Ex-OpenAI Trio’s Applied Compute Hits $100M Run Rate

Applied Compute, the custom-model start-up founded by three former OpenAI technical staffers, says it has reached $100 million in annualized revenue run rate less than 18 months after its founding. In a post on 2 November, it also said it has raised $350 million at a $3.25 billion valuation, in a round led by Elad Gil.

By our arithmetic, the new valuation is 32.5 times that run rate. The $100 million run rate is also the valuation Upstarts reported for the whole company when it raised a $20 million seed in June 2025, citing sources; the latest valuation is 2.5 times the $1.3 billion post-money mark of April 2026.

Business Pill · Fine-Tuning

A short explainer on fine-tuning: adapting a general model to a specific task with your own data. It teaches the general idea only and says nothing about Applied Compute.

What Applied Compute Announced

“Less than 18 months after founding Applied Compute, we’ve grown from zero to $100 million in annualized revenue run rate,” the company wrote. Kleiner Perkins, Lux Capital, NVIDIA, Blackstone, Shopify Ventures and HubSpot Ventures are among the other investors it lists in the round.

It says it has more than doubled the number of enterprises it serves since its last fundraise, and names Microsoft, Shopify, Mercor, Harvey, DoorDash, Cognition, Clay and Decagon as customers. The post gives no revenue for any customer, and no figure for total funding.

The Business Behind the Run Rate

Applied Compute builds custom models for companies on top of open-weight models. “We started Applied Compute because we believed companies would want to own and improve the AI powering their products and operations,” the post says.

It gives two customer results. For Harvey, it says it trained a model that improved answer and citation quality “at less than half the inference cost of a frontier alternative”. For Clay, it says a custom router cut production inference costs “by up to 47% without sacrificing quality”.

The company says customers have run “more than 1,700 training runs” on its AC2 platform since July, that its inference service serves more than 1 trillion tokens a day, and that 40% of customers use it for both training and inference. Its team has tripled since January, and 85% are researchers and engineers, it says.

From $100 Million to $3.25 Billion

In June 2025, Upstarts reported, citing sources, that Applied Compute had raised $20 million in a round led by Benchmark partner Victor Lazarte, valuing it at $100 million. It named the founders as Rhythm Garg, Linden Li and Yash Patil, all three recent Stanford students who worked on technical projects at OpenAI.

In April 2026, Applied Compute announced $80 million at a $1.3 billion post-money valuation, led by Kleiner Perkins with Elad Gil among the participants, according to Latham & Watkins, which represented it. The 2 November post puts the new valuation at $3.25 billion; it does not say whether that is pre- or post-money.

By our arithmetic, the new round of $350 million is 3.5 times the $100 million run rate, and the valuation rose 2.5 times in about seven months.

The Thread So Far

On 2 November we reported that Sequoia backs Trajectory at $300M, a start-up in the same post-training market. On 3 November we covered a survey finding that only 10% of AI builders post-train to cut costs, in which 30% said they use specialists such as Applied Compute.

Connecting the Coverage

Our related pieces on post-training:

• Sequoia Backs Trajectory at $300M; Tuned Model Tops Opus 4.6: a rival in the same market.
• Only 10% of AI Builders Post-Train to Cut Costs, Like Cursor: the survey on who post-trains.

The Structural Read

As we read the post, Applied Compute sells both halves of a custom model: training it and serving it, with 40% of customers already using both.

As we read the customer results, the pitch is cost: Harvey’s model at less than half the inference cost of a frontier alternative, and Clay’s router cutting inference costs by up to 47%.

As we read the rounds, each valuation step has been larger in dollars: from $100 million to $1.3 billion, then to $3.25 billion.

Three Implications

FOR APPLIED COMPUTE It says it will use the new funding to expand compute capacity, grow its team and accelerate deployment in the enterprise.

FOR ITS CUSTOMERS It names Microsoft, Shopify, Mercor, Harvey, DoorDash, Cognition, Clay and Decagon as companies using its platform.

FOR THE POST-TRAINING MARKET In our earlier coverage, Trajectory raised at $300 million in the same market; Applied Compute’s $3.25 billion is about 11 times that, by our arithmetic.

The Business Engineer Lens

This story maps onto the Business Engineer framework The Four Intelligence Moats.

The framework lists four moats, including “The Container Moat — built by closed data loops inside a customer’s environment, nascent but the deepest position in the stack.”

As we read it, Applied Compute sells the tools to build that moat: companies train and run models on their own data and own the result. The framework does not discuss Applied Compute.

What Is Not Established

The $100 million run rate and the $3.25 billion valuation are Applied Compute’s own figures; it gives no audited revenue, and a run rate annualizes a recent period. The June 2025 valuation is Upstarts’ report from sources, not a company statement. A post on X from @MTSlive dates an $80 million round to last October; we use the April 2026 announcement reported by Latham & Watkins, and we do not total the funding.

The 32.5x, 3.5x and 2.5x figures are our arithmetic and mix a valuation whose basis is not stated with an annualized run rate. We did not contact Applied Compute.

The Bottom Line

Applied Compute says it reached a $100 million run rate in under 18 months and raised $350 million at $3.25 billion. By our arithmetic, that valuation is 32.5 times the run rate, and the run rate alone now matches what Upstarts reported the whole company was valued at in June 2025.