Live
Abstract illustration of a film reel unspooling into rising stacks of glowing frames, one frame cracked, on a dark purple background
Creator Economy

Higgsfield’s $1 Billion Run Rate Is Real Math. Here’s What It Leaves Out.

Most companies bury the arithmetic. Higgsfield put it in the fourth paragraph. In the blog post announcing that it had crossed a US$1 billion (CA$1.39 billion) annualized revenue run rate, the AI video company spelled out exactly how it got there: take the revenue from the last four weeks and multiply it by 13.

That is not a confession. It is a fairly standard way for a fast-growing private startup to describe itself, and Higgsfield deserves some credit for saying it plainly. But it is also a reminder that the number at the top of every headline about the company is a projection, not a sum. Nobody has added up a billion dollars of Higgsfield sales. What exists is roughly four weeks of very strong revenue, extrapolated across a year the company has not yet lived.

The milestone was first reported by Bloomberg on Sept. 24, and the company’s own post followed on Sept. 29, dated and signed by the “Higgsfield.AI Team.” Every figure in it is company-claimed and unaudited. Some of them are remarkable anyway.

What Higgsfield says it built in 18 months

According to the company’s post, Higgsfield reached the billion-dollar pace 18 months after launching its AI-native creative suite. It reports more than 32 million users across 238 countries and nearly 1 million paying subscribers. It says revenue grew 20 times over the past year, that contracted business revenue is up 10 times since June, and that net revenue retention in its business segment sits at 300%. And: “Today, 390 of the Fortune 500 run campaigns on Higgsfield.”

The most telling number is the least flashy one. Subscriptions, the company says, now make up just over 60% of revenue, down from more than 90%. That is a business changing shape in real time, from a consumer app that sells monthly plans to individual creators toward something that sells to marketing departments.

Co-founder and CEO Alex Mashrabov, who previously co-founded AI Factory (sold to Snap in 2020) and led generative AI work at Snap, framed it as a shift in who uses the product and why.

“For the first time, the scale of content creation can match the scale of distribution.”

Alex Mashrabov, Higgsfield co-founder and CEO, in the company’s milestone post

In the same post he went further, predicting that “most direct-to-consumer businesses in the world are going to be winning and increasing their sales with AI-generated ads made on Higgsfield.” AdExchanger quoted him describing a “snowball effect”: once businesses see a positive return from AI-generated ads, “they keep generating more and more.”

Multiply by 13

Illustration: one solid, glowing wave period boxed at the left of a baseline, followed by twelve identical dashed copies repeating to the right, with a dotted arc spanning from the real segment to the end of the projection.
Illustration: prompt/power

Here is the math, because it matters. A year has 52 weeks, or thirteen four-week blocks. If four weeks of revenue times 13 equals US$1 billion, then Higgsfield brought in roughly US$77 million (about CA$107 million) in that four-week window. That is our back-of-envelope figure, not one the company has published, and “crossed $1 billion” means the real four-week number could be somewhat higher.

AdExchanger’s Sept. 25 roundup ran the item under the headline “Higgsfield’s $1 Billion Run Rate Comes With An Asterisk,” and noted that this kind of short-window calculation lacks the reliability of the year-over-year figures public companies report, which smooth out seasonal swings. It also pointed out that Anthropic and OpenAI describe their own run rates using similar short lookbacks. Higgsfield is playing by the industry’s current rules. The rules are just generous.

Push the arithmetic one step further and it gets more interesting. If just over 60% of a US$1 billion pace comes from subscriptions, that is something above US$600 million a year spread across nearly 1 million paying subscribers, or very roughly US$50 (CA$70) a month each. Again, that is our estimate from the company’s own rounded figures. It suggests Higgsfield’s paying base is not mostly hobbyists on entry plans, and it puts the other 40%, roughly US$400 million annualized, in business contracts and other non-subscription revenue, a category that by the company’s own account used to be less than a tenth of the business.

Three things a four-week snapshot cannot tell you:

  • Whether the window was typical. A promotion, a viral model launch or a big campaign contract landing in those four weeks inflates the annualized figure for as long as the company keeps quoting it.
  • Whether the earlier milestones are comparable. RuntimeWire tracked the trajectory: about US$500 million in June, US$700 million at the August fundraise, US$1 billion-plus now. It also noted the earlier figures came without disclosed measurement windows, so nobody outside the company can confirm all three used the same method.
  • What any of it costs. Generating video is compute-hungry. Revenue run rate says nothing about margin, and Higgsfield has not published one.

The climb itself is steep by any method. Forbes reported in February that Higgsfield was at a US$200 million run rate in January and US$300 million by early February, with 300,000 paying users and 15 million creators on the platform. Mashrabov’s stated goal then was US$1 billion by year’s end. He got there with three months to spare, at least by his own yardstick, and roughly tripled the paying base along the way.

Meanwhile, the private market priced the story accordingly. TechCrunch reported on Aug. 17 that Higgsfield raised a US$400 million (CA$556 million) Series B led by DST Global at a US$5.4 billion (CA$7.5 billion) valuation, roughly quadruple the US$1.3 billion it was valued at in January.

The same Forbes story had a second half

The February Forbes piece that documented the early growth was headlined around something else entirely: what it called the dark side of that growth. Rashi Shrivastava reported that Higgsfield had refunded US$1.35 million (about CA$1.88 million) to users hit by slowdowns, after a Black Friday promotion offered “unlimited” access at a 65% discount, US$25 a month, and users said generation became unusable without buying extra credits.

The same report described promotional videos, made by Higgsfield’s internal marketing team and outside creators, that put racist dialogue in the mouths of characters like Shrek and Moana, and nonconsensual deepfakes of Sydney Sweeney, Zendaya and President Trump. Creators in its Higgsfield Earn program reported trouble withdrawing payments and unexplained account bans; the company told Forbes that 90% of submissions were paid.

