One million users in two months, and almost no business.

That was Higgsfield in 2024. Eighteen months after relaunching, it crossed $1 billion in annualized revenue, starting from $1 million. By its founder's count, Cursor needed 24. (Strictly, that $1 billion is an annualized run rate, not classic recurring revenue. Chapter 5 explains why that matters.)
Higgsfield is an AI video platform that lets brands, agencies and creators make ads and short-form video without a film crew.
It matters because it did not win by building the best video model. It won by building the best place to use everyone else's models.
And it got there after spending most of its seed money on the wrong customer, pivoting, and surviving a public backlash over its own marketing.
Below is the full story, told in five chapters, followed by a playbook you can run in your own company over the next 90 days.
The Enduring Founder has no relationship with Higgsfield. This is an independent playbook built from public sources, not sponsored content.
1. The Origin: A $166M Exit and a Seat Inside Snap (2020–2023)
Higgsfield is not a first-time founder's story.
Alex Mashrabov co-founded AI Factory, a company building mobile face-animation technology. Snap bought it in 2020 for $166 million, and Mashrabov went on to lead generative AI at Snap.
That seat gave him a close view of how people create short video on their phones, and of how hard it still was to make something that looked professional.
In October 2023 he founded Higgsfield with Yerzat Dulat, a Kazakhstan-based co-founder. Mahi de Silva joined in early 2025 as co-founder and chief strategy officer.
The company is headquartered in San Francisco, but most of the team sits in Almaty. By September 2026, about 300 of its roughly 400 employees were in Kazakhstan, a choice Mashrabov ties to a 15% personal income tax and a deep pipeline of maths and physics talent.
→ A previous exit doesn't guarantee the next one. What it buys you is access: to investors, to talent and to a clear view of where the market is broken.
2. The Wrong Bet: A Mobile App That Couldn't Keep Its Users (2023–2025)
Most people assume Higgsfield found product-market fit on day one. It didn't.
In April 2024 it announced an $8 million seed round led by Menlo Ventures and launched its first app, Diffuse, on iOS and Android. The pitch was explicitly mobile-first: Mashrabov argued that prioritizing phones over desktop workflows would let creators make social content anytime, anywhere.
Curiosity was not the problem. The consumer app reached 1 million users in two months. Retention was the problem: people played with it, then left, and they did not want to pay.
The search for fit took more than a year. Mashrabov later put the company's total seed funding at $16 million, more than the $8 million announced in 2024, and said over $10 million of it was gone before the web platform launched. He later called that money "tuition." His conclusion was blunt: at that stage, AI could not fix consumer retention.
The signal was hiding in the usage. The users who stayed were the ones making commercial work, and they needed fine control over shots, effects and iteration, which a phone screen could not give them.
So Higgsfield stopped building for everyone who wanted to make video and started building for the people who make video for a living: professional creators, agencies and marketing teams. The product moved to the browser.
→ When the deck says one thing and the usage says another, believe the usage. The users who stay are telling you who your customer is.
3. The Launch: No Model, Six Releases a Week, $10M in Six Weeks (2025)
The web platform launched on March 31, 2025. The breakthrough feature was camera control: cinematic moves that professional creative directors could direct with a click.
Higgsfield launched on its own model, which Mashrabov described as "open source plus plus plus." Within weeks, it changed course.
Instead of competing with the labs, it became an aggregator. Today it runs Google Veo, Kling, Seedance and its own models side by side, and it picks the model itself in more than 40% of cases.
→ While everyone else raced to train foundation models, Higgsfield bet that the money was in the workflow on top of them.
The second advantage was speed. In 2025 the team shipped about six releases a week. One of them, one-click visual effects, reached $10 million in ARR in about six weeks.
Not owning the model did not mean ignoring its cost. Higgsfield spends more than $4 million a month on AI models. It says it earns margins above 80% on its own and open-weight models, versus 20–30% on closed-source ones.

That gap explains why open-source models went from under 30% of usage to over 60% in a year. Routing a job to a cheaper model that does it equally well is not a technical detail. It is the margin.
→ You can build a billion-dollar company on rented models, but only if you control which model runs each job and what it costs you.
4. The Explosion: When Brands Became the Customer (2025–2026)
The revenue curve did not bend when the product got better. It bent when the buyer changed.
Analysts at Sacra estimated Higgsfield's run rate at about $11 million in May 2025 and $100 million by November. In January 2026 the company said it had hit $200 million. By early February it passed $300 million, by June $500 million and by August $700 million.
On September 29, 2026, Higgsfield announced a run rate above $1 billion, 18 months after launching its creative suite.

