$300M in 11 Months: The Rise and Reckoning of Higgsfield AI
How Higgsfield AI hit $300M ARR in 11 months with multi-model aggregation, influencer distribution, and extreme shipping velocity, then faced a reputation crisis.
Eleven months. That’s how long it took Higgsfield AI to go from zero revenue to a $300 million annual run rate. Twenty million users. Five million videos generated per day. A $1.3 billion valuation. Kazakhstan’s first unicorn. Celebrity users including Madonna, Snoop Dogg, and Will Smith, all showing up organically.
By every metric that matters in Silicon Valley, Higgsfield was the fastest-growing AI startup anyone had ever seen. Faster than OpenAI, Slack, and Zoom at the same stage, according to the company.
Then Forbes published “Racist Videos And Payment Problems: The Dark Side Of This AI Startup’s Super-Fast Growth.” The company’s X account got suspended for inauthentic behavior. Creators who’d promoted the platform turned into its loudest critics. The internet gave it a new name: “Shitsfield AI.”
How do you build the fastest-growing AI startup in history and torch your reputation in the same year? That’s the question this teardown answers.
TL;DR
How did Higgsfield hack its growth? A mobile-to-web pivot targeting professional creators, multi-model aggregation (12+ AI models including Sora 2), influencer-powered distribution through its Earn program, and a shipping velocity of 4-7 releases per week, all executed by a founder who’d already sold a company to Snap for $166 million.
What can you learn?
The aggregation play works. You don’t need your own model to build $300M ARR. You need to be the best interface to everyone else’s models.
Influencer distribution is a loaded weapon. It scales faster than any other channel, but the content you incentivize is the content you get blamed for.
Speed without guardrails is a liability. Shipping 200+ features in 11 months meant no ethical review process caught the problems until Forbes did.
Company Background
Alex Mashrabov’s mother designed rockets for the Soviet space program. He started programming at 10 and was ranked top-3 globally in competitive programming by 20. He co-founded AI Factory, which built neural networks that could run directly on mobile devices. Snap acquired it for $166 million in 2020, and the technology became the backbone of Snapchat’s face filters and Cameos. Mashrabov became Head of Generative AI at Snap, where he helped ship AI features to billions of users.
In late 2023, he left Snap and co-founded Higgsfield with Yerzat Dulat, a Kazakh AI researcher, and Mahi de Silva, a veteran with decades at Apple, NVIDIA, and Triller. The company was named after the Higgs field in physics: the field that gives particles mass. The idea: ideas that pass through the Higgsfield platform gain mass as videos.
In short: the guy who invented Snapchat’s viral face filters went and built a platform to make every video on social media AI-generated.
Key Metrics Snapshot
Revenue trajectory (the fastest in AI history):
March 2025: Web platform launches
April 2025: $10M ARR within weeks (”We’d never seen anything like it,” per GFT Ventures investor)
May 2025: $11M ARR
August 2025: $50M ARR, cashflow positive (5 months in)
November 2025: $100M ARR (7 months)
January 2026: $200M ARR (9 months, doubled in 2 months)
February 2026: $300M ARR (11 months)
Users & scale: 20M+ users, 5M videos/day, 50M+ total videos created, 3B+ social media reach
Funding:
Apr 2024: $8M Seed (Menlo Ventures, K5 Global, Alpha Intelligence Capital)
Sep 2025: $50M Series A (GFT Ventures lead). $1B valuation. Kazakhstan’s first unicorn.
Jan 2026: $80M Series A extension (Accel, GFT, Menlo). $1.3B valuation.
Total raised: $138M
Other: 300K paying subscribers, ~100 employees, credit-based pricing from Free to $125/month
ACT 1: THE RISE
Growth Strategy 1: The Pivot That Changed Everything
Higgsfield didn’t start as a web app. It started as a mobile app called Diffuse, an image-to-video generator where you could upload a selfie and turn it into an animated clip. Think TikTok-native, phone-first, casual creator tool.
It didn’t work.
Mashrabov explained the problem bluntly: “Generative AI doesn’t change the economics of mobile apps. High churn, small ticket sizes, limited expansion.” The mobile app had users, but they weren’t paying enough, weren’t sticking around, and weren’t upgrading. Classic consumer mobile trap.
The inflection came when they abandoned mobile-first and rebuilt for desktop. They targeted a completely different user: professional social media marketers, content teams, and ad creators. People who needed volume and quality, not just novelty.
