Fal, the San Francisco company that rents out capacity to run video, image and audio models, closed a $1.3B round led by Bessemer Venture Partners. Nvidia, Sound Ventures and T. Rowe Price joined. The deal doubles its valuation from $8B in April to $16B.
The company had already banked a Series E led by Sequoia, and disclosed both rounds at once: $1.65B raised across the pair. Annualized revenue passed $800M earlier this year, according to Bloomberg’s reporting on the round.
What buyers want is inference, the cheaper half of the AI business. Running an already-trained model for customers who favor open-weight alternatives has become a volume game with thin margins, since compute leases stay expensive.
Rivals are on the same treadmill. Baseten, Fireworks AI and Modal have all raised aggressively at climbing valuations, and none of them escape the cost problem. Every dollar of gross margin depends on how well the operator packs workloads onto GPUs it does not own.
Fal was founded in 2021 by Burkay Gur, a former Coinbase machine learning lead and Oracle engineer, and Gorkem Yurtseven, previously a developer at Amazon. The company has spent five years turning that hosting layer into a distribution channel for generative media models.
The bet behind the valuation is that inference becomes the durable toll road of the AI economy, with training concentrated in a handful of labs and demand for outputs spread across thousands of products.
Whether that holds depends on pricing pressure. As open-weight models improve and serving stacks commoditize, customers gain the option to move. Investors are wagering Fal can stay ahead of that curve long enough to earn back a $16B price.