It did not have the market to itself. In January 2026, Sacra put Higgsfield at about $230 million in ARR and Kuaishou's Kling, one of the models it resells, at about $240 million. The aggregator was growing as fast as the lab it depended on.
The fuel was direct-to-consumer brands making video ads. Ad production is iterative: when a brand sees one AI-generated ad pay back, it orders more variants for new products, audiences and markets.
The numbers behind that loop are striking:
The average customer spends about $1,000 a year, roughly five times what a Canva user pays, according to Mashrabov.
Net revenue retention was above 300% at month 12.
One customer went from a $99-a-month subscription to a $6 million-a-year deal in six months.
Business customers now make up slightly more than half of revenue. Mobile consumer usage is under 10%.
Higgsfield also says it powers campaigns for 390 of the Fortune 500. Mashrabov says the company doesn't buy ads and owns its distribution. As the next chapter shows, that distribution still cost money: it paid creators to post about it.
Investors followed the curve. In January 2026, an $80 million Series A extension led by Accel, with GFT Ventures and Menlo Ventures, valued the company at $1.3 billion. In August, a $400 million round led by DST Global, with Goldman Sachs Alternatives, Intel Capital and Tribe Capital, among others, valued it at $5.4 billion post-money. On October 23, TCV and NVIDIA joined through an extension of that round, on undisclosed terms. It was NVIDIA's first investment in the company.

The most telling detail is the multiple. The run rate grew 3.5x between the January and August rounds, but the valuation grew 4.2x. Investors were not just paying for more revenue. They were paying for faster, stickier, more B2B revenue.
→ Valuation follows the quality of revenue, not just its size. Expansion from business customers is what moves the multiple.
5. The Conflict: Shock Marketing, Creator Backlash and the ARR Debate (2026)
Hypergrowth came with a bill, and part of it was paid in reputation.
In February 2026, Forbes reported on the darker side of Higgsfield's growth. The company had offered creators $200 to post about a new tool, and a media kit sent to them included stock templates with Higgsfield logos added. Creators also received folders of racist and obscene videos featuring well-known cartoon characters, plus non-consensual deepfakes of celebrities.
De Silva called the shared material a mistake. He also said, "We fully admit that we push the envelope," explaining that controversial content gets attention on X.
Mashrabov acknowledged that the company's "internal processes and external communications haven't always kept pace with our core values."
The second debate was about the number itself. Higgsfield's $1 billion is the latest four weeks of revenue multiplied by 13. The company is explicit that it is not revenue earned over the past year.
That matters more for Higgsfield than for a classic SaaS company. Subscriptions fell from over 90% of revenue to just over 60%, and the rest is on-demand usage that can move week to week. In an earlier interview with SaaStr, Mashrabov described a slightly different formula built on monthly figures multiplied by 12.
To be fair to the company, it publishes its method, which is more than many AI startups do. It also says it has been cash-flow break-even for the past couple of months.
Users also complained that a 2025 Black Friday "unlimited" plan slowed to a crawl after a few videos. Heavy discounting is still part of the playbook: in mid-November 2026, its pricing page again advertised up to 64% off annual plans for Black Friday. The same week, Anthropic named Higgsfield a launch partner for linked billing on the Claude Platform, a sign of how far it has moved toward business customers.
→ Shock gets you attention once. Trust is what lets you keep charging. And when your revenue is usage-based, the way you define ARR is part of your credibility.
What's next: the five chapters tell you what Higgsfield did. The next section turns it into something you can run, organized by the stage you're in. Inside: a pivot test with real retention benchmarks, so you don't pay $10 million in "tuition" like Higgsfield did; the gross margin and retention numbers investors will compare you to; a Higgsfield ad workflow you can run this week for a few dollars per video; a 90-day plan with a pass/fail check at each step; a one-page scorecard; and four copy-paste prompts.
Subscribe to CLUB to read the rest.
Become a paying subscriber of CLUB to get access to this post and other subscriber-only content.
UpgradeAlready a subscriber? Sign in
A subscription gets you:
- Everything included in Premium.
- Early access to new tools and the specialized AI and product-market fit newsletter.
- One-on-one feedback on your business, with actionable guidance on strategy, positioning, and growth
- Masterclasses with Pablo Fernandez Alvarez
