The company that runs the models behind a growing share of the internet’s AI-generated images and video just doubled its price in six months. Fal, the San Francisco inference startup, has raised $1.3 billion in a Series F led by Bessemer Venture Partners at a $16 billion valuation, twice the $8 billion mark set in April, Bloomberg reported.
Key takeaways
- Fal raised $1.3 billion in a Series F led by Bessemer Venture Partners, with Nvidia, Sound Ventures, HOF Capital, and T. Rowe Price participating, at a $16 billion valuation, per Bloomberg reporting carried by TechCrunch.
- The company announced the round alongside a previously closed $350 million Series E led by Sequoia Capital in April; the two rounds total $1.65 billion.
- Fal surpassed $800 million in annualized revenue earlier this year, per a person familiar with the matter cited by Bloomberg. CEO Burkay Gur says demand outgrew capacity this summer to the point of rate-limiting customers.
- The 185-person company will spend the capital on compute and hiring. Bessemer partner Talia Goldberg is joining the board.
What Fal sells, and why it is suddenly worth double
Inference is the unglamorous half of AI: running a model that has already been trained to produce outputs. Every generated image, video clip, or voice track is an inference call, and Fal has positioned itself as the toll booth for generative media specifically, hosting and running video, image, and audio models through a single API. Demand has been strongest from customers building on open-source models, per TechCrunch.
The growth numbers explain the valuation step. Fal passed $800 million in annualized revenue earlier this year, according to Bloomberg’s source, and the constraint has shifted from finding customers to serving them. “We hit a point in summer where we were not able to serve our customers and had to rate limit them,” Gur told Bloomberg. Raising $1.3 billion is, in large part, a compute purchase.
Bessemer, which has backed Fal since its Series B, describes the customer base as including Adobe, Canva, Shopify, World Labs, and Amazon – the investor’s own characterization. The firm is betting that generative media is becoming one of AI’s largest markets and that media workloads need purpose-built infrastructure rather than stacks designed for text models.
From serving models to improving them
Fal was founded in 2021 by Burkay Gur, previously a machine learning leader at Coinbase and an engineer at Oracle, and Gorkem Yurtseven, formerly a developer at Amazon. The company has spent the past year moving up the stack: per Bessemer, its fal Research arm released H3 Max, a post-trained version of MiniMax’s H3, in August, and recently partnered with Black Forest Labs on Flux 3 Max. Gur told Bloomberg the company has no short-term plans to pre-train its own models, though he did not rule it out; some of its model collaborations involve licensing fees or revenue-share arrangements.
The competitive set is well funded and growing: Baseten, Fireworks AI, and Modal are all raising large rounds at higher valuations. The shared caveat, per several people familiar with inference companies’ financials cited by TechCrunch, is that gross margins stay thin because compute – bought or leased – eats the revenue. Fal’s answer to that problem is now a $1.3 billion balance-sheet entry of its own.
FAQ
What is Fal?
Fal is a San Francisco startup founded in 2021 that hosts and runs AI image, video, and audio models for developers and enterprises through a single API – the inference layer for generative media. It was co-founded by CEO Burkay Gur and Gorkem Yurtseven.
How much did Fal raise and at what valuation?
$1.3 billion in a Series F led by Bessemer Venture Partners at a $16 billion valuation, double its April 2026 valuation of $8 billion, per Bloomberg. Announced at the same time: a previously closed $350 million Series E led by Sequoia, bringing the combined total to $1.65 billion.
What is AI inference?
Inference is the process of running an AI model that has already been trained, so it can generate outputs – an image, a video, a line of text. Training happens once; inference happens every time anyone uses the model, which is why inference capacity scales with usage.
Is Fal the same company as FalconX?
No. Fal is a generative-media inference startup. FalconX is a separate company, a crypto prime broker that confidentially filed for an IPO earlier this year, per Grey Journal’s earlier coverage.
More from Grey Journal: Trajectory Raises $45 Million From Sequoia to Let Domain Experts Retrain AI Models.
