Bülent Tekmen

Türkiye Will Be a Top-Ten AI Economy

Sep 15, 2026

Bülent Tekmen

Becoming one of the world’s ten leading AI economies will be extraordinarily difficult. It will demand exceptional work, sustained commitment, and the ability to keep learning through failure. We will have to build things that do not yet exist and earn people’s trust in them. I am clear about the scale of that task.

I believe Türkiye will become a top-ten AI economy. This is a future I am committed to helping build.

Top ten is a demanding claim. Türkiye is the sixteenth largest economy in the world, and AI value tends to follow economic size. Ranking tenth means producing more than our weight implies. That has to come from particular industries where we are already better than our average, and from people willing to take responsibility for outcomes nobody has asked them to deliver.

By an AI economy I mean something specific: the capacity to turn intelligence into trusted action: products, companies, exports, and measurable productivity inside existing institutions. Not the number of models trained here, and not the number of people using a chatbot. The question is how much of the value created with AI is captured by companies built here, sold from here, and reinvested here.

The obvious worry about cheap intelligence is that it makes expertise worthless. The opposite is happening.

Between November 2022 and October 2024, the cost of using a model at GPT-3.5-level performance on a language-understanding benchmark fell more than 280-fold. Stanford AI Index That is a comparison at a fixed capability level. But when something gets that cheap that fast, the value does not disappear. It moves — to whatever the cheap thing cannot do on its own.

A model can inspect a part. Someone still has to know which defect matters. A model can score a loan. Someone still has to decide whether its mistakes are acceptable. A model can answer a customer. Someone still has to answer for it when it is wrong. None of that got 280 times cheaper. Some of it got more valuable, because there is suddenly much more to point it at.

That is the whole opportunity. The scarce input is no longer intelligence. It is the person who knows what to do with it — and three years ago, that person’s tools cost 280 times more.

Türkiye has a great many of those people, in startups, factories, in financial institutions, and in businesses that have sold abroad for decades. AI gives us a chance to turn what they know into products.

It also has something harder to measure. I believe Türkiye produces entrepreneurs with unusually high agency: the ability to take responsibility for an outcome, find a way through uncertainty, bring people together, and keep going until something works. George Mack’s High Agency describes the quality well — thinking clearly, taking action, and questioning assumptions that others accept.

The clearest Turkish example is Baykar. It was founded in 1986 as a machine-parts company. In 2000 it began researching unmanned aircraft. It flew its first autonomous test in 2004, delivered its first mini UAV in 2007, flew the Bayraktar TB2 fully autonomously in 2014, and exported it in 2018. By 2025 it had $2.2 billion of exports — 88 percent of its revenue — and TB2 contracts in 36 countries. Baykar

For scale, compare the most celebrated defence technology startup of the past decade. Anduril, founded in 2017, reported $2.2 billion of revenue in 2025 and raised $5 billion in May 2026 at a valuation of $61 billion; most of its revenue comes from its own government. Source Baykar’s 2025 revenue was $2.5 billion, and 88 percent of it came from abroad. Anduril has far more capital. But in 2025 the Turkish company was the larger business by revenue, and the more international one.

That is what agency looks like when it is sustained: a team that decided to build a capability it did not yet have, and kept building — through small deliveries, long tests, and years when the market was not there — until it could sell to the world.

The question for a country is how to produce that story many times over. TEKNOFEST is one answer. The aerospace and technology festival drew more than a million visitors to Istanbul in 2025, and more than 565,000 teams and 1.1 million competitors applied to its competitions across 58 technology categories. More than a million people applying to build something is not a festival statistic. It is a talent pool, and the work is to turn it into companies. Some of them will found the Baykar of an industry that does not exist yet.

Baykar is not alone.

Togg, founded in 2018 by five Turkish companies, started production in 2022, and passed 100,000 electric vehicles delivered by early 2026. Source Peak built Toon Blast and Toy Blast in Istanbul and sold the company to Zynga for $1.8 billion in 2020. Dream Games built Royal Match into one of the highest-grossing mobile games in the world and was valued at $5 billion in 2025. Trendyol became Türkiye’s first decacorn at a $16.5 billion valuation in 2021 and is now expanding into the Gulf. Yemeksepeti pioneered online food ordering here and was sold to Delivery Hero for $589 million in 2015, then the largest technology exit the region had seen. Insider sells its software to global brands from Istanbul and raised $500 million in 2024, led by General Atlantic. And fal.ai, founded in San Francisco in 2021 by Burkay Gür and Görkem Yurtseven, runs the models behind products at Perplexity, Canva, and Quora, and was valued at $4.5 billion in December 2025.

