Applied Compute raises $350M at $3.25B on a $100M revenue run rate

Applied Compute raises $350M at $3.25B on a $100M revenue run rate

2 November 2026: Applied Compute says it has raised $350 million at a $3.25 billion valuation in a round led by Elad Gil, less than 18 months after it was founded. The company, which helps enterprises train and run their own versions of open-weight models, says it went "from zero to $100 million in annualized revenue run rate" in that time.

Not yet independently verified. The valuation, run rate and customer results are the company's own figures, read on its post alone. Its headquarters and exact founding date are not stated on the page. We will update this when it can be confirmed, and remove this note.

A long investor list with customers on it

Kleiner Perkins, Lux Capital, Hanabi Capital, NVIDIA, Blackstone, Shopify Ventures and HubSpot Ventures took part, along with Diffusion, Definition, Timeless, Layer Global, Neo, Abstract, Nova, Quantum Light, First Harmonic and Swish. Shopify is also named as a customer, so at least one investor sits on both sides of the relationship.

The company's post, signed by founders Yash, Rhythm and Linden, says the money will "expand our compute capacity, grow our team, and accelerate deployment in the enterprise". It gives no breakdown between those three. The team has tripled since January, and the company says 85% are researchers and engineers.

What AC2 sells to enterprises

The product behind the numbers is AC2, the Applied Compute Agent Cloud, introduced three months ago. It trains, serves and continuously improves open-weight models for a customer's own work, rather than sending everything to a frontier model through an API. The company counts more than 1,700 training runs since July, says AC Inference serves "more than 1 trillion tokens a day", and says 40% of customers use both training and inference.

Named customers are Microsoft, Shopify, Mercor, Harvey, DoorDash, Cognition, Clay and Decagon. The number of enterprises served has "more than doubled" since the previous round, although the post does not say how many that is.

The savings claims behind the pitch

Two case figures carry the argument. The company says it built Harvey a Review Table model "at less than half the inference cost of a frontier alternative", and that a custom router cut Clay's production inference costs "by up to 47%". Both come with the usual limit that they describe the best cases a vendor chooses to publish.

Still, the pitch is a plain one for finance teams. Companies whose AI bills have grown with usage are being offered a way to swap per-token frontier pricing for a model they control, in exchange for depending on a young supplier and keeping up with model updates. A buyer comparing this route should ask what happens to a trained model if the contract ends, and whether quoted savings hold once training costs are counted.

Our machine learning software category lists seven products, and artificial intelligence software lists seventeen.

Company profile on TrustList: Applied Compute

Sources

• Applied Compute: $100M run rate and fundraise, 2 November 2026, 2 November 2026