Paper Unicorns Are Cheap. Building A $1B Revenue AI Company Takes A Multimarket Cap Table

By Ilman Shazhaev 

A $1 billion valuation still makes a startup a unicorn on paper. Yet when 250 companies joined The Crunchbase Unicorn Board in the first half of 2026, the speed of wealth creation raised another question: How many of these businesses can keep their customers once the initial excitement fades?

I would set a higher benchmark for unicorns with $1 billion in revenue sustained by customers who keep paying.

I spent five years at Harbin Institute of Technology and Shanghai Jiao Tong University, where “fail fast, improve faster” shaped my approach to building technology. Having studied in China before building an AI company in Abu Dhabi, I have also seen how the same growth story can raise different questions among investors, eventually boiling down to: “What will customers keep paying for?”

The revenue test investors should demand

Fast revenue deserves attention. Founders who turn pilots into paying deployments are solving a real problem, and strong sales can make the next round easier. An annualized run rate built from discounts, experimental budgets or generous compute subsidies, however, offers little assurance that customers will stay.

Since 2023, the price of hitting a fixed level of AI performance has dropped thirteenfold every year. Startups that just resell model access now face brutal price competition. Look at OpenRouter and its $113 million Series B: developers can already swap between models through one shared interface.

Businesses built around proprietary workflows face a different test. Software embedded in a company’s approvals, records and daily decisions can be far harder to replace than the model doing the underlying work. I would check whether customers renew at full price and expand their contracts. Then I would see whether margins stay healthy after inference costs.

Why the cap table should cross borders

Once founders know who keeps buying, they should raise across regions, because a local backer can double as market entry. Borders barely slow an AI product down. Enterprise sales still depend on local budgets, procurement procedures and relationships. A foreign investor becomes valuable when those connections turn a product demonstration into a paid contract.

The $30 million Series A for London- and Gulf-based 1001 AI offers a useful example. U.S. venture firm Lux Capital led the financing, alongside Saudi PIF-backed Sanabil Investments. The deal cannot establish future retention by itself, but it shows how capital, talent and customers can sit on different maps, with engineers in London and customers in the Gulf.

Ask every investor what they bring besides money. One warm introduction to the right buyer can beat a higher valuation if it turns into a contract you can actually deliver and renew.

Different money, different obligations

The choice between traditional venture funding and sovereign-backed money is often framed as a choice between fast and patient capital. I find the more practical distinction lies in the obligations attached to each check.

In conversations with U.S. and European venture investors, I have often been asked how quickly we could turn product adoption into repeatable sales before competitors caught up.

Venture firms need returns within their fund’s life. That puts commercial momentum and a sizable exit at the center of their investment decisions. The pressure can sharpen execution. It can also backfire when founders who chase fast expansion without counting delivery costs may fill their pipeline with discounted experiments instead of enduring enterprise relationships.

Sovereign-backed investors usually ask me a very different set of questions. They want to know where user data gets processed and which technical capabilities stay in the country. Will local customers still be able to rely on the service years from now?

Some funds carry an explicit domestic development mandate. For them, those concerns can count as investment criteria alongside financial returns. The upside is a partner who cares about local infrastructure and enterprise adoption. The price can be longer diligence and localization spending. Contract terms that make it harder to expand elsewhere can also cause a problem.

Neither mandate is universal. Many sovereign-backed funds pursue conventional financial returns, while traditional VCs can support long enterprise sales cycles. Government-affiliated capital doesn’t guarantee procurement wins, either.

Founders should read the fine print: governance rights, exclusivity requests, promised commercial introductions. Then judge whether the investor’s regional reach is worth the constraints.

Fund what customers will keep

The best capital strategy follows the customer’s path, from a working product to a renewal. Venture money speeds up development and sales once the business model works. Regional strategic investors earn their place when a new market’s infrastructure and procurement rules call for local expertise.

I would rather show an investor a small cohort of customers renewing without discounts than a glossy revenue forecast that depends on the next funding round. The numbers to watch are net revenue retention, production deployments and margins after delivery costs.

Market sentiment will keep minting and erasing paper unicorns every quarter. The founding teams that build $1 billion in lasting annual revenue will likely balance venture momentum with regional obligations. Their global cap tables will mirror the enterprise markets they serve.

Ilman Shazhaev is the founder and CEO of Dizzaract, an AI and gaming infrastructure company headquartered in Abu Dhabi and the largest game development studio in the Middle East. His work focuses on distributed AI infrastructure, human behavioral data and the interactive environments in which AI systems are trained and tested. Shazhaev is a United Nations expert under the UNODC program, advising on how AI can help solve social problems in developing economies. He is also the author of 46 scientific articles, including published research on AI that reads a player’s psychological state from ordinary gaming hardware, and holds 10 registered invention patents.

Illustration: Dom Guzman