Fal raises 1.3 billion at a 16 billion valuation, Bessemer leading

Fal, a San Francisco startup that hosts AI image, video and audio models, has raised 1.3 billion dollars. Bessemer Venture Partners led the Series F round. NVIDIA joined as a strategic investor. Sound Ventures and T. Rowe Price also took part. TechCrunch, citing Bloomberg, reports the deal values Fal at 16 billion dollars. That is double the 8 billion dollar valuation Fal held in April.

Fal announced the Series F today, alongside an earlier Series E led by Sequoia Capital. The two rounds together raised 1.65 billion dollars. The company confirmed the investors in a post on its own blog. The blog does not give the amounts or the valuation. Those figures come from Bloomberg.

What Fal does

Fal runs trained AI models on behalf of customers. A team sends an image prompt, a video prompt or an audio prompt. Fal returns the output from the model the team picked. The company names its customers in its blog post. They include entertainment studios, ad agencies, e-commerce brands, game developers and marketing teams. They run image, video and audio models inside their own pipelines. Some of those models are Fal's own. Many are open-source models the company hosts.

The company was co-founded in 2021 by Burkay Gur and Gorkem Yurtseven. Gur used to lead machine learning at Coinbase and worked at Oracle. Yurtseven was a developer at Amazon. TechCrunch's report lists both.

How big the business is

TechCrunch reports that Fal surpassed 800 million dollars in annualized revenue earlier this year. The figure is from Bloomberg's report. Fal has not published a revenue number on its own blog. The round is not just big. It arrives at a moment when every provider that runs AI inference is raising money on the same story. TechCrunch names Baseten, Fireworks AI and Modal as peers that have also raised large rounds at higher valuations.

TechCrunch adds a caveat. People familiar with inference companies' financials told it that gross margins remain thin. The cost of buying or leasing compute eats most of the price the customer pays. A big round in this space is a bet that scale brings cheaper compute and better margins over time.

Why this matters for a team that uses Fal

A founder who runs an image, video or speech model through Fal just got a signal. The company has the cash to keep paying for chips and to keep adding model hosts. The round reads as a vote that the inference market keeps growing.

The practical check is the same as before. Fal lists its prices by model and by second of output. A team that spends real money on inference should keep a second provider on standby. The names in this round, Fireworks, Baseten and Modal, are also candidates. The cheapest one depends on the model and the batch size. Send the same prompts through two providers. A test like that shows how much depends on Fal and what can be moved.

Nothing in the round says what Fal plans to charge next quarter. Nothing says which open-source models the company will add. The news is that Fal has the money to keep running. A large round at double the April valuation is where the inference market is today.

Source

TechCrunch by Marina Temkin, citing Bloomberg's report, and Fal's own blog post on the Series E and Series F.