Fal doubles valuation to $16B as AI inference demand surges

Fal runs the infrastructure that lets other companies call on already-trained AI models to generate video, image, and audio outputs — the inference layer of the AI stack. Bessemer Venture Partners led the new $1.3 billion raise, with Nvidia, Sound Ventures, and T. Rowe Price also participating. The startup disclosed its Series E and Series F rounds together, saying the two combined totaled $1.65 billion, with Sequoia having led the earlier round.

The Series F lands amid soaring demand for inference services — running an already-trained model to produce outputs — particularly from customers building on open-source models. Fal’s annualized revenue had already topped $800 million earlier this year, according to Bloomberg’s report. Rival inference providers Baseten, Fireworks AI, and Modal have likewise been growing fast and raising large rounds at rising valuations, though people familiar with their financials told Bloomberg that gross margins across the group remain thin because compute is expensive to buy or lease.

The company launched in 2021, started by Burkay Gur — who had led machine learning at Coinbase and worked as an engineer at Oracle — together with Gorkem Yurtseven, previously a developer at Amazon.

Why it matters

The jump to a $16 billion valuation in roughly seven months shows investors still betting heavily on AI inference infrastructure, even as thin margins across the sector raise questions about how durable that growth is.