Fal doubles valuation to $16B with $1.3B round led by Bessemer

Fal has raised $1.3 billion in a funding round led by Bessemer Venture Partners, with Nvidia, Sound Ventures and T. Rowe Price participating, TechCrunch reported on November 3. The deal values the San Francisco startup at $16 billion, double the $8 billion valuation it carried in April, according to Bloomberg.

Fal does not train models. It hosts and runs AI video, image and audio models that others have already trained, letting developers send requests to its infrastructure instead of building and operating their own GPU clusters.

The company announced its Series E and Series F at the same time, saying it raised a combined $1.65 billion across both, with Sequoia leading the earlier round. The newest round lands as demand for inference — running a trained model to generate output — surges, particularly among customers relying on open-source models.

Fal surpassed $800 million in annualized revenue earlier this year, Bloomberg reported, a figure that puts it near the front of the unlisted inference providers.

The whole category is heating up. Baseten, Fireworks AI and Modal are growing quickly and raising large rounds at higher valuations, a sign that investors now treat the inference layer as one of the more predictable parts of the AI stack.

Margins remain the weak point. Several people familiar with the financials of inference companies told TechCrunch that gross margins stay thin because of the high cost of buying or leasing compute: the more revenue these companies book, the larger the hardware bill.

Fal was founded in 2021 by Burkay Gur, a former Coinbase machine learning leader and Oracle engineer, and Gorkem Yurtseven, previously a developer at Amazon. It has become a default inference endpoint for generative media workloads.

The interesting question is not the headline number but unit economics. Once model quality converges across providers, the advantage belongs to whoever can run models more cheaply and more reliably, and Nvidia's presence on the cap table makes the compute relationship one to keep watching.