Mashrabov’s response, as quoted by Forbes, was an apology of the corporate-conditional kind: “Rapid scaling brings real challenges. We acknowledge that our internal processes and external communications haven’t always kept pace with our core values, and we have made mistakes.”

A week earlier, the company had found a different way to offend. The Register reported on Feb. 6 that Higgsfield bragged on X that its AI motion design tool had put an end to more than 20 creative jobs. Artists were furious. “Seems like Higgsfield found out today that there are lines they just shouldn’t cross in their marketing,” Motion Management CEO Aharon Rabinowitz said, per The Register, which added that Higgsfield did not respond to its requests for comment.

Remember the 20 jobs. They come back.

“Unlimited,” again, in August

If February was the reckoning, August suggested the lessons were partial. On Aug. 7, Higgsfield launched an unlimited offer for Seedance 2.5, the video model it had just added. WION reported that the company advertised average generation times under five minutes, while customers described waits of two hours or more. The outlet found that unlimited jobs run in a shared “Standard” queue limited to one job at a time, that Higgsfield’s terms allow it to throttle or move usage to a slower queue, and that refunds are limited to seven days after purchase with no credits used. The Series B was announced 10 days later, while the complaints were still coming in.

The same month, the creator problem moved from paid promos to audience trust. The Verge reported on Aug. 21 that two of YouTube’s best-known filmmaking creators, Matti Haapoja and Sam “Kold” Kolder, faced backlash over videos showcasing Seedance 2.5 on Higgsfield. The videos were not labeled as ads, and neither the creators nor Higgsfield answered The Verge’s questions about whether they were paid partnerships. Dataconomy noted that other creators then began posting screenshots of partnership offers from PR firms affiliated with Higgsfield, and The News reported that Kolder’s video description carried an affiliate link offering 30% off annual plans.

To be fair to everyone involved: neither creator has been shown to have broken a disclosure rule on the public record we could find, and an affiliate link is itself a kind of disclosure. The damage was reputational. For an audience of working filmmakers, two of their own appearing to sell the tool that a company had marketed as a job-ender landed badly.

The turn toward the people who buy ads

Now look again at that revenue mix. Subscription share falling from 90% to 60% means a growing slice of Higgsfield’s money comes from somewhere other than individual creators paying monthly. The company’s post points to campaigns for Fortune 500 brands and direct-to-consumer advertisers. That is a different customer, with legal departments and brand-safety reviews.

Higgsfield is courting that customer through OpenAI. On Sept. 21, OpenAI published a customer story on how Higgsfield uses GPT-6 Astra to generate ad variations, such as localizing one ad for different countries. “Higgsfield also enables smaller businesses to sell more products by generating ads using video AI,” Mashrabov says in it. Then, on Sept. 29, OpenAI launched Dots at DevDay: always-on agents that pursue goals in the background, available to ChatGPT Pro and Business Premium users. Within a day, Higgsfield announced Dots x Higgsfield on X, pitching an always-on creative crew that keeps producing while you are away and that you can check in on or pause by text, call or email.

It is not alone in that pitch. Adweek reported on Sept. 30 that Runway’s ad agent can recreate existing ads, spin out variations, localize them and publish approved versions to Meta, Google and TikTok, then pull performance data back in. Canva got there earlier: Canva Grow 2.0, unveiled at Cannes Lions in June, generates static and video ads and pushes them to Meta, TikTok and LinkedIn. The race is no longer about who makes the prettiest clip. It is about who sits closest to the ad budget.

Does the baggage make the trip?

Our read: the move upmarket helps Higgsfield with some of its problems and sharpens others.

It helps with billing. Enterprise customers sign contracts with service-level terms, not US$25 “unlimited” plans with a shared queue. If business revenue keeps growing at the pace the company claims, the consumer throttling fights matter less to the top line, and contracted revenue is steadier than a four-week snapshot of subscription sign-ups.

It sharpens the content question. A Fortune 500 marketing team is precisely the buyer that cares whether the vendor’s own promotional channels ran celebrity deepfakes, and that will ask how a tool prevents its output from looking like someone it has no right to depict. Brand-safety reviewers read Forbes too. And an always-on agent producing ads while nobody is watching is, by design, a product with fewer human checkpoints.

It also leaves the creator relationship unresolved. Higgsfield’s growth was built on creators: the tutorials, the showcase reels, the 32 million users. Its pivot is toward clients who, in many cases, used to hire those same creators. The company has already told the world, once, what it thinks that trade looks like.

On revenue, Higgsfield showed its work: four weeks, times 13. The other number it once advertised, more than 20 creative jobs ended, came with no method at all.

Sources

// Columnist, Creator Economy & Culture
Casie Stewart

Casie Stewart covers the creator economy, social platforms and wellness tech for prompt/power: who gets paid, who gets seen, and what the algorithm is really asking of the people who feed it. She has been publishing online since 2005, long before "creator" was a job title, and has spent most of that time on the other side of the platforms she now writes about. Based in Toronto. Her rule for any new app: if it promises to save you time, check how much of it you just spent reading the onboarding.

Latest from prompt/power

  1. The Family Safe Word: How to Beat AI Voice-Clone Emergency ScamsOct 7
  2. Reflection AI’s Beam, Explained: The 501B Open-Weight Model Aimed at ChinaOct 7
  3. Your SSN or SIN Leaked in a Breach? Do These 8 ThingsOct 7
  4. Apple’s Oct. 13 Event Rumour, Plus iPhone Duo Pre-Order Dates for CanadaOct 7
  5. Why Grindr Is Paying US$250M for Calgary PrEP Clinic FreddieOct 7

Leave a Reply

Your email address will not be published. Required fields are marked *