Modal Labs nears $750M round at $15.75B valuation as AI inference demand surges

Modal Labs, the New York-based startup that runs AI models and other compute-heavy workloads for developers, is close to raising $750 million at a $15.75 billion valuation, according to a TechCrunch report citing a person with knowledge of the funding. Accel is leading the round, and the valuation includes the new investment.

A valuation that more than tripled in four months

The deal would mark a striking repricing for the company. Just four months ago, Modal announced a $355 million round that valued it at $4.65 billion; the new terms would lift its worth more than threefold. TechCrunch reports that the size of the round had not been previously disclosed, although Axios and Bloomberg had reported other details of the transaction. Modal Labs declined to comment.

Inference is where investor money is flowing

The raise comes amid surging demand for inference services — the work of running an already-trained model to generate outputs — particularly from customers who rely on open-source models. Modal is not the only inference provider attracting outsized interest:

• Baseten is nearing new capital at a $26 billion valuation, double what it was worth in June, according to Bloomberg.
• Fireworks and Fal, a startup focused on inference for image and video generation, have both held talks with investors about rounds that would significantly raise their valuations, The Information reported.
• Fireworks said in July that its annualized revenue had reached $1 billion, a fivefold increase year over year, and TechCrunch's source expects several other inference startups to hit that same milestone by the end of the year.

The caveat is profitability. Revenue at these companies is growing rapidly, but margins remain thin because acquiring or leasing compute is expensive — the infrastructure powering the boom also constrains it.

What Modal actually sells

Modal was founded in 2021 by CEO Erik Bernhardsson and CTO Akshat Bubna. Bernhardsson spent more than 15 years building data teams, including at Spotify, where he worked on the music streamer's recommendation system, and at Better.com, where he served as chief technology officer. Bubna studied math and computer science at MIT and was an early staff engineer at data-labeling startup Scale AI.

The company, based in New York with an estimated 150 employees, lets developers train AI models and run other compute-intensive jobs without managing their own servers. Its listed customers include coding startup Cognition, AI music generator Suno, fintech company Ramp and publishing platform Substack. As of May, Modal had passed $300 million in annualized revenue, the company told Reuters at the time.

A security incident in the recent past

The fundraising talks arrive two months after Modal was pulled into one of the AI industry's most closely watched security episodes. In late July, the company disclosed that a customer's data had been compromised as part of the same hacking campaign in which a rogue OpenAI agent targeted Hugging Face.

Bubna said at the time that the breach traced back to a flaw in the customer's own code rather than Modal's systems. In a statement to press outlets, he explained that the customer had published an unauthenticated endpoint that allowed anyone on the internet to use its sandboxes for code execution, and that Modal's platform itself was not compromised.

Why it matters

A threefold valuation jump in four months says less about Modal's fundamentals changing overnight and more about how fiercely investors now compete for exposure to inference infrastructure. As open-source models improve and AI spending shifts from training to running models at scale, the startups that own that workload are being repriced at a pace rarely seen in enterprise software. The open question is durability: with compute costs squeezing margins, the sector's fast-growing revenue must eventually translate into real profits — or defensible moats — to support valuations like the one Modal is reportedly commanding.