What government support exists for AI projects in Türkiye?
As of October 2026, an AI project in Türkiye has three main doors: grants from TÜBİTAK's Technology and Innovation Support Programmes Directorate (TEYDEB), loans and interest support from KOSGEB, and the investment incentive certificate issued by the Ministry of Industry and Technology. Only two schemes are specific to AI: the TÜBİTAK 1711 Artificial Intelligence Ecosystem Call and the KOSGEB AI Credit. The others are general programmes with no topic restriction, so an AI project can use them if it meets their conditions.
This article is for information only. Hatson Tech does not prepare funding applications and does not advise on KOSGEB or TÜBİTAK applications. We checked everything below against the official programme pages on 22 October 2026. Rates, limits and calendars change from call to call; KOSGEB's AI Credit directive alone was amended twice in September and October 2026. Read the current call text before you decide anything. All official pages linked here are in Turkish.
What is the difference between a grant, a loan and an incentive?
A grant is not repaid but requires genuine R&D; a loan is repaid but needs no R&D; an incentive reduces the tax and financing burden of an investment.
- Grant (TÜBİTAK 1501, 1507, 1707, 1711): A share of accepted spending is paid to you and never repaid. The project must contain an unsolved technical problem, pass reviewer or panel evaluation and report every period.
- Interest-free loan (KOSGEB AI Credit): Finance tied to buying specific services, secured by a bank guarantee.
- Interest support (KOSGEB SME Digital Transformation): A bank lends; KOSGEB covers the interest on the loan used for approved purchases.
- Investment incentive (incentive certificate): Tax, customs and financing support for a fixed investment registered on the certificate.
In practice, connecting an off-the-shelf AI service to existing software is usually not R&D and will not fit a grant; a loan, interest support or an ordinary budget is the better route.
Which programme fits which kind of AI project?
The table separates the programmes by two questions: who applies, and who does the R&D? Figures are as stated on the official pages on 22 October 2026.
| Programme | Who applies? | Form of support | AI project it fits | Condition to check |
|---|---|---|---|---|
| TÜBİTAK 1501 Industrial R&D | An SME-scale capital company | Grant: 75% for the first five projects, then 60%; TÜBİTAK's contribution capped at TRY 20 million per project | Product or process development with real technical uncertainty, run by your own team | If the R&D sits mainly with a contractor, that is a critical finding |
| TÜBİTAK 1507 SME R&D Start-up | An SME-scale capital company | Grant: 75%; project budget up to TRY 3.5 million; up to 18 months | A narrower project from a company new to R&D | For a firm's first five projects, at least two of them collaborative |
| TÜBİTAK 1707 Order-Based R&D | A customer company of any size together with at least one SME supplier | Customer pays at least 40% of the supplier's costs; TÜBİTAK grants the supplier 40% of accepted spending | Having an SME develop an AI solution for you as an R&D project | Market research and a technical feasibility study are expected |
| TÜBİTAK 1711 AI Ecosystem | A consortium of a customer, an AI SME and a university or public research centre | SME customer pays 10% of the provider's costs and TÜBİTAK grants 70%; for a large customer, 20% and 60% | A sector-specific AI solution in one of the priority areas | Opens once a year; applications without a consortium are not evaluated |
| KOSGEB AI Credit | An SME holding a Technology Venture Badge (Teknogirişim Rozeti) | Interest-free loan up to TRY 5 million per business; 24 months, first 12 without repayment | Cloud and AI services such as compute, storage, data processing and model use | Only from KOSGEB's listed providers; a bank guarantee letter is required |
| KOSGEB SME Digital Transformation | A small or medium-sized manufacturer | Interest support on a bank loan of TRY 1 to 20 million | Production projects that involve buying cameras, servers, software and hardware | A maturity report from an authorised consultant; purchases must be new |
| Investment incentive certificate (Digital Transformation Programme) | The investing company | Tax, customs and financing support | Larger investments that integrate technology into business processes | Spending before the application date is not covered |
What do TÜBİTAK 1501 and 1507 fund, and who can apply?
1501 and 1507 are general programmes that give non-repayable grants for R&D carried out by SME-scale capital companies; unless a call says otherwise, there is no topic restriction. According to TÜBİTAK's 1501 programme page, sole proprietorships, ordinary partnerships, cooperatives, foundations and associations cannot apply. The support rate is 75% for a firm's first five supported projects and 60% from the sixth; TÜBİTAK's contribution is capped at TRY 20 million per project, and a project cannot exceed 36 months.
The 1507 SME R&D Start-up Grant Programme is for companies starting out in R&D: a 75% rate, a project budget of up to TRY 3.5 million and up to 18 months, aimed at a firm's first five projects provided at least two are collaborative. Both run through calls that open twice a year. Applications go through TEYDEB's PRODİS system with an e-signature, after the company has completed organisation-level pre-registration. Eligible costs include personnel, travel, equipment and software, materials, consultancy and services.
For AI projects, priority topics matter. The second 1501 call of 2026 awards 5 extra points to projects under the "Technological Leap in Industry" and "Digital Leadership" headings of TÜBİTAK's 2026–2028 priority list, and that list places AI and data science under Digital Leadership. In the same call, an organisation may submit at most two proposals across 1501 and 1507, each lasting up to 24 months.
