Analysis
Crunchbase's Unicorn Board added 250 new companies in the first half of 2026, and most of them are AI startups chasing a $1 billion valuation -- but a sharper framework from Dizzaract co-founder and CEO Ilman Shazhaev, published on Crunchbase News, argues that hitting $1 billion on paper and proving $1 billion in durable revenue are two very different tests -- and most of this year's new unicorns have only cleared the first one. Pulse has previously covered Crunchbase's own funding data closely, including its report on fintech's third-quarter Series A pullback even as overall dollars stayed resilient.
What the "multimarket cap table" argument says
Shazhaev, who runs an Abu Dhabi-based AI and gaming infrastructure company, sets a higher bar than the usual unicorn scoreboard: a true $1 billion-revenue AI company needs customers who keep paying, not just a fast climb to an eye-catching valuation. His framework asks three questions of any AI startup -- do customers renew at full price, do pilots convert into production deployments, and do margins stay healthy once inference costs are netted out. Since 2023, he notes, the price of a fixed level of AI performance has fallen roughly thirteenfold a year, citing Epoch AI research, which is exactly why revenue growth alone is a weaker signal now: it has gotten structurally cheaper to post it.
Who's building this way
- 1001 AI -- $30 million Series A, led by Lux Capital with Saudi PIF-backed Sanabil Investments participating: a London- and Gulf-based enterprise AI company Shazhaev cites as an example of a cap table built for specific-market access rather than brand alone.
- OpenRouter -- $113 million Series B: a model-access reseller Shazhaev cites as a cautionary example of a business that can scale revenue quickly without necessarily building the retention or margin profile his framework demands.
- Dizzaract -- the Abu Dhabi-based AI and gaming infrastructure company Shazhaev founded and runs, which he describes as the largest game development studio in the Middle East.
“We think the framework is right about the mechanism and understates the tension inside it.”
Our read
We think the framework is right about the mechanism and understates the tension inside it. Sovereign-linked capital like Sanabil's does buy market access, but it also comes with a different definition of success than a 10-year venture fund needs -- data-residency commitments and local-capability requirements that can outlast the fund's own return horizon. A founder optimizing a cap table for market access this year can end up negotiating around those same investors' strategic priorities three years later, right when a traditional VC would be pushing hardest for an exit.
Room for disagreement: however, a $30 million Series A at Lux Capital's lead is still a small sample to generalize a "multimarket" strategy from, and plenty of AI companies have scaled on a single geography's enterprise demand without needing a sovereign-linked backer at all. The thirteenfold compute-cost decline cited here also cuts toward lower margins industry-wide, not just toward easier revenue -- which argues for scrutinizing unit economics at least as hard as geographic cap-table composition.
What to watch next
With 250 new AI unicorns minted in just six months, the forcing function Shazhaev describes -- net revenue retention, production deployments, post-inference margins -- is the diligence checklist worth applying to every one of them before the next funding round, not after a down round forces the question.
