Alex Mashrabov

AI · AI video generation
@alexmashrabovUnited StatesAge 31

Co-founder and CEO of Higgsfield

A competitive-programming alumnus who co-founded AI Factory, sold it to Snap, stayed to lead generative AI, then burned through most of a second seed round before pivoting toward marketers and reporting over $1 billion in annualized revenue

One-line read: Mashrabov is a serial founder who first became a number two, achieved an exit, and then returned as the number one—the second outcome came not from building his own model, but from packaging other people's models into a workflow marketing teams could order directly, and using a creator network instead of paid advertising.

Started after workingSerial founderScaled company$100M+ · 2026-09VC-backedSilicon ValleyFrom UzbekistanBorn Mar 1995Top-school trainedBig-tech backgroundForbes 30 Under 30ICPC World Finals 2014 / 2016

Key numbers

Each figure includes its definition, date, and evidence level.

1995Born · age 31
~23First venture · 2018
Self-reported$1B↗annualized revenue · 2026-09
Self-reported$5.4B↗Series B valuation · 2026-08
Media report$400M↗Series B funding · 2026-08
Self-reported30M+↗registered users · 2026-08
Media reportapproximately $166M↗AI Factory acquired by Snap · 2020-01

Key stages

What he was doing at each age

24

Company sold to Snap

29

Burned through most of the seed round

30

Shifted toward marketers

31

Annualized revenue surpassed $1 billion

Geographic moves

The city behind each change in direction

  1. Moscow
    19 Competitive programmer
  2. San Francisco
    ~28 Co-founder and CEO29 Burned through most of the seed round30 Shifted toward marketers30 Co-founder and CEO30 Co-founder and CEO31 Co-founder and CEO31 Annualized revenue surpassed $1 billion

Selected work

Higgsfield

An AI creation platform that aggregates multiple video and image models, offering shot presets, character consistency, and finished-content workflows for advertising and social media

AI Factory

A mobile computer-vision company behind Snapchat Cameos' face-swapping short-video feature; acquired by Snap

How the business works

Sector
AI · AI video generation / creative tools
Users and payers
Social media marketers, advertising agencies, and brand content teams, plus a large number of individual creators. An interview in June 2026 put agency revenue at about 70%; in September 2026, he said just over half of revenue came from enterprises—the figures have different dates and definitions and cannot be compared directly
Pricing
Subscriptions + credit usage. The official blog listed the individual plans in September 2026 as Starter $15/month, Plus $49/month, and Ultra $129/month ($99/month billed annually), plus seat-based team plans and enterprise contracts; plan names changed more than once during 2026
Customer acquisition
He says the company does not buy paid ads and relies mainly on creator distribution and social sharing; this was also the source of the controversy in February 2026
Main costs
Usage fees and compute for third-party video models (Veo, Kling, Seedance, and others); one of the main uses of the Series B funding is securing compute capacity
Evidence level
Profitability is self-reportedAll revenue figures are annualized and come from the company or Mashrabov; “cash-flow positive” appears only in interview statements; in a September 2026 LinkedIn post, he said gross profit had been positive since the start of the year and net revenue retention was 300%, without giving a specific gross margin. Profit has not been disclosed
Moat
The product is not the model, but the workflow layer of shot presets, character consistency, and marketing templates, plus the speed of integrating new models (analytical judgment)
Main risks
Upstream model providers raising prices, restricting supply, or building their own applications; content moderation and brand safety; user trust problems caused by early “unlimited” plans
Growth flywheel · analysis

Integrate a new model as soon as it launches → creators use it to make shareable videos → sharing brings free traffic → marketing teams pay for reusable workflows → revenue buys more model credits and compute (analytical judgment)

Higgsfield does not rely on a single in-house model. It connects multiple video models, including Google Veo, Kling, and Seedance, in one interface, then layers on finished-content workflows such as shot-motion presets, character consistency, Cinema Studio, and Marketing Studio. In interviews, Mashrabov acknowledged that betting early on a single in-house model and chasing leaderboards was a detour.SaaStr interview

People pay not to “generate a video,” but to deliver advertising and social media assets more cheaply and quickly: the same interview put average annual customer spending at about $1,000 and annual subscriptions at about 40% of revenue. This explains how it could grow without buying ads—its customers are people who make a living from content.

There is no public data on the profit side. Video is one of the most compute-intensive generative tasks, and the more revenue grows, the more money goes to upstream model providers; the company says it is cash-flow positive but has not disclosed its gross margin. So what can be confirmed here is how the money moves, not that the business has proven profitable.

Timeline

  1. Birth

    Starting out

    In a March 19, 2026 LinkedIn post, he wrote that he had turned 31 “a few days ago,” so his birth is recorded as March 1995; the exact date is not public and is not inferred. No primary source for his birthplace has been found.

    Source ↗
  2. Represented the Moscow Institute of Physics and Technology at the ICPC World Finals

    19 years oldStarting outRussia · Moscow · Competitive programmer

    He placed 53rd with the MIPT team in 2014, then represented Innopolis University at the World Finals in 2016 and placed 38th; in between, he placed third in CodeChef SnackDown and eighth in VK Cup in 2015. In an interview, he said he reached the “top three in the world” at 19, and that for children from post-Soviet countries, international competitions were almost the only way out.