Some of these companies passed through Endeavor Türkiye, which has selected 79 ventures and supports 128 entrepreneurs. The Turcorn 100 programme, launched in 2022 to help Turkish startups reach billion-dollar valuations, had 43 companies in its candidate pool by mid-2026 against a target of 100 by 2030. Source If you want to know who comes next, those two lists are the shortest reliable answer.

These companies took different paths. I draw one lesson from all of them: Turkish teams can turn specialised knowledge into products the world buys. AI gives us a way to repeat that across more of the economy.

The common starting point is people. Many of those who could help lead this already work at the frontier abroad. Koray Kavukcuoğlu runs Gemini as Senior Vice President of Google DeepMind. Tufan Erginbilgiç runs Rolls-Royce. Bülent Altan led avionics work at SpaceX. Ekin Doğuş Çubuk co-founded Periodic Labs. We should connect that experience to teams building here through joint laboratories, company partnerships, and sustained mentoring, so that what they know compounds at home as well as abroad.

I see the clearest opportunity in work I know closely: financial services. It is the strictest test of this thesis, because every decision has a customer, a regulator, and a consequence attached, and someone is accountable for it. Turkish financial institutions digitised early and operate at scale in a large, young, mobile-first market; the judgment their people have built doing that is exactly the kind of tacit knowledge cheaper intelligence makes more valuable. And it exports without a factory. A better way to assess risk or handle insurance claims, built inside an institution and proven on real customers, exceptions, and local rules, can be sold to institutions in markets that look like ours — the Gulf, Central Asia, Africa — where banks face the same conditions and want systems they can operate themselves.

Institutions also need systems they can operate under their own control, with custody of their data and a way to keep working when a supplier changes. I wrote about why that matters in When the Models Go Dark. Building and supporting those systems is another export opportunity.

These opportunities fit businesses we already have. Their first customers can be close enough to work with every day, and that proximity lets a team learn quickly. The ambition should be to turn what it learns into a product that works internationally.

Startups need affordable computing and patient capital. People need practical ways to learn new skills, ideally through real work.

Other countries can do this too. Our advantage will depend on how well we connect capable teams to customers, capital, and distribution.

Textiles is the example I think about most. Türkiye is the world’s fifth largest textile exporter, at $12 billion, tied with Italy — behind China at $141 billion, then India, the United States, and Germany. Source We hold that position without the advantage the industry was originally built on. Production costs that once favoured Türkiye now favour other countries, and Turkish firms still keep their place — on speed, quality, and closeness to their customers rather than on price. That is the lesson. A cost advantage travels. Know-how stays — as long as the people who hold it do. Our next export industries should be built on knowledge from the start.

We have to create our own edge.

Patrick Collison’s Fast is a useful reminder of what determined teams have done with far less. The Empire State Building was constructed in 410 days, in 1930 and 1931. Kelly Johnson’s team designed and built the XP-80 jet fighter prototype in 143 days, in 1943. Tony Fadell started work on the iPod in January 2001, and the first units reached customers that November, roughly 290 days later. All of this was achieved decades before modern AI tools. A team that can now call on reasoning, research, coding, and design whenever it needs them should be asking what those benchmarks look like today, and adding Turkish examples to the list.

Sam Altman puts the discipline in one sentence: “Plans should be measured in decades, execution should be measured in weeks.” What I Wish Someone Had Told Me The ambition can span years. Every week should bring something closer to reality: a working product, a solved problem, a better answer, a team capable of more than it was. I collect the ideas behind this — Mack’s High Agency, Collison’s Fast, Altman on ambition and execution — on my reading page.

We should judge progress by exports, repeat customers, and measured productivity gains, and publish those numbers every year. Research and skilled work must grow here too. A successful company should train people and generate capital that help the next one get started. Bringing people together around that shared ambition is one reason we want to establish the Türkiye Artificial Intelligence Foundation.

I have also put a date on being wrong. By 2030, at least ten Turkish companies should each be earning more than $100 million a year from AI products sold abroad, and at least one of them more than $1 billion. If that has not happened, I will have been wrong, and I will say so rather than move the date.

At Colendi, we are putting this conviction into practice. We founded the company in 2017: a payment, settlement and credit-scoring platform by 2020, an AI banking platform in 2024, an AI-native insurance platform in 2025, and today a licensed digital bank with banking-as-a-service infrastructure.

Our ambition is to make ColendiBank the first and largest AI-native bank in the Middle East and Africa by 2030.

Banking is where this thesis gets tested most strictly. For a bank, a demo is not enough. It has to turn intelligence into decisions it is accountable for — credit, risk, payments, claims — under supervision, at scale, for customers who will notice when it is wrong. That is what we are building: intelligence turned into trusted action, on rails we operate ourselves.

Türkiye’s place among the leading AI economies will be earned by the people who decide to build that future.

Colendi is our commitment to it.

If it works, the next company should be easier to build here.

Notes