The product shift was specific. They built over 100 cinematic templates with pre-defined camera controls: dolly-ins, aerial pullbacks, crash-zooms, 360 orbits, FPV drone shots. These templates turned the experience from “type a prompt and hope” to “pick a style and click.” Click-to-video, not prompt-to-video.
“That made videos cinematic. Once that went live, traction exploded,” Mashrabov told Evolving Edge.
Within weeks of the web launch in March 2025, they hit $10M ARR. The mobile app still exists as a companion product, but the revenue engine is the web platform.
Steal this: If your initial distribution channel isn’t producing the right economics, question the channel, not the product. Higgsfield’s core technology didn’t change. Their go-to-market completely did. The same AI, repackaged for professionals with money to spend, unlocked everything.
Growth Strategy 2: Multi-Model Aggregation
Here’s the thing most people don’t realize about Higgsfield: they don’t have their own AI model. They’re a platform that sits on top of everyone else’s models.
Their stack uses 12+ models, including OpenAI’s Sora 2, Google’s Veo 3.1, WAN 2.5 and 2.6, and Kling. For planning and scene composition, they use GPT-4.1 and GPT-5. OpenAI featured Higgsfield on their blog, calling out how they use GPT models to translate vague creative intent (”make it dramatic”) into structured video instructions that Sora 2 can execute.
Higgsfield is reportedly OpenAI’s largest Sora 2 customer.
The speed of integration is what made this work as a Higgsfield growth strategy. Mashrabov: “Whenever a new state-of-the-art model dropped, we integrated it within 24 hours. Customers learned they could trust Higgsfield to always have the latest, best-performing AI.”
This is the exact same playbook that worked for Perplexity in search (aggregating LLMs) and Cursor in coding (wrapping models in a better UX). You don’t need to train the best model. You need to be the best interface to the best models.
The risk is obvious: if you don’t own the model, you don’t own the moat. Mashrabov acknowledges this: “The economics of generative AI companies are challenging. Gross margins aren’t like traditional SaaS, they’re well below 80-90%.” That’s why growth and cheap acquisition matter so much. You need volume to offset thin margins.
Steal this: In a fast-moving AI market, the company that aggregates and ships fastest often beats the one building from scratch. But know the trade-off: you’re trading defensibility for speed. That works when you’re growing 60% monthly. It gets uncomfortable when growth slows.
Growth Strategy 3: Influencer-Powered Distribution
This is the one that made Higgsfield’s growth numbers possible. And it’s the one that nearly destroyed them. But let’s start with why it worked.
Higgsfield launched Higgsfield Earn, an automated creator monetization program. Creators link their social accounts, join active campaigns, and get paid based on their content’s real-time engagement. The numbers: 10,000+ creators commissioned, $1M+ distributed, 50,000+ submissions processed, 90% approval rate.
They also introduced Soul ID, a system for building persistent AI influencers: virtual personas with consistent faces, styles, and personalities that creators can monetize like real influencer accounts.
The strategy was deliberate. Instead of spending on performance marketing, Higgsfield turned creators into their paid distribution network. Mid-tier and micro-influencers with loyal, engaged audiences got polished templates and pre-made video styles to showcase. The content went viral, the creator got paid, and Higgsfield got users.
The social reach numbers tell the story: 3 billion+ social media impressions. Creator-driven organic reach. That’s more reach than all their competitors combined, according to Mashrabov.
Celebrity adoption amplified it further. Madonna, Snoop Dogg, and Will Smith all used Higgsfield to create content, all organically according to the company. When a celebrity shares something made with your tool and you didn’t pay them, that’s the highest form of product validation.
This is the strategy that drove 300,000 paying subscribers in under a year. But here’s the foreshadow: when you incentivize creators to make content that goes viral, you’re optimizing for attention. And attention doesn’t care about your brand guidelines.
Steal this (carefully): Creator-as-distribution-channel is one of the most powerful GTM motions available right now. It works when the incentive structure rewards quality. It breaks when it rewards shock value. More on that in Act 2.
Growth Strategy 4: Shipping Velocity as a Competitive Weapon
Most startups talk about shipping fast. Higgsfield built a culture around it.
The numbers: 4 to 7 feature releases per week. Sometimes one release per day. 200+ total releases since launch. Ten new video presets created daily, with underperforming ones cycled out based on engagement data. Every new state-of-the-art model integrated within 24 hours of release.
Mashrabov described it plainly: “We’ve built on the ‘996’ culture: intense, fast, and deeply committed.”
