The Central Asian Ecosystem and Unicorns: An Honest Conversation About Resources and Reality

Bakht Niyazov, city — Almaty, founder of Justart VC

Recently, a startup joined our venture studio. Great people with a deep understanding of a specific and potentially lucrative niche. I asked them to go through the accelerator because I believed structure and networking would not hurt. They agreed. But they dropped out of the program after a couple of hours.

I tried to convince them to stay. Their response was roughly this: “Maybe it’s better to spend these two or three months on something more sensible than talking to a tracker about various topics? In that time, we could earn a couple of hundred thousand dollars. And if we need help, we’ll ask you to find us a targeted advisor from the relevant industry.”

I had nothing to argue. And it was precisely this short story that made me think seriously.

Over the last decade, a full-fledged startup infrastructure has formed in Central Asia. Accelerators, hubs, grants, competitions, hackathons, demo days. The state has invested significant resources; the private sector has invested time and money; hundreds of teams have gone through programs. There are success stories. There are live companies that have grown.

But if we look honestly: not a single unicorn or unicorn candidate in the region has emerged from a cohort accelerator. And I think the issue is not that people worked poorly. The issue is that the world has changed, and the question today is: does the current ecosystem model match the reality we are in? And if not, what is a better use of resources?

The Decade’s Result: What We Got from the Acceleration Model

An honest assessment requires acknowledging both achievements and limitations. Local accelerators solved a primary task: they created a funnel, introduced the culture of entrepreneurship, and grew the first generation of founders who know what product, PMF, and unit economics are. Isn’t that cool?

But the secondary task — producing companies with global potential — remains unsolved. There are virtually no large tech companies born within these programs. The average “graduate” of an accelerator is a small local business that survived but did not scale properly. There is nothing shameful in this — as long as we honestly admit what we are actually producing.

The problem begins when we continue to use the same model with the same processes and expect a different result. Many say: we will show results in our market first, then capture all of Central Asia, then go to MENA, then to the US. We continue to force startups to write all this in their pitches. Has any of them actually succeeded this way? No. Because it doesn’t work that way. It hasn’t worked that way for a long time. But no one explains this during the training. Why?

What Artificial Intelligence Changed: Why the Old Logic Stopped Working

All of the above is not news. Criticism of accelerators has been heard before. I myself have criticized them. But now, two factors have converged that make the gap between the old model and reality critical.

First is speed. In the AI era, a startup can go from idea to first users in weeks, not months. Tools that previously required a team of ten people and six months of development are now available to a single founder with a laptop. This means that any program that “freezes” a team for two or three months in a cohort format inherently operates in a slowdown mode, not an acceleration mode. Wow!

Second is the devaluation of basic knowledge. The main product of a traditional accelerator has always been education: marketing basics, financial models, sales scripts, go-to-market strategy. By 2025–2026, any strong language model can generate all of this in minutes, for free, customized for a specific market and sector.

The informational advantage that lectures and trackers once provided has effectively disappeared.

What remains? What AI cannot provide: human judgment from someone who has built at scale. Investor trust that opens the right doors. Access to capital at the right moment. The ability to see potential in a founder before the market does. This is today’s real added value of venture infrastructure — and this is exactly what is almost missing from the classic accelerator model.

Anatomy of the Tracker Model: What Exactly Doesn’t Work

The methodological anchor of most CIS programs is the figure of the tracker. This model came from Russia, where it was actively promoted by a Russian player who was “toxic” in many aspects, and became the de facto standard. The first trackers also came from Russia. It is worth understanding what a tracker actually represents in practice.

A tracker is usually a hired employee who has undergone short-term training. Their task: weekly meetings with the team, metric control, HADI cycles, moving through the process. By design, this is an instrument for reducing variance: it pulls up laggards, levels the cohort, and increases the survival rate of the average project.

