Higgsfield passes $1 billion in annualized revenue as TCV and NVIDIA extend its Series B

TCV and NVIDIA join a round that closed in August

Higgsfield, the AI video and image platform used by brands, agencies and creators, said on Friday, 23 October 2026, that it has extended its Series B with new money from TCV and NVIDIA. According to the company's announcement on PR Newswire, the extension builds on the Series B that DST Global led in August, with Goldman Sachs Alternatives, Tribe Capital, AI Capital Partners, Smash Capital, GFT Ventures and Valor Capital also taking part.

Neither Higgsfield nor its new backers said how much the extension adds or whether it changes the company's valuation. When Higgsfield announced the original round in August, it said it had raised $400 million at a $5.4 billion valuation with annualized revenue of $700 million. The new release sets that revenue figure aside for a much larger one.

TCV is a growth investor whose past bets include Netflix, Spotify, Facebook, ByteDance and Electronic Arts, and the release leans on that list to frame Higgsfield as a media company rather than just another model wrapper. NVIDIA's participation was disclosed the same day Jensen Huang made his first visit to Kazakhstan, where Higgsfield's co-founder and chief technology officer, Yerzat Dulat, comes from.

We reported the NVIDIA stake from the Kazakhstan side on Saturday, in our story on how Firebird won a US licence for 45,000 NVIDIA GPUs during Huang's Astana visit. NVIDIA did not say how much it put in then, and the Higgsfield release does not say either.

The headline number: $1 billion a year, 18 months in

The figure Higgsfield wants people to remember is revenue. The company says that 18 months after launch it crossed $1 billion in annualized revenue in September. Annualized revenue usually means a recent month's revenue multiplied by twelve, so it is a run rate rather than a year of booked sales, and the company did not publish audited accounts or say which month the calculation used.

Even read cautiously, the jump is large. Going from a $700 million run rate in August to more than $1 billion by September, if both figures are measured the same way, would mean roughly 40 percent growth in about a month. Higgsfield attributes much of it to enterprise customers. It says enterprise is its fastest-growing business, growing more than 100 percent month over month, and that revenue under contract from business customers has grown tenfold since June.

The company also cites spending data from Ramp, the corporate card and expense company, which it says identifies Higgsfield as the fastest-growing image and video AI vendor, taking market share from rival platforms faster than any other provider. Ramp's data covers spending by its own customers, so it shows business adoption among those companies, not the whole market.

On usage, Higgsfield claims more than 40 million users in 238 countries and territories. The United States is its largest single market and Europe its largest region. It says it processes 12 trillion tokens across 150 million generations every month. All of these figures come from the company, and none has been independently verified.

What Higgsfield actually sells

Higgsfield does not present itself as a single model. It brings together leading video, image and audio models with its own creative tools, and sells the result to teams that need a lot of finished visual work. The release says brands and agencies, including Fortune 500 companies, use it to turn a single campaign concept into up to 100 ad variants for different markets and platforms.

That aggregator approach has shown up in its product news. Earlier this month Higgsfield connected its Katana tools to Claude so that an assistant can drive its generation tools directly, and it has also pitched a product that turns a product page into a finished ad campaign. Rivals are moving toward the same space from the model side; HeyGen, for example, recently launched a model that makes picture and sound in one pass.

The company's argument for why this business can be so large is about the economics of tokens. It says industry benchmarks put the effective yield of generative video at $30 to $150 per million equivalent tokens, compared with $0.15 to $0.60 for text. In other words, the same compute earns far more when it produces video. Commercial work also tends to be generated many times over, because a campaign needs a first version, variants to test against each other and versions for each market and language.

Higgsfield also points to broad advertising numbers to size its market: a WPP Media forecast that global advertising revenue will reach $1.3 trillion in 2026, and an IAB estimate that US digital video ad spending will exceed $80 billion this year, with nearly two in three buyers using generative AI for video creative, up from half in 2025.

Why NVIDIA is in the round

For NVIDIA, Higgsfield is a showcase customer for the idea that the money in AI is moving from training models to running them at scale. "AI factories produce intelligence in the form of tokens and the application layer is where those tokens become useful," said Rev Lebaredian, NVIDIA's vice president of physical AI simulation, in the release.

Higgsfield says it has run on NVIDIA hardware at every stage, from L40 and A40 GPUs through the Hopper and Blackwell generations, and is now expanding to NVIDIA's next generation, Vera Rubin. It uses NVIDIA Dynamo, an inference serving framework, to deploy custom workloads on its own GPUs. Its 15-minute animated short film, Kök Börü, was made on Blackwell Ultra systems, and the company open-sourced the project and turned it into a free course on Higgsfield Academy, where it says more than 150,000 creators have completed lessons.

Chip makers investing in their own customers is common in this cycle, and it cuts both ways. It signals confidence, but it also means part of a startup's funding returns to the investor as hardware spending. The release does not say whether any purchase commitments came with NVIDIA's stake.

Kazakhstan's first AI unicorn

Higgsfield is headquartered in San Francisco but keeps close ties to Kazakhstan through Astana Hub, the country's technology park. It was co-founded by chief executive Alex Mashrabov, Kazakhstani AI engineer Yerzat Dulat and Mahi de Silva. Mashrabov previously co-founded AI Factory, the computer vision company behind Snapchat's Cameos and face filters, which Snap bought in 2019. The release calls Higgsfield Kazakhstan's first AI unicorn.

"Every brand, studio and creator now needs more visual content than traditional production can supply, and a new category is forming to meet that demand," Mashrabov said. TCV founding general partner Jay Hoag placed Higgsfield alongside the media platforms the firm has backed over 30 years, and general partner Neil Tolaney said the team was delivering "a pace of product velocity and market adoption that we have rarely seen."

The company says the money will go to product development, global infrastructure, hiring and expanding its sales and marketing.

What to watch

The questions now are about durability and margins. A run rate built on monthly subscriptions from individual creators can move quickly in both directions, which is why Higgsfield is stressing contracted enterprise revenue. It has not said how much of the $1 billion comes from enterprise, how much it pays the outside model providers whose models it offers, or what its gross margin is.

Competition is also intensifying. Model makers such as Google, OpenAI, Kling and HeyGen sell video generation directly, and each new model release can shift which platform creators prefer. Higgsfield's bet is that brands want one place to manage campaigns across many models. The new investors are betting that it can hold that position; the next disclosed revenue figure will show whether September's growth continues.