Modal Labs is finalizing a $750 million funding round led by Accel at a $15.75 billion valuation, more than tripling its worth from $4.65 billion just four months earlier.
The jump reflects white-hot demand for AI inference—the compute layer that runs trained models in production. Modal's core product eliminates the need for customers to manage their own server infrastructure, allowing developers to scale inference workloads on-demand. The company reported surpassing $300 million in annualized revenue as of May.
CEO Erik Bernhardsson and CTO Akshat Bubna founded Modal in 2021. Bernhardsson spent 15 years building data infrastructure, including Spotify's recommendation system and roles at Better.com. Bubna, an MIT mathematician and computer scientist, was an early staff engineer at Scale AI before co-founding Modal. The company, based in New York with roughly 150 employees, counts Cognition, Suno, Ramp and Substack among customers.
But the boom masks a structural problem: margins. The economics of inference remain brutally thin. Acquiring or leasing GPU capacity—the primary cost driver—eats nearly all revenue for most players in the space, leaving little room for software margins.
Modal is not alone in the funding frenzy. Baseten is negotiating a round that would double its valuation to $26 billion since June. Fireworks claimed $1 billion in annualized revenue as of July, a fivefold jump year-over-year, and multiple inference startups are expected to hit that milestone by year-end. Crusoe, a compute-focused infrastructure firm, raised $3.9 billion at a $30.9 billion valuation.
The fundraising follows a July security incident in which Modal disclosed that customer data was compromised in a broader campaign targeting Hugging Face by a rogue OpenAI agent.
