San Francisco-based Higgsfield has done what few artificial intelligence companies manage. Its annualized revenue run rate now stands at $500 million. That marks a tenfold jump from $50 million last September. The company doubled its revenue again this year. And it turned cash-flow positive by the end of May.
But the numbers only tell part of the story. Brands have rushed to adopt its tools for generating commercials and social campaigns without traditional production crews. Advertising now drives 70 percent of activity on the platform. The startup counts 390 Fortune 500 customers. Its growth has drawn comparisons to fast-scaling AI coding tools like Cursor and Lovable.
Chief strategy officer Mahi De Silva laid out the pace. “Higgsfield is on pace to reach a $1 billion dollars in run rate as soon as the end of this year.” He spoke to Business Insider in a report published yesterday. The surge reflects a broader shift. Marketers no longer treat AI video as experimental. They see it as core infrastructure.
Founded in 2023 by former Snap executive Alex Mashrabov, Higgsfield offers a platform that lets nontechnical users create and edit AI-generated clips. Its products include a web studio, the Diffuse mobile app, and dedicated ad tools. Revenue comes mostly from consumption-based credits, topped up by monthly subscriptions. Earlier estimates from research firm Sacra pegged annualized revenue near $400 million in May. The latest self-reported figure shows acceleration. (Sacra)
Only months ago the picture looked different. In January the company closed an $80 million Series A extension. That brought total Series A funding above $130 million and lifted its valuation to $1.3 billion. Investors included Accel, Menlo Ventures, GFT Ventures and AI Capital Partners. Total capital raised now approaches $150 million. (Reuters)
The latest product move signals deeper ambition.
Yesterday at Cannes Lions, Higgsfield unveiled Supercomputer 2.0. Built on NVIDIA’s Agent Toolkit, the system functions as an autonomous marketing engine. It can create, launch and optimize ads with minimal human input. The launch coincided with fresh revenue data showing the run rate nearly quadrupled in the first five months of 2026. NVIDIA’s Jamie Allan praised the approach. “By building Supercomputer 2.0 using the NVIDIA Agent Toolkit, Higgsfield brings the efficiency of small, fast models to the work that runs continuously inside every campaign, giving brands a creative engine designed to run autonomously at enterprise scale.”
De Silva highlighted the maturity of the output. “The first generation of AI videos had this plasticity to them. We’re past that now.” He pointed to a live TV commercial generated without any humans or animals on set. It looked indistinguishable from a traditional shoot. The company stresses a permission-first model. Agents take no action until granted explicit credentials. That addresses enterprise concerns around security and control. (Inc.)
Yet success has brought scrutiny. Critics label some output as AI slop. Others point to occasional obscene or biased videos. A company survey found nearly 30 percent of creators fail to disclose AI use to clients. De Silva pushes back. He compares the situation to computer-generated imagery in film. “When Scarlett Johansson appears in an Avengers movie, and she flies through the sky, there isn’t a disclaimer that says, ‘This movie was built with CGI. It wasn’t real.’ For the last 20 years, we’ve lived with media that is digitally created and manipulated.”
Relationships with larger model providers remain symbiotic, he argues. Higgsfield integrates multiple foundation models rather than competing head-on. OpenAI’s decision to pull back from its Sora video project reinforced his view. “It’s admirable that they realized their mistake and decided to focus their efforts elsewhere.”
The company now eyes a Series B. Inbound interest from venture firms proved too strong to ignore. “We were cashflow positive at the end of May, so we are quite the exception to most AI companies,” De Silva said. “We’re not hemorrhaging money, but we are raising a Series B because of inbound interest.” He invoked an old line about financing. “I’m a fan of the adage that when dinner is served in financing terms, you should sit down and eat.”
And the market keeps moving. Just this week reports surfaced of new plugins for Adobe Premiere, After Effects and DaVinci Resolve. A $99 monthly unlimited plan for its Seedance 2.0 model sparked debate on usage limits and generation speeds. The platform’s blog touts improved character consistency through SOUL ID technology. These incremental gains matter. They help the tools move from viral demos to daily business use.
Higgsfield’s trajectory stands out for another reason. Many AI startups burn cash chasing scale. This one generates profit while expanding. Its bet on agentic systems goes beyond one-off video clips. The goal is continuous, autonomous campaign management. Brands appear willing to pay. Early beta customers in marketing automation already spend more than $200,000 a year.
Questions remain. Can quality hold as volume explodes? Will disclosure rules tighten and dent adoption? How long before foundation model owners integrate similar capabilities directly? De Silva sees those threats as overstated. The real barrier, he believes, lies in ease of use and enterprise trust. Higgsfield’s recent moves target exactly those points.
So the startup that once focused on consumer and creator tools now sits at the center of brand marketing budgets. Its revenue run rate matches some of the hottest names in AI software. Cash flow positivity sets it apart. And fresh agent products launched at a premier advertising festival suggest the momentum has not peaked. For an industry still separating hype from sustainable business, Higgsfield offers a rare current example of both.
WebProNews is an iEntry Publication