcement learning (RL) training and self-distillation. Asynchronous RL allows data generation and model weight adjustment to run simultaneously, reducing training time. Self-distillation eliminates the need for teacher models, further streamlining the process.

Applied Compute’s Shift-1 model router automatically directs user requests to the most cost-effective LLM for each task. It also splits complex workloads across multiple models, which the company says lowers inference costs for customers like Microsoft and DoorDash.

Revenue growth and plans for capital

Applied Compute launched less than 18 months ago and has already surpassed $100 million in annualized recurring revenue. The funding announcement confirmed that Microsoft and DoorDash are among its customers.

Proceeds from the round will be used to scale infrastructure purchases and expand market reach, addressing demand for both training and inference workloads.

How this compares to earlier AI infrastructure stories

Earlier reporting on AI infrastructure startups has focused on partnerships like Dell and AMD’s rack-scale solutions or Red Hat’s integration of Nvidia’s OpenShell for agent security. Applied Compute’s rapid revenue growth and $350 million raise stand out for targeting the full stack—training, inference, and routing—while achieving enterprise adoption faster than most cloud upstarts.

Unlike the incremental updates seen in projects like Krafton-backed Ludo Robotics, Applied Compute’s asynchronous RL and self-distillation claim to cut training cycles materially, potentially compressing time-to-production for teams building custom LLMs.

What this means

Lazyfounder analysis — our interpretation, not reported fact.

Applied Compute’s $3.25 billion valuation and $100 million revenue run-rate underscore how quickly capital is flowing into infrastructure startups that combine cutting-edge training methods (asynchronous RL, self-distillation) with off-the-shelf hardware. For operators building or buying AI stacks, the funding signals two things: (1) cloud providers like Microsoft are already treating Applied Compute as a viable alternative to in-house clusters, and (2) the $350 million war-chest will force incumbents to price more aggressively, narrowing margins for everyone else in the training-as-a-service market.

Key takeaways

• The $3.25 billion valuation sets a new benchmark for AI infrastructure startups targeting enterprise workloads.
• Microsoft and DoorDash’s adoption shows Applied Compute’s platform is already deployable at scale.
• Asynchronous RL and self-distillation claim to cut training time; if replicable, this could reset expectations for ROI on model development.
• With Shopify and HubSpot Ventures in the round, e-commerce and CRM platforms may soon offer Applied Compute’s stack as a managed add-on.

Related on Lazyfounder

• Nvidia Corp.
• Microsoft Corp.
• More startup news
• Dell and AMD partner to build rack-scale AI infrastructure for enterprises
• Red Hat integrates Nvidia’s OpenShell into AI platform to secure autonomous agents