TwelveLabs closing a $100 million Series B to build what it calls "video superintelligence," co-led by NEA and NAVER Ventures with Amazon, Radical Ventures and Index Ventures participating, is a useful marker of a shift underway in growth-stage AI investing: nine-figure checks are increasingly going to narrow, vertical AI applications, not just horizontal foundation-model labs.
TwelveLabs' pitch is specific -- its Marengo 3.0 model understands sound, speech and motion across video at scale, turning raw footage into a searchable semantic layer for enterprise customers -- rather than a general-purpose foundation model competing with GPT or Claude directly. The company plans to use the new capital to open offices in New York and London, on top of existing San Francisco and Seoul operations, chasing global enterprise demand for video-specific AI tooling.
The pattern isn't isolated. LeapXpert's $180 million growth round for AI-powered governed enterprise communications, and Celea Therapeutics' $180 million Series B for a single biotech indication, both landed in the same window, each representing a narrow, defensible application rather than a horizontal platform play. Three different verticals -- video AI, compliance AI, biotech -- all commanding checks that would have been considered foundation-model-scale just two years ago.
“Three different verticals -- video AI, compliance AI, biotech -- all commanding checks that would have been considered foundation-model-scale just two years ago.”
Context matters here: in 2023 and 2024, nine-figure-plus AI rounds concentrated almost entirely in a handful of foundation-model labs -- OpenAI, Anthropic, Mistral, Cohere -- competing directly on general-purpose model capability. The fact that video-understanding, enterprise-compliance and single-indication biotech companies are now commanding similar check sizes suggests investors have grown comfortable underwriting vertical AI applications as standalone, venture-scale businesses rather than treating them as smaller bolt-ons to the foundation-model story.
For growth-stage investors, this is a validation of the 'picks and shovels don't have to be horizontal' thesis: a well-defended, enterprise-traction-backed vertical AI company can now credibly raise at the same scale as infrastructure plays, provided it owns a genuinely hard technical problem (video understanding, regulatory compliance, drug development) rather than a thin wrapper around an existing foundation model.
For founders, the implication is that a strong vertical AI thesis with real enterprise or clinical traction no longer needs to pretend to be a platform play to raise at scale -- specificity itself has become fundable at nine-figure size, provided the moat is real.
The bear case: not every vertical will support this size of round, and the risk of overcapitalizing narrow, single-application companies before their market size is proven remains real -- TwelveLabs, LeapXpert and Celea are all still pre-scale relative to their new valuations.
What to watch: whether other vertical AI categories -- legal, insurance, logistics -- see similar nine-figure rounds in the second half of 2026, and whether any of this cohort's next rounds come at markups or down rounds once growth metrics are tested against these valuations.