In the space of seven months, fal's price went from $4.5 billion to $16 billion. On 3 November 2026 the company announced a $1.3 billion Series F led by Bessemer, with Nvidia, Sound Ventures, HOF Capital and T. Rowe Price joining. It was the second giant round of the year, after a $350 million Series E in April led by Sequoia.
The company's legal name hints at an older life. It is Features & Labels Inc., founded in 2021. The sources read for this profile do not say what it first built, so that story is left to later reporting. What they do say is who ran it: Burkay Gur as chief executive, who had led machine learning at Coinbase and worked as an engineer at Oracle, and Gorkem Yurtseven as chief technology officer, a former Amazon developer. Their birthplaces and schools were not found.
The sources read for this profile do not say what it first built, so that story is left to later reporting.
What fal sells is plumbing. As image, video and audio models multiplied, builders needed a place to run them without managing their own servers. fal gives them a single API for generative media models. Its in-house researchers post-train some models to make them faster, and in some partner deals fal pays licensing fees or shares revenue. The customers are entertainment studios, advertising agencies and game developers using the models in production.
The numbers moved fast. Revenue was reported at about $200 million annualized in October 2025 and about $400 million by early 2026. A person familiar with the matter told Bloomberg that annualized revenue passed $800 million in 2026. At $16 billion that is roughly 20 times revenue, a multiple one analyst site noted. Investors have put in more than $1.9 billion in total, and the 2026 rounds alone add up to $1.65 billion.
Behind the figures is a business with a built-in squeeze. Compute is expensive, and gross margins across inference providers remain thin. The rival Fireworks AI closed $1.5 billion at a $17.5 billion valuation in July, so the race is not close to settled. The new money is earmarked for compute and hiring; the company had only 185 employees.
There is not much of the human story in public yet. No childhood tale, no famous rejection. The record shows two engineers from big-company infrastructure jobs who built a picks-and-shovels business at the moment media generation exploded.
The lesson for founders is old but fits: when everyone is chasing the shiny model, the tool that makes all of them usable can become the larger business. The risk is that the margin belongs to whoever owns the chips.