·--:--:-- ·Istanbul UTC+3 · Live
EN TR

Türkiye's AI Action Plan Takes Effect: 1 GW of Data-Centre Capacity and 10 TWh of Low-Carbon Contracts by 2030

Türkiye's AI Action Plan 2026-2030 Enters Into Force: Energy Is the First Layer of the Architecture, with 1 GW of Capacity, 10 TWh of Low-Carbon Purchase Agreements and a 60 Per Cent Low-Carbon Condition by 2030

UzEnergyNews · Top Story · ·Türkçe sürümü →

Presidential Circular 2026/9, signed on 17 August and published in the Official Gazette on 18 August 2026, brings Türkiye's Artificial Intelligence Action Plan 2026-2030 into force, replacing the 2021-2025 national strategy; the plan itself was unveiled at the Türkiye Artificial Intelligence Summit in Istanbul on 13 June, with the full text published by the Ministry of Industry and Technology. It is built as four axes of four actions each on a five-layer architecture in which energy is the first layer. The compute action sets three energy targets for 2030: at least 1 gigawatt of AI and data-centre installed capacity, at least 10 terawatt-hours a year of low-carbon electricity purchase agreements, and a condition that at least 60 per cent of contracted supply be low-carbon — renewables or nuclear — tracked through verifiable source certificates. The transmission operator TEİAŞ and the energy regulator EPDK appear on the action's cooperation list; grid and energy optimisation is a priority pilot area, and energy is among five sectors named for a regulatory sandbox. The parameters that turn the targets into load — megawatts, the low-carbon share and the procurement calendar — are assigned to a National AI Compute Strategy due within six months.

The circular, signed on 17 August and carried in the Official Gazette of 18 August 2026 under No. 33344, puts the Artificial Intelligence Action Plan 2026-2030 into force and repeals Circular 2021/18, which had established the National AI Strategy for 2021-2025. The plan was presented publicly on 13 June 2026 at the Türkiye Artificial Intelligence Summit in Istanbul, and the full text is published on the Ministry of Industry and Technology's site, prepared by its National Technology and Artificial Intelligence directorate. The architecture is explicit: sixteen actions in four axes — Fark Et, İstifade Et, Üret, Yönet, rendered by the state news agency as Recognize, Benefit, Produce and Govern — resting on five technical layers, energy first, then computing hardware, infrastructure, models and data, and applications, with talent, financing, and regulation and trust as cross-cutting components.

The Energy Targets of Action 5

Action 5 — titled, in the plan's own words, building a competitive and low-carbon national compute backbone — carries the plan's energy numbers. By 2030: at least 1 gigawatt of installed AI and data-centre capacity, alongside at least 1 exabyte of storage and a 100-terabit-per-second national backbone; at least 10 terawatt-hours a year of low-carbon electricity purchase-agreement volume; and a condition that at least 60 per cent of contracted electricity be low-carbon, defined as renewables plus nuclear, monitored through verifiable source certificates (plan pages 25-26). The action's cooperation list includes the Ministry of Energy and Natural Resources, the transmission system operator TEİAŞ and the energy regulator EPDK. The plan attaches an investment mobilisation target of at least 10 billion dollars, weighted to the private sector, with incentives routed through the HIT-30 programme's data-centre and AI calls; there is no separate budget line attached to the energy targets themselves.

Where Energy Recurs in the Plan

Energy runs through the rest of the document. Grid and energy optimisation is one of four priority areas of the sectoral pilot programmes, with energy networks named among the test sites (Action 8). AI Growth Zones — special-status areas with energy, land and telecoms infrastructure prepared in advance — are to host multi-tenant AI Factories described as production campuses powered by low-carbon electricity (Action 9). The investment package for international cloud providers includes an energy connection and a low-carbon power purchase agreement, with electricity costs presented comparably to investors (Action 13). Pilot programmes with Turkic States, Organisation of Islamic Cooperation and D-8 partners begin with agriculture and energy applications, and joint data-centre investment with shared compute capacity appears in both the international action and the compute action itself (Actions 15 and 5). Energy is one of five sectors named for a regulatory sandbox, alongside finance, health, mobility and telecommunications (Action 16). The Investment Office is to run a single point of contact covering environmental, zoning, licensing, energy and telecom applications for investors.

The Arithmetic of 1 GW and 10 TWh

The two headline numbers sit in a defined but unfinished relationship. One gigawatt running for all 8,760 hours of a year draws 8.76 terawatt-hours; matching 10 terawatt-hours implies facility overhead of about 14 per cent on top of a fully loaded gigawatt, and any load factor below full use widens the gap in the other direction. Read differently, if the 10 terawatt-hours is the low-carbon share and that share is 60 per cent of contracted volume, total contracted supply would be near 17 terawatt-hours — which no reading of 1 gigawatt reaches. The plan does not choose among these readings: power usage effectiveness and water usage effectiveness are defined as monitored indicators in the action's framework without numeric targets, and the parameters that would settle the question — electric power in megawatts, the low-carbon share and the procurement calendar — are assigned to a National AI Compute Strategy and long-term investment plan due within the plan's first six months. On grid access, the plan lists grid-connection capacities, renewable and nuclear matching scenarios and waste-heat recovery among the items the National AI Board will evaluate, with the results feeding the siting and capacity planning of the Growth Zones; it does not define an allocation mechanism for data-centre connections, does not name a certificate scheme for the 60 per cent condition — the existing YEK-G guarantee system is not referenced — and does not set a tariff class for data-centre electricity.

The Market the Targets Land In

Türkiye's market already has its first hyperscale anchor: Turkcell and Google Cloud announced the country's first hyperscale regional data centre in November 2025, an investment its parties put at 3 billion dollars over ten years. EY's Europe Central study of May 2026 places Türkiye among the markets expected to take a growing share of Europe's data-centre build-out as the established hub cities saturate. The questions this publication has tracked in other jurisdictions now carry a Turkish schedule: the National AI Compute Strategy within six months of the plan's publication; the National AI Board's evaluation of connection capacities; any EPDK treatment of data-centre load; and the first public progress report, due within twelve months on the National AI Progress and Transparency Portal, annual thereafter with an independent evaluation component.

Sources: Presidential Circular 2026/9 (Official Gazette, 18 August 2026, No. 33344); Artificial Intelligence Action Plan 2026-2030, full text published by the Ministry of Industry and Technology — energy provisions in Action 5, pages 25-26 (unveiled 13 June 2026 at the Türkiye Artificial Intelligence Summit); Anadolu Agency (13 June and 18 August 2026); Invest Office of the Presidency and Hürriyet Daily News (November 2025) for the Turkcell-Google Cloud data centre; EY Europe Central, AI and Energy: The Two-Way Dependency (May 2026). Prior UzEnergyNews reporting: Paying for the New Load — Tariff Design for Hyperscale Demand (Weekly Insight #14, 5 August 2026). Copyrighted material is not reproduced. Analysis by UzEnergyNews.