Higsfield AI CEO Alex Mashrabov: $200M ARR in two months, sub-$1 video production, and the rise of AI-native ad agencies

Summary

Higsfield AI has crossed $200M ARR, doubling from $100M in just two months — a growth rate that signals the company is catching a significant structural shift in how brands produce video advertising.

CEO Alex Mashrabov points to the emergence of AI-native social media agencies — lean teams of 10 to 30 people — as the primary demand driver. Fortune 500 brands are increasingly routing budgets through these smaller shops rather than traditional agencies, with the explicit goal of using AI to produce commercial ads end to end.

Cost Structure and Unit Economics

The per-video production cost on Higsfield's platform has fallen to sub-$1, compared to an industry benchmark Mashrabov's interlocutors peg at roughly $2,100 per asset for human-generated UGC content — a reduction of approximately 100x. Customers are spending thousands of dollars per month on the platform, implying production volumes in the thousands of videos.

The model cost trajectory cuts both ways. Video models that ran at 10 to 50 billion parameters last year are heading well beyond 100 billion parameters, which will push inference costs higher. Mashrabov frames this as a manageable tradeoff given the capability gains, particularly around persistent brand context — models that remember prior campaigns, brand assets, and performance history without re-uploading each session.

Performance Data Loop and Ad Tech Roadmap

Higsfield has built direct integrations with major social platforms including Meta and is collecting performance data to run reinforcement learning on content creation. Mashrabov frames 2026 as the year of data accumulation and 2027 as the inflection point for full ad tech integration, when connecting generative output to live performance signals becomes the central competitive battleground.

Personalization at Scale

The most compelling near-term use case is mass personalization. Ridge Wallet, cited as a customer by its founder Sean Frank, uses Higsfield to generate custom ads targeting individual college team fanbases — a level of creative localization that was economically unviable with traditional production. Mashrabov's stated goal is batch content generation: hundreds of video variations produced simultaneously, which he describes as building a "Coursera for video."

AI Detection and Consumer Reception

On the question of AI disclosure and detection standards, Mashrabov expects EU-driven regulatory frameworks to materialize by end of year. More striking is his observation on consumer behavior: the best-performing AI-generated ads look distinctly AI-generated, not realistic, and direct-to-consumer brands do not care about the distinction as long as conversion metrics hold. Performance is overriding aesthetics concerns entirely.

Infrastructure Outlook

Mashrabov is watching custom silicon closely, calling out Gemini and a model he references as Banana Pro as early demonstrations of multimodal capability that he argues make traditional pixel-level tools like Adobe Photoshop largely redundant for the social media marketing workflow. The shift is from pixel manipulation to semantic, prompt-based editing — a workflow change with direct implications for the Adobe product suite's relevance in performance marketing.