Türkiye replaced its first national AI strategy (2021–2025) with a sharper document in August 2026: the 2026–2030 AI Action Plan. It has 16 priority actions across four axes, which the plan calls notice, benefit, produce and govern, and it puts numbers on almost everything. GridStudio was founded in Istanbul in 2011 and still runs engineering there, so we read it closely.
The infrastructure numbers
| Target | By |
|---|---|
| At least 1 GW of AI and data-centre capacity | 2030 |
| At least $10 billion of investment, mainly private | 2030 |
| At least 1 exabyte of storage | 2030 |
| A 100 Tb/s national backbone | 2030 |
| 10 TWh a year of low-carbon power purchase agreements | 2030 |
| At least 2% of the public investment programme budget for AI projects | Plan period |
The power target matters as much as the capacity target. A gigawatt of AI load needs firm, low-carbon supply, cooling and grid connections that take years to secure. Projects that start site and power work in 2026 will be the ones that can use the plan’s incentives.
Compute for the long tail
The plan’s GPU credit programme starts at 2 million GPU-hours a year for startups, SMEs and researchers and rises to 20 million GPU-hours a year by the end of 2028. There are also at least 1,000 AI vouchers for SMEs in the first year, a National AI Research Fund of at least 10 billion lira and a 15 billion lira AI Growth Fund.
For enterprises, this changes the build-or-rent question. Experimentation can move onto subsidised capacity, and owned infrastructure can be sized for steady production load instead of peaks.
Government as the first customer
At least 30 pilot applications across 10 ministries in the first 12 months, and at least five AI-supported citizen services piloted and scaled on e-Devlet. At least 2,000 public datasets are to be published in machine-readable form in the National Data Library, 100 of them via API.
Those datasets are what make public-sector AI practical: an assistant is only as good as the records it can reach.
Models, languages and export
The plan calls for at least 10 licensing or joint-development agreements for sectoral foundation models by the end of 2028, at least 25 corporate model deployments abroad, and a Turkic-languages large model covering the Oghuz, Kipchak and Karluk families, built with Organization of Turkic States members. The first version is due by the end of 2027.
That is a regional opportunity. Kazakh, Uzbek, Turkmen and Azerbaijani speakers are poorly served by general models, and governments across Central Asia are building national AI services.
Sandboxes and skills
Regulatory sandboxes are planned in at least five sectors: finance, health, energy, mobility and telecommunications. On skills, the plan targets 5 million people through a national AI literacy programme within two years, and 10,000 advanced AI specialists plus 100,000 AI application professionals by the end of 2027.
What to do now
- Ministries and agencies: pick pilot services with clear rules and digital records, and design the audit trail before the pilot starts.
- Investors and operators: secure sites, power and cooling plans for AI-density racks now; capacity, not chips, will be the constraint.
- Enterprises: move experiments onto subsidised GPU capacity when it opens, and keep production workloads on capacity sized for steady load.
- Everyone: check KVKK data-transfer rules before choosing where a model runs.
GridStudio’s Istanbul team works on AI infrastructure, AI energy infrastructure and Turkish-language model work. See our Türkiye page.
Questions
It was put into force by a Presidential Circular published in the Official Gazette in August 2026, and reported on 18 August 2026.
At least 1 GW of AI and data-centre capacity by 2030, at least 1 exabyte of storage, a 100 Tb/s national backbone, 10 TWh a year of low-carbon power agreements and at least $10 billion of mainly private investment.
A programme of GPU-hour credits for startups, SMEs and researchers, starting at 2 million GPU-hours a year and rising to at least 20 million a year by the end of 2028.
The Ministry of Industry and Technology coordinates it, the National AI Board monitors implementation, and the first progress report is due within 12 months.