Higgsfield says it crossed a $1 billion revenue run rate

Higgsfield, the AI video generation startup founded by former Snap generative-AI head Alex Mashrabov, says its annualized revenue run rate has crossed $1 billion, 18 months after its browser-based product went live.

Mashrabov announced the number in a post on X, and OfficeChai reported on 25 September that he framed it as a 20x increase in revenue since September 2025. The company circulated a chart placing its 18-month climb to $1 billion ahead of Anthropic and Cursor at around 22 months, OpenAI at roughly 37 months, Surge AI at about 47 months and Ramp at around 66 months.

The company's own supporting figures, as reported from its announcement, are a net revenue retention rate of 300 percent, enterprise adoption growing 115 percent month over month since June, and positive gross margin since the start of the year. A separate company statement put total users above 30 million worldwide and said the platform is used across the Fortune 500, with enterprise adoption up tenfold since June alone. Those numbers are Higgsfield's own and are not audited.

The curve behind the headline is steep even by AI-startup standards. The annualized run rate stood at roughly $200 million at the end of 2025, climbed to $500 million by June 2026 and reached $700 million in August alongside a $400 million Series B that valued Higgsfield at $5.4 billion, led by DST Global with participation from Goldman Sachs Alternatives and Intel Capital. Seven months before that round, the company was valued at $1.3 billion.

A run rate multiplies a recent short period of revenue across a full year. It is a projection rather than revenue already collected, and the same caveat applies to the figures Anthropic and OpenAI have disclosed to investors.

Mashrabov used the announcement to argue against the idea that foundation model makers eventually absorb most of the value created on top of them. His position is that better underlying models have widened what enterprises can do and made the applications built into their workflows more valuable, not less.

Alongside the revenue claim sits the company's clearest piece of production evidence. Higgsfield's feature film Hell Grind runs 95 minutes and was assembled by a 15-person team in Almaty, Kazakhstan over 14 days, according to Gate News's account of Higgsfield's own statement. The budget was about $500,000, roughly 80 percent of it GPU cloud computing. The team ran 16,181 generations and used 253 shots, a ratio of about 64 to one.

Higgsfield also published the production's prompts, character models and scene assets, a 19-minute behind-the-scenes breakdown and a prompt toolkit packaged as a Claude Skill that becomes usable about 30 seconds after the files are uploaded.

For creators the useful part is not the revenue figure but the ratio. A 64-to-one generation rate is what a finished 95-minute AI feature actually costs in compute.

Higgsfield has not published audited financial statements, so the $1 billion figure stands as a company statement.