The count is 25. The category breakdown is the point.
BestBrokers and Crunchbase report that 98 startups have reportedly reached unicorn status year-to-date in 2026, with AI companies representing approximately 25% of new entrants, roughly 25 firms. That headline figure matters less than what the 25 are building. The 2023-2024 unicorn wave was foundation model-heavy. This one isn’t.
Three names lead the reported valuation list for AI entrants, all sourced to Crunchbase and BestBrokers with a disclosure: these company names are as reported and haven’t been independently verified against primary filings for . Ineffable Intelligence, a UK-based AI company, is reportedly valued at $5.1B with $1.1B in disclosed funding, making it the highest-valued new AI unicorn in the class if the figures hold. Two US-based companies, reported as “humans&” and “Ricursive Intelligence,” are reportedly valued at $4.5B and $4.0B respectively. The ampersand in the first name and the non-standard spelling in the second are as reported by Crunchbase; confirm these against primary sources before acting on them.
Robotics is the structural shift worth watching. BestBrokers data estimates robotics companies accounted for approximately 11 of the new unicorn class, reportedly making it the second-largest category after foundation/application AI. That’s an estimate from a single source, not a primary count, but it’s consistent with the broader capital flow pattern this quarter.
The real story is that the 2026 unicorn class reflects where infrastructure constraints are creating value, not where model capability is. Companies building compute, embodied AI hardware, and energy-adjacent AI infrastructure are reaching billion-dollar valuations because the frontier labs need what they’re building. Foundation model companies have consolidated around a handful of heavily capitalized incumbents. The infrastructure layer hasn’t consolidated yet.
That’s the pattern repeating across ‘s capital data. Wayve’s $1.05B Series C in embodied AI, orbital compute investment, and now a robotics-led unicorn class, the common thread is physical infrastructure for AI deployment, not another LLM.
Don’t conflate this with a rotation away from AI software. Enterprise application AI is still generating unicorns. The shift is at the category level: the fastest-growing new unicorn cohort isn’t consumer AI or foundation model competitors, it’s the companies solving the build-out constraints that are slowing frontier model deployment.
Watch the Q3 unicorn data for whether robotics holds second place or gives way to energy-adjacent AI infrastructure. If AI data center grid constraints deepen through summer, and current capacity data suggests they will, the energy infrastructure layer is the likely next unicorn cluster. The infrastructure pivot isn’t a trend. It’s a response to a specific physical bottleneck.