    Source ↗
  3. Co-founded AI Factory

    About 23 years oldTurning pointCo-founder (not the lead)

    The company was led by Looksery founder and former Snap engineering director Victor Shaburov after he left Snap in May 2018, with Mashrabov and Greg Tkachenko as co-founders. Before that, Mashrabov was a machine-learning engineer on Yandex's search team (according to Kommersant). The company had an office in San Francisco, and most of its roughly 70-person team was in Ukraine. No public primary source describes his exact responsibilities at the time.

    Source ↗
  4. Snap acquired AI Factory; reported transaction price approximately $166 million

    24 years oldExitCo-founder

    The price was first reported by Ukrainian media outlet AIN; Snap confirmed the acquisition to TechCrunch but not the amount. This was the company's transaction valuation, and how much each founder received has not been disclosed. A 2026 Higgsfield press release gives the acquisition year as 2019, while media reports disclosed it in January 2020.

    Source ↗
  5. Stayed at Snap to lead generative AI

    About 25 years oldTurning pointHead of generative AI at Snap (executive retained after the acquisition)

    He went from co-founder of an acquired company to an executive leading a team at a major platform, working on filters and generative features serving hundreds of millions of users. Different sources use both Director and Head for his title.

    Source ↗
  6. Left Snap and founded Higgsfield with Yerzat Dulat

    About 28 years oldTurning pointUnited States · San Francisco · Co-founder and CEO

    His first startup as the lead founder. Dulat is CTO; Mahi de Silva is chief strategy officer and is not listed in the founders' section of the company press release, but Mashrabov himself calls him a co-founder on LinkedIn.

    Source ↗
  7. Raised a seed round for a consumer mobile app, but the direction did not work

    29 years oldFailureUnited States · San Francisco · Co-founder and CEO

    In April 2024, the company announced an $8 million seed round led by Menlo Ventures for Diffuse, a mobile video app for ordinary users. He later said in interviews and on LinkedIn that the seed stage raised $16 million in total, burned more than $10 million, failed to produce a usable product for a full year, and was even advised to shut down; he described the single in-house model and consumer product as tuition.

    Source ↗
  8. Web version launched publicly, shifting toward professional creators and marketing teams

    30 years oldTurning pointUnited States · San Francisco · Co-founder and CEO

    The same generative technology was repackaged for desktop, with shot-preset templates and finished-content workflows, and sold to people who make a living from content. Revenue began growing rapidly afterward.

    Source ↗
  9. Series A expanded to $130 million, with a $1.3 billion valuation and $200 million in annualized revenue

    30 years oldMilestoneUnited States · San Francisco · Co-founder and CEO

    About nine months after launch, users exceeded 15 million; annualized revenue doubled from approximately $100 million in two months. Accel, AI Capital Partners, Menlo Ventures, and GFT Ventures participated.

    Source ↗
  10. Forbes investigated racist marketing videos and unpaid creators

    30 years oldFailureUnited States · San Francisco · Co-founder and CEO

    The report said the company had distributed racist content and unauthorized celebrity deepfake material to creators, and that some creators had difficulty getting paid. Company executives acknowledged that the videos were wrong. The growth method itself became the company's main reputational risk.

    Source ↗
  11. Raised $400 million in Series B financing at a $5.4 billion valuation

    31 years oldMilestoneUnited States · San Francisco · Co-founder and CEO

    DST Global led the round. The company announcement also disclosed $700 million in annualized revenue, more than 30 million users, and 390 Fortune 500 companies using the product. The valuation quadrupled in eight months.

    Source ↗
  12. Self-reported annualized revenue reached $1 billion

    31 years oldMilestoneUnited States · San Francisco · Co-founder and CEO

    Mashrabov disclosed the figure to Bloomberg; the calculation is revenue for the last four weeks multiplied by 13. It is a revenue run rate, not audited annual revenue; the $500 million in June, $700 million in August, and $1 billion this time may not use identical measurement windows.

    Source ↗

He was not the number one in his first startup

The AI Factory story is often compressed into “Mashrabov sold the company to Snap for $166 million.” At the time, reports identified Victor Shaburov as the leader—he founded Looksery, which sold to Snap in 2015, then worked as Snap’s engineering director before leaving to start another company in May 2018. Mashrabov and Greg Tkachenko were co-founders, and the team had about 70 people, most of them in Ukraine.

The significance for what came later is not the amount, but two experiences: he lived through the complete process of turning computer-vision technology into a consumer feature and selling it as a whole to a platform; after the acquisition, he stayed at Snap to lead generative AI, for the first time running a product line serving hundreds of millions. Those experiences gave Higgsfield the conditions for fundraising and an early team.

The tuition for his second startup

In 2023 he left Snap and founded Higgsfield with Yerzat Dulat, this time as CEO. The initial route was the one he knew best: build a proprietary video model and package it as a mobile application for ordinary users. In April 2024 the company announced an $8 million seed round and the mobile product Diffuse.

