Higgsfield raises $400M Series B with Mirae Asset Capital, hits $5.4B valuation

DST Global led the round. New backers include Tribe Capital, Goldman Sachs Alternatives' Growth Equity arm, Smash Capital, Fifth Wall, Valor Capital, Intel Capital, and Liberty Global Tech Ventures, alongside returning investors Accel and Menlo Ventures. Supermodel Natalia Vodianova Arnault also joined as an investor and advisor.

Here's the scorecard the company shared alongside the announcement:

Metric	Figure
Annual recurring revenue (ARR)	$700 million (about 1 trillion won)
Global users	30M+ across 238 countries/regions
Fortune 500 companies using the platform	390
Monthly content generations	20M+
Total funding raised to date	~$538 million

Mirae Asset Capital and NTT Docomo join the round

For readers in Korea, what makes this round notable is the investor list itself. Mirae Asset Capital, the investment arm of one of Korea's leading financial groups, and NTT Docomo Ventures, the venture investment unit of Japan's largest telecom operator NTT, both came in as strategic investors.

Higgsfield frames this pairing as a foothold for expansion across the Asia-Pacific region. Korea and Japan already rank among its top five markets by global user count, and both countries have some of the world's strongest demand for high-quality commercial advertising and premium entertainment production — a fit the company sees as well matched to its enterprise capabilities.

The new funding will go toward expanding a dedicated Asia-Pacific go-to-market team and recruiting top-tier AI talent.

"Every business needs video content, but producing it at the quality, speed, and scale enterprises demand remains complex and expensive. AI is fundamentally changing that. Our partnerships with Mirae Asset and NTT in particular will be pivotal to bringing our enterprise-grade AI production platform to the vibrant creative economies of Korea and Japan." — Alex Mashrabov, co-founder and CEO of Higgsfield

The founder behind Snapchat's Cameos, back for round two

Higgsfield was founded in 2023. Its founder, Alex Mashrabov, had already built and sold a company in this space once before.

AI Factory, the computer vision startup he co-founded earlier, generated facial expressions and movement from a single photo. That technology became the basis for Snapchat's popular Cameos feature, and Snap acquired the company for roughly $166 million in December 2019. After the acquisition, Mashrabov ran generative AI at Snap, overseeing the My AI chatbot and AR effects, before leaving in 2023 to start Higgsfield.

Co-founder and CTO Yerzat Dulat has an interesting story of his own. He's a machine learning researcher from Kazakhstan with no university degree, someone who built his reputation entirely through open-source work. Oddly enough, his GitHub handle was higgsfield long before the company existed — the same name the startup would later take. That account hosts a GPU orchestration framework built for training models with billions to trillions of parameters.

The company name comes from particle physics' Higgs field — invisible, present everywhere, and the source of mass for particles. The idea is that any concept passing through the platform gains the "mass" of finished video.

Supercomputer: describe it, and the video appears

The engine behind this growth is Supercomputer, an agentic product launched in May 2026 that saw its user base grow 42x within three months of launch.

Despite the name, it's not hardware — it's a chat interface that operates Higgsfield on your behalf. Describe what you want, and the agent plans the work, picks the right models and presets on its own, and produces the final output. Before running anything, it shows exactly how many credits the job will cost and waits for approval, a detail that matters a lot to people managing production budgets.

It also connects to more than 30 tools — Slack, Google Drive, Notion, Gmail, Figma, and others — so it can slot directly into existing team workflows.

For full multi-scene productions, Higgsfield relies on Cinema Studio. Version 4.0, released August 12, supports generations up to 30 seconds at 1080p, accepts up to 50 reference images, and adds roughly 30 new camera-movement presets along with more than 50 color grades. It also introduces a "performance console" for dialing in an actor's emotion and intensity.

It all started with turning camera moves into numbers

What Higgsfield got right from day one wasn't image quality — it was camerawork.

That started with DoP (Director of Photography), the company's proprietary video model unveiled in March 2025. True to its name, it was built to act like a cinematographer, and it launched alongside roughly 50 camera-motion presets — bullet time, crash zoom-in, robo-arm, and more.

That library has since grown to 229 presets, with names evocative enough to picture instantly — earth zoom-out, FPV drone, roll transition, and so on. Users can blend up to two presets in a single generation and tune each one's intensity from 0 to 1 in increments of 0.01. Instead of describing a camera move in words, you dial it in with precision, the way a real production crew would. The presets essentially encode techniques used on actual film sets.

Higgsfield keeps adding proprietary models of its own, including the image model Soul, the reference-based image model Popcorn, and the voice model Speak. On top of that, a single subscription gives access to more than 50 external models — Google's Veo, OpenAI's Sora, Kuaishou's Kling, ByteDance's Seed-Dance, Alibaba's Wan, MiniMax's Hailuo, and others. The platform's biggest advantage is that users never have to switch tools every time a new model ships.

390 of the Fortune 500 already use it

The most striking figure in this round is that 390 Fortune 500 companies are now customers. That means Higgsfield has moved well past being a tool for individual creators and become a working part of enterprise marketing operations.

Its enterprise plan includes unlimited seats, 10x concurrent generation, a shared workspace with approvals and comments, SSO and permission management, and a dedicated account manager. Plugins are available for Premiere Pro, After Effects, Photoshop, DaVinci Resolve, Figma, and Blender, and outputs are cleared for commercial use.

By web traffic, Higgsfield already leads the category. According to one analytics firm's June 2026 tally, the platform pulled in 34.89 million monthly visits, good for a 15.7% share of the AI video tool market.

We've covered work made with Higgsfield ourselves more than once, including a 110-minute AI feature film with a real cast, the first AI-animated film to screen alongside Pixar and Disney at SIGGRAPH, and the K-pop action series "Zephyr".

Education is getting investment too

Higgsfield is also expanding its education programs, framing them as a way to close the AI skills gap.

Higgsfield Academy offers free training in commercial-grade AI video production, and more than 400,000 people have gone through it so far, completing 67,000 lessons.

Higgsfield for Good launches in September 2026. The program helps schools and nonprofits localize multilingual visual learning materials, and CTO Yerzat Dulat is personally leading a STEM-focused initiative for Central Asia.

Editor's take

The AI video industry has had a rocky year. OpenAI shut down its Sora app, and Google replaced Veo with Gemini Omni inside the Gemini app. It's been a year that laid bare just how expensive it is to build these models in-house.

Higgsfield chose a different path: build in-house where it has a real edge, like camera control, and otherwise aggregate the best models in the world so creators never have to think about which one to pick. In a market where new models drop every week, that bet has paid off so far.

Higgsfield's toolkit is already widely used in Korean content production, and it's worth watching what comes next.