But venture is pure capitalism. The entire portfolio economy is built not on average results, but on one or two extreme cases. This means that the correct infrastructure should be optimized for producing such super cases, not for the survival rate of the median portfolio. A model that systematically reduces variance mathematically cannot produce those couple of super cases. This is not a question of people’s quality or effort; it is a question of design.

Another flaw. Finding Product-Market Fit is not filling out tables or following regulations. It is entrepreneurial judgment: the ability to read the market, make the right pivot at the right moment, feel when to push and when to stop. This cannot be taught through HADI cycles. And certainly, a person who has never searched for PMF for their own product cannot teach this.

As a result, a structural paradox arises: a hired manager without operational scaling experience and sometimes without understanding the essence of entrepreneurial risk leads a startup founder down a path they have never walked themselves. Some of them now position themselves as super-mentors; some have moved to Uzbekistan hoping to earn money. But this gap is not accidental. It is built into the system’s architecture. When I see some former trackers teaching startups how to grow, I involuntarily recall Nassim Taleb’s remark about ornithologists who seriously taught migratory birds how to fly correctly.

A Case That Speaks for Itself

In 2025, the first true tech unicorn with Kazakh roots appeared — Higgsfield AI, a generative video platform. Today, the company’s valuation exceeds $4 billion — with an annualized run-rate revenue that has surpassed $400 million and continues to grow. The company is openly targeting $1 billion ARR by the end of 2026, which, given current dynamics, makes it a real contender for decacorn status — a company worth over $10 billion — already this year. The company rejected an acquisition offer from Meta. Investors include Accel, Menlo Ventures, and GFT Ventures.

Co-founder Alex Mashrabov previously led the generative AI division at Snap. The team was built with a global focus from day one. About 95% of the engineering staff are from Kazakhstan.

Higgsfield did not go through local accelerators. Not because they are bad. But because the founders already had the necessary set for the global market: experience in a world-class industry, direct access to investors, and a clear understanding of what they needed to build. A program with trackers would not have added anything; it would only have slowed them down.

This is the key question for the ecosystem: how to build infrastructure that finds, supports, and does not hinder such founders?

Comparison: How Infrastructure Works Where Unicorns Appear Systematically

The key difference, of course, is not in the quality of the program. It is in the economics of incentives. Y Combinator earns only when the founder earns. An accelerator monetizing educational content earns regardless of the outcome. These two models produce fundamentally different results — even at the same level of effort and team diligence.

Where to Direct Resources

If the goal is the systematic production of companies with real global potential, then investment should go to three things that AI cannot replace.

Selection, not training. The true value of a strong institution is not the program, but the signal to the market and the quality of selection. Finding the right founder and removing obstacles from their path is more important than “retraining” a random founder. This means: hyper-targeted matching with a mentor from the relevant industry, rather than general lectures for all cohorts at once.

Capital and deal access, not regulations. Smart money at the right moment, open doors to international investors, willingness to share risk with capital — this is what truly changes a startup’s trajectory. A venture studio where an experienced entrepreneur

builds together with the founder as a co-founder, rather than controlling them as a tracker, is a different model of participation and a different responsibility for the result.

Global orientation from day one. Unicorns from small markets do not scale within them; they scale outward. Infrastructure tailored for a local demo day and local market produces local companies. Infrastructure with direct access to international capital and global distribution produces others.

Conclusion

This is not a criticism of people or a denial of what has been done. I myself recently launched our accelerator in Tashkent with the MOST team. But I am interested in responding to changes. Our acceleration wave, specifically our IRA program, fulfilled its task of its time; it created a primary ecosystem where there was none before. Now there is a next-level task.

The world has changed. AI has rewritten the rules of speed, access to knowledge, and infrastructure requirements. A founder with a phone today has access to tools that five years ago required a full team and a funding round. In this reality, the scarcest resource for a startup is not information or a mentor with a methodology. What is more important now is time, capital, distribution, and the trust of the right people at the right moment.

The question is not whether the old model worked correctly. The question is what we build next — and for whom.