That route did not work. He later said in an interview that more than half of the seed money was burned before they found direction, and described both the single in-house model and the consumer product as tuition.

What happened after the pivot

On March 31, 2025, the web version launched publicly for professional creators and marketing teams. The product stopped emphasizing “our model” and instead connected multiple models, building shot presets, character consistency, and an advertising-production workflow on top.

The subsequent numbers came from the company or Mashrabov and use annualized revenue: about $200 million in January 2026, $500 million in June, $700 million in August, and $1 billion in September. Valuation rose from $1.3 billion in January 2026 to $5.4 billion in August. These are financing prices and revenue run rates, not profit.

The other side of the growth model

Higgsfield’s distribution depends heavily on creators: it pays creators to publish videos made with the product and gets traffic through sharing rather than ads. In February 2026, Forbes reported that materials distributed to creators included racist content and unauthorized celebrity deepfakes, while some creators said they could not receive payment. Company executives acknowledged that the videos were wrong.

This belongs next to the revenue curve: the same mechanism explains both why the company grew quickly and where its greatest risk lies.

Open verification items

Unknown numbers stay blank instead of pretending to be reliable.

  • The birth month and year come from his March 19, 2026 LinkedIn post (“I turned 31 a few days ago”), recorded as March 1995; the exact date is not public. The source link is his activity page, not a permanent link to the individual post.
  • Place he grew up: one biographical profile says he grew up in Andijan, Uzbekistan, and he said in an interview that his father was from Uzbekistan; competition records list Russia, and Russian-language media call him Russian. His origin is provisionally recorded as Uzbekistan pending confirmation from a primary source.
  • The meaning of “top three in the world at 19” is unclear: the interview excerpt refers to competitive programming, while his LinkedIn post refers to international draughts.
  • The seed-round amount has two reported figures: the April 2024 announcement said $8 million, while he has repeatedly said the seed stage raised $16 million in total. This may include an unannounced extension and has not been verified.
  • AI Factory acquisition timing: the company press release says 2019, while the media disclosure came in January 2020; Snap did not confirm the $166 million figure, and personal proceeds have not been disclosed.
  • The years he worked at Yandex could not be found; regarding education, one Russian-language database says he holds computer-science master's degrees from MIPT and Innopolis and attended Berkeley Haas, but this is the only source.
  • Operating figures such as “cash-flow positive,” “gross profit positive since the start of the year,” 70% of revenue from agencies, and average annual customer spending of about $1,000 come only from his statements and have not been independently verified.
  • The selection year and list category for Forbes 30 Under 30 have not been verified.
In their words
We burned more than 10 million out of 16 million raised in seed fundraising.
—— Alex Mashrabov · Sep 28, 2026 ↗

More quotes

By the age of 19, I was able to get to top three in the world in competitive programming.
Sep 28, 2026#Starting pointSource ↗

20VC interview, excerpted by Scuttlebutt. The closest result in public competition records is third place in CodeChef SnackDown in 2015; he separately said on LinkedIn that he reached the world's top three in international draughts as a child, and the two statements may not refer to the same thing.

The next wave of value will be created by the applications that put this technology to work, and Higgsfield is leading that shift in visual media.
Aug 17, 2026#Application-layer viewSource ↗

A signed quotation in the Series B press release, representing the company's public position.

Video is one of the most compute-intensive domains in AI.
Aug 17, 2026#CostsSource ↗

Explains why the Series B funding needs to be used heavily for compute.

My read

Mashrabov is a serial founder who first became a number two, achieved an exit, and then returned as the number one. Two things are especially easy to misread. First, treating $166 million as his personal achievement—AI Factory was led by Victor Shaburov, and Mashrabov was one of its co-founders; that figure was the company’s transaction valuation, and Snap never confirmed it. Second, reading Higgsfield as a “video-model company”—it only took off after abandoning a single in-house model.

Three mechanisms actually mattered, and each can be copied separately. First, sell the same technology to a different group of payers: consumer applications had poor retention and low average revenue per user; after switching to marketers and agencies, customers used the product to deliver client work and were willing to pay for saved production costs. Second, sell the workflow, not the model: shot presets, character consistency, and finished-ad templates are the layer model providers do not want to build but marketing teams use every day; speed of integrating new models is itself a product. Third, let users’ work advertise for you: video tools naturally produce shareable content, and he says the company does not buy ads.

There are three cautions. One, this is not an ordinary starting point: world-class competitive-programming results, one exit, and experience leading a generative-AI team serving hundreds of millions at Snap underwrote fundraising and hiring; being able to burn more than $10 million of seed capital and continue is itself a tolerance for error most people do not have. Two, $1 billion is “the last four weeks × 13” annualized run rate, not profit or realized annual revenue; every dollar of video-generation revenue first shares value with upstream model providers, and that margin is the first thing squeezed if they raise prices, restrict supply, or compete directly. Three, growth has a cost: creator distribution brought reports of racist content and unpaid creators in February 2026; the scale gained through speed and the reputational bill charged by speed are the same mechanism.