The product surface area expanded at a pace that’s genuinely hard to believe. In 11 months since the web launch, Higgsfield shipped:
Cinema Studio (Sep 2025): Flagship production workspace with optical camera physics, 50+ camera presets, 3D scene access
Cinema Studio 2.0 (Feb 2026): Multi-shot sequencing, 16x efficiency improvements
Cinema Studio 2.5 (Mar 2026): Genre logic, character emotion control
Click-to-Ad: Paste a product URL, get a social-first video ad
Higgsfield Audio: Text-to-speech, voice swap, video translation
Sora 2 Trends: Auto-generates trend-matched videos from trending formats
Soul ID / UGC Factory: Persistent AI characters for brand content at scale
Higgsfield Earn: Full creator monetization platform
They didn’t hire their first salesperson until October 2025, five months after launch, with $50M already in the bank. Product velocity was the sales team.
And here’s the number that still surprises people: Higgsfield was cashflow positive from the $50M ARR mark. In a space where most competitors are burning capital to subsidize compute, Higgsfield was profitable.
“We release product updates almost every day. This rhythm keeps us learning faster than anyone else in the space, and that’s unlikely to change.” — Alex Mashrabov
Steal this: Shipping velocity is a compound interest machine. Each release teaches you something. Each integration of a new model retains users who would otherwise leave for the shiny new tool. The danger? When you ship this fast, you skip the step where someone asks “should we?”
Four strategies. $300M ARR. 20 million users. Kazakhstan’s first unicorn. Everything was working. And then it wasn’t.
ACT 2: THE RECKONING
The Cracks: Misleading Marketing
The first cracks weren’t dramatic. They were mundane: users feeling deceived.
Influencers in the Higgsfield Earn program received overly polished templates that produced demo videos far superior to what average users could create. These slick clips went viral, drawing in millions of users who signed up expecting the same quality. When they couldn’t replicate it, the frustration landed on the influencers, not on Higgsfield.
Then came the stock footage problem. Multiple sources reported that Higgsfield passed off stock videos from Envato as AI-generated in their marketing materials. When your entire brand promise is “AI-generated video,” using stock footage in demos is the equivalent of a restaurant microwaving frozen meals and calling it chef-prepared.
The pricing drew heat too. Higgsfield offered “unlimited” plans at a 65% discount. Users subscribed, used the platform heavily, and then found their generation speeds throttled to the point of unusability without purchasing additional credits. Trustpilot reviews tell the story: “100% scammers. Their so-called ‘unlimited’ plans are pure marketing.”
None of this was unusual for a fast-growing startup cutting corners. Lots of companies do this and survive. What happened next was different.
The Cracks: The Content Scandal
The Higgsfield Earn program incentivized creators to make content that generated maximum engagement. That optimization function had no ethical constraints built in.
According to Forbes, Higgsfield offered to pay creators to share clips that were deliberately offensive. Screenshots and three sources confirmed to Forbes that the company distributed marketing materials containing racist content: Shrek characters using anti-Asian slurs, a Moana character declaring racist dialogue, nonconsensual deepfakes of celebrities like Sydney Sweeney and Zendaya.
In December 2025, a Higgsfield user created “Island Holiday,” a video depicting people named in the Epstein files alongside fictional characters. It went viral on X.
The payment side of Earn deteriorated in parallel. While the program claimed a 90% approval rate, creators reported withdrawal problems, sudden account bans, and unpaid work.
Mahi de Silva, Higgsfield CSO: “We fully admit that we push the envelope... it’s more controversial content that gets attention.” He called the racist videos a “mistake” unrepresentative of the company’s values.
In February 2026, two things happened in quick succession. Forbes published its exposé. Then Higgsfield’s X account was suspended for inauthentic behavior. Creator Tim Soret called the hype “fake and it’s bought.” Dustin Thornton slammed the “ruthless tactics.” The internet coined “Shitsfield AI.”
CEO Mashrabov acknowledged missteps: “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.”
The Lesson: When Growth Hacking Eats Itself
This is the part every growth team should study, because the playbook that built $300M ARR is the same playbook that created the scandal. The strategies didn’t fail. They worked exactly as designed. The problem is what “working” looked like when nobody was checking.
The influencer distribution strategy that drove 3B+ impressions also created the incentive for shock content. When you pay creators for engagement and don’t review what they produce, you’re outsourcing your brand to whoever generates the most clicks. The clicks came. So did the racist videos.
The shipping velocity that produced 200+ releases meant the company was moving too fast to build ethical review processes. When you ship daily, there’s no step in the workflow where someone asks “should this template exist?”
