Modal Labs, a company that lets developers run AI models without managing their own servers, is closing in on a $750 million funding round led by Accel at a $15.75 billion valuation, according to a person with knowledge of the financing. The size of the round had not previously been reported.
The deal would value the New York-based company at more than triple the $4.65 billion it reached just four months ago, when it announced a $355 million round. That pace of revaluation is striking even by the standards of an AI market that has repriced its winners repeatedly over the past two years. Modal Labs declined to comment.
The money is chasing a specific kind of business. Inference, the work of running an already-trained model to generate answers, has become one of the fastest-growing layers of the AI stack, driven in particular by customers that rely on open-source models rather than the closed systems sold by OpenAI and Anthropic. Those customers need somewhere to run their models, and Modal sells exactly that.
The economics are harsher than the demand would suggest. Inference providers rent graphics processors from cloud companies and resell that compute to developers, a business with thin margins and heavy capital costs. The rent does not get cheaper when demand spikes; it gets scarcer. Fireworks, a rival, said in July that it had reached a $1 billion annualized revenue run rate, a figure that measures the scale of the business without saying much about its profitability.
The inference layer sits below the models that get the attention. When a company builds on an open model, it still has to buy or rent the hardware that actually generates each answer, and that work is increasingly outsourced to specialists. Modal’s pitch is that a developer should be able to deploy a model to production in a few lines of code, scale it up and down with demand, and never think about the servers underneath.
Modal sits in a crowded and fast-moving field. Baseten has raised money at a $26 billion valuation, and Fireworks and Fal are both in talks with investors, according to people familiar with those discussions. The sector is attracting capital on the assumption that inference demand will keep compounding, and that the companies that win early developer trust will hold it as the market matures.
Modal has spent the past several years building that trust. Founded in 2021, the company now has about 150 employees, and its customers include Cognition, Suno, Ramp and Substack, a list that spans coding tools, AI-generated music, fintech and publishing. The breadth of that roster is part of the pitch: inference is a horizontal need, not a niche one, and whoever serves it well can sell into many industries at once.
The company’s growth has not been entirely smooth. In late July it disclosed that a customer’s data had been compromised in a broader hacking campaign, which its chief technology officer attributed to a flaw in the customer’s own code rather than Modal’s systems. The incident surfaced during the fundraising process and did not appear to slow the round, a measure of how eager investors are for exposure to inference infrastructure.
The open-source movement is what feeds the demand. Models from Meta, Mistral and DeepSeek, among others, have made capable AI available to anyone who can rent the hardware to serve it, and a growing share of that serving work lands on platforms like Modal. The more companies choose open models over proprietary ones, the more the infrastructure layer underneath them gains value.
Analysts said the revaluation reflects more than one company’s momentum. The gap between the $4.65 billion of May and the $15.75 billion of September is a bet on the whole category, and on the idea that the demand for running open-source models will keep outgrowing the supply of places to run them. It is also a bet that thin margins can be fixed with scale, a proposition that remains unproven at this size.
The valuation math is the part investors are still debating. At $15.75 billion, Modal is being priced against a future in which inference is a large, recurring revenue stream rather than a thin-margin rental business. Whether that future arrives depends on whether the company can keep its developer base growing while its competitors, several of them richer, chase the same customers.
What Modal has that its investors are paying for is a developer base that already depends on it. The company’s customers write their applications against Modal’s platform, and switching costs rise the longer they stay. In an infrastructure business, that stickiness is worth more than any single quarter of revenue, and it is the asset the new round is priced around.
The round is not closed, and a person familiar with the matter said terms could still shift before the financing is final. If it lands as described, Modal will join a short list of infrastructure companies valued like the models they host, and the pressure to convert a developer following into a durable, profitable business will arrive with the money.
