Three months ago, a Sequoia Capital-led funding round valued Fal at $4.5 billion. The company is now in talks to raise between $300 million and $350 million at a valuation of around $8 billion, according to a person with direct knowledge of the fundraise, as first reported by The Information.
Fal builds cloud infrastructure that lets developers run AI models generating images, video, and audio. Its annualised revenue recently reached $400 million, up from $200 million in October, according to The Information.
The company charges developers based on usage, including per second of video output, and serves three million developers, including at Adobe, Canva, and Shopify. Investors include Andreessen Horowitz, Notable Capital, Meritech Capital, and Kindred Ventures. Which firm is leading the new round is not yet known.
The round is not being raised as a single transaction at one price.
Louis Eriakha
One Round, Two Different Prices
Fal is splitting the fundraising into two instalments, each priced at a different valuation. The higher-priced instalment carries the $8 billion figure. The other is priced below that.
Investors who get into the lower-priced instalment receive a paper markup on their investment automatically when the higher-priced instalment closes. No revenue milestone, product launch, or new customer is required for that markup to occur. The deal structure produces it upon closing.
The Information reports that this arrangement has become more common as investors compete for access to fast-growing AI startups. Firms in the lower-priced tranche get in at a cheaper entry point, while firms in the higher-priced tranche secure allocation at the prevailing market price. The Information also reports that the structure has added to concerns in the venture community about a bubble in AI startup valuations, where the mechanics of a deal can move a company's price tag separately from changes in the underlying business.
AI startups have been raising successive rounds quickly since OpenAI launched ChatGPT three years ago, and investors competing for the fastest-growing companies have increasingly accepted terms including the dual-tranche structure as part of that environment, according to The Information.
What Fal Builds and Who Uses It
Founded in 2021 by engineers who previously worked at Amazon and Coinbase, Fal operates in AI inference, the process of running a trained model to produce an output on demand. Developers use its infrastructure to generate images, video, and audio inside their own products. It competes with inference providers, including Replicate, as well as traditional cloud providers.
Developers on the platform can run Google's Nano Banana, Black Forest Labs' Flux, and hundreds of other models. Fal raised $314 million across three separate funding rounds last year alone, according to The Information.
Ogbonda Chivumnovu