The “loudest player wins” philosophy that Mashrabov articulated openly, “cheap acquisition only happens when you stay on top of people’s minds,” translated directly into a culture where controversial content was a feature, not a bug. Until it became the story.
Compare this to the competition. Runway has grown to $90M ARR more slowly but has maintained brand integrity as a professional creative tool. Synthesia reached $146M ARR in the avatar space while building genuine enterprise trust. Neither has a “racist videos” Forbes headline.
(For more on how AI voice and video companies approach growth differently, see my ElevenLabs teardown.)
How ElevenLabs Hit $330M ARR: The Three-Platform Flywheel That Conquered Enterprise AI
Inside the API-first strategy that powers 1B+ users and 41% of Fortune 500 companies
The question hanging over Higgsfield now: does $300M ARR survive when the reputation doesn’t? The company still has 20M+ users, a $1.3B valuation, and backing from Accel, Menlo Ventures, and GFT. Mashrabov has pledged mandatory reviews for marketing materials.
But “Shitsfield AI” is the kind of nickname that sticks. And in a market where Runway, Adobe Firefly, and OpenAI’s own tools are all competing for the same professional creators, trust is the one thing you can’t growth-hack your way back to.
Key Takeaways
The aggregation play is real. You don’t need your own model to build a massive business. Higgsfield built $300M ARR on top of Sora 2, Veo 3.1, and others. Speed of integration is the competitive advantage.
Pivot timing matters more than pivot direction. Higgsfield’s mobile app wasn’t working. Their web pivot didn’t change the core tech, only the audience and packaging. That single decision unlocked everything.
Influencer distribution is the fastest and most dangerous growth channel. It scales virally, costs less than performance marketing, and compounds through social reach. But the content you incentivize becomes your brand, whether you review it or not.
Shipping velocity compounds, but so do mistakes. 200+ releases in 11 months is extraordinary. It’s also 200+ opportunities for something to go wrong without ethical guardrails.
Cashflow positive while growing 60% monthly is the real flex. Most AI startups are burning cash at this stage. Higgsfield proved it’s possible to scale and be profitable.
Reputation is a lagging indicator. Revenue can grow for months while trust erodes underneath. By the time Forbes publishes the exposé, the damage is already baked in. Build the guardrails before you need them.
The AI video market is still early. Higgsfield has the product, the users, and the revenue. Whether they can rebuild the trust is the question that will determine if this is a story of a company that stumbled and recovered, or a company that grew so fast it forgot what it was building for.
This is my personal opinion: the product is genuinely impressive, and the growth strategies are worth studying. But the next chapter of Higgsfield’s story depends entirely on whether they can ship ethics as fast as they ship features.
FAQs
How did Higgsfield AI grow so fast?
Higgsfield reached $300M ARR in 11 months by pivoting from mobile to web for professional creators, aggregating 12+ AI models (including Sora 2 and Veo 3.1), building an influencer-powered distribution program called Higgsfield Earn, and shipping 4-7 feature releases per week.
Does Higgsfield AI have its own AI model?
No. Higgsfield operates as a multi-model aggregation platform, orchestrating 12+ third-party models including OpenAI’s Sora 2, Google’s Veo 3.1, WAN 2.5/2.6, and Kling. They use GPT-4.1 and GPT-5 for planning and scene composition, making them reportedly OpenAI’s largest Sora 2 customer.
What went wrong with Higgsfield AI?
Higgsfield’s Earn program incentivized creators to produce viral content without ethical guardrails, leading to racist and offensive videos. A Forbes exposé revealed misleading marketing, stock footage passed off as AI-generated, throttled “unlimited” plans, and unpaid creators. Their X account was suspended for inauthentic behavior.
How does Higgsfield’s growth compare to other AI startups?
Higgsfield claims to be faster than OpenAI, Slack, and Zoom at the same stage, reaching $300M ARR in 11 months with 20M users. By comparison, Runway reached $90M ARR more slowly, and Synthesia hit $146M ARR — both without major reputation scandals.
What is the Higgsfield Earn program?
Higgsfield Earn is an automated creator monetization program where influencers link their social accounts, join campaigns, and get paid based on engagement. It commissioned 10,000+ creators and distributed over $1M, driving 3B+ social media impressions — but also became the source of the company’s content scandal.
Is Higgsfield AI profitable?
Yes. Higgsfield became cashflow positive at the $50M ARR mark, just five months after their web launch. This is unusual for AI startups, which typically burn cash to subsidize compute costs. However, the company acknowledges that gross margins are below traditional SaaS levels of 80-90%.