$1.3b · Series D+ · Media · San Francisco, US
fal has raised a $1.3b Series F led by Bessemer Venture Partners, with NVIDIA, Sound Ventures and T. Rowe Price participating, according to TechCrunch. The financing values the San Francisco generative-media infrastructure company at $16b.
fal announced the Series F alongside a separate Series E led by Sequoia Capital earlier in 2026. The company and reporting describe $1.65b raised across the two rounds; that combined figure is not the size of the Series F covered here.
Media models create a specialised inference market
fal gives developers one API for image, video, audio and 3D models, alongside serverless infrastructure for deploying custom models and larger compute clusters for training. The company says entertainment studios, advertisers, e-commerce companies and game developers are embedding these models into production workflows.
Serving media models is a distinct infrastructure problem. As Bessemer explains, image and video generation repeatedly refines whole frames, creating different memory, latency and throughput requirements from text models that produce one token at a time. fal's commercial proposition is that specialised kernels, scheduling and autoscaling can make that workload fast enough and predictable enough for customer-facing products.
The business sits between model labs and application developers. Developers gain access to many models through one integration, while labs can use fal for distribution and serverless capacity. That can create a two-sided advantage as more workloads improve infrastructure utilisation and more models make the API useful. It also creates a capital burden: maintaining low latency while video demand grows requires substantial GPU capacity before all of that usage becomes durable margin.
The round funds both capacity and model development
fal says the Series E and F will support infrastructure and the team through 2027, including new research. Bessemer says the company has expanded from inference into workflow tools, enterprise infrastructure and post-trained model development. Moving up the stack can deepen customer relationships because deployment, optimisation and model improvement sit in one environment.
It also changes the competitive test. fal must show that its performance and developer distribution remain differentiated as cloud providers, model labs and other inference platforms invest in the same workloads. The $1.3b round gives it room to reserve capacity and build products ahead of demand; the operational proof will be whether that scale produces better unit economics and stronger customer retention rather than only a larger compute footprint.