Which programme fits if a supplier will build the software?
If an outside SME will develop the AI solution as an R&D project, look at 1707 and 1711, not 1501. The 1501 call text lists, among the critical findings that can reject a project before review, the case where the applicant's contribution to the R&D is absent or limited and the R&D will be done by the contracted organisation. Buying services is an eligible cost, but the weight of the R&D has to sit with the applicant.
In the 1707 Order-Based R&D call, a customer company of any size applies jointly with at least one SME supplier, which does the R&D. The customer pays the supplier at least 40% of each period's project costs, and TÜBİTAK grants the supplier 40% of accepted spending. The budget is capped at TRY 10 million and the duration at 24 months; the customer's own staff for testing and field validation are supported up to 18 person-months. A market study and technical feasibility review are expected in the proposal.
The 1711 Artificial Intelligence Ecosystem Call is specific to AI and opens once a year; the 2026 call closed on 2 October 2026. The consortium includes the customer, at least one SME as technology provider and at least one university research lab or public research centre, with TÜBİTAK's Artificial Intelligence Institute also part of the process. If the customer is an SME, it pays 10% of the provider's declared costs and TÜBİTAK grants the provider 70% of accepted spending; for a large customer the figures are 20% and 60%. The customer's own staff time for absorbing the technology is supported at 60% up to 18 person-months, and support lasts at most 24 months. The 2026 priority areas were smart manufacturing; smart agriculture, food and livestock; fintech; climate and sustainability; and smart education.
In both programmes, TÜBİTAK deals with the customer: the company with the need runs the application, not the software firm. Companies that count as related parties under tax law cannot be customer and supplier in the same project. Intellectual property and the use of project outputs are settled in the cooperation agreement between the parties, so decide at that stage who may use your data and for what.
What does KOSGEB offer for AI projects?
KOSGEB's AI-specific scheme is the AI Credit; manufacturers can also use the SME Digital Transformation Support Programme. Neither is a grant: both are tied to finance you repay.
The AI Credit carries no interest or commission; the cap is TRY 5 million per business over 24 months, the first 12 free of repayments. Applicants need a valid Technology Venture Badge (a Ministry of Industry and Technology certificate for technology-based ventures) and a GO Dijital wallet account, and the credit is issued against a definitive bank guarantee letter. It can only be spent with providers on KOSGEB's published list and on items in its eligible expense table, which currently covers high-performance compute, data storage, data processing, AI platforms and foundation model use. In short, it finances cloud and AI infrastructure; it is not general funding for commissioning custom development from a software company.
The SME Digital Transformation Support Programme is for small and medium-sized manufacturers (NACE Section C); micro businesses are excluded. Before applying, the company needs a digital transformation or maturity assessment report from a consultant authorised by TÜSSİDE, MEXT or İHKİB. KOSGEB pays the interest on a loan from a partner bank of TRY 1 to 20 million, with a maximum term of 36 months, for machinery, equipment, software and hardware approved by its board. Purchases must be new.
When does an investment incentive certificate matter for an AI project?
The incentive certificate matters when the AI project is part of a fixed investment, such as a plant renewing hardware, software and automation together. The Decree on State Aid for Investments (No. 9903, published on 30 May 2025) defines investments that integrate technology into business processes to cut costs and raise productivity and quality as the "Digital Transformation Programme", and lists them as priority investments.
Three details matter to buyers. Spending made before the certificate application date is not covered, so applying after you have started paying is too late. The 25% cap on intangible assets such as licences and know-how does not apply to Digital Transformation Programme investments. And the decree's lira amounts are updated every year by the revaluation rate, so check minimum investment amounts against the current rules rather than an old summary. Certificates are issued electronically through E-TUYS, and the decree covers applications made until 31 December 2030.
What makes an AI project fundable?
A fundable project states an unsolved technical problem, who will solve it and how the result will be measured. The critical findings in TÜBİTAK's call texts are explicit about what gets rejected: projects whose R&D was finished before applying, routine work, work behind the current state of technology, investment-heavy projects, proposals with too little information, and budgets that do not match the R&D can all be rejected without review. So can a project team with no company employee holding at least a bachelor's degree relevant to the subject. Test your project against this scorecard before you start writing:
| Question | Weak signal | Strong signal |
|---|---|---|
| What technical problem will be solved? | "We will integrate AI into our processes." | One paragraph says what current methods cannot do and why the outcome is uncertain. |
| Who does the R&D? | A service provider does all of it; the applicant only pays. | In 1501 and 1507, the core work is in the applicant's team; in 1707 and 1711, customer and supplier roles are separate in the work packages. |
| How will success be measured? | "Productivity will improve." | Current measurement, target value, test data and method are defined. |
| Is the data ready? | "We have lots of data." | Which data, who owns it, whether it contains personal data and who grants access are known. |
| Is the budget tied to work packages? | One large hardware line, unexplained consultancy. | Every cost line maps to a work package and an output. |
| Is the team adequate? | No employee with a relevant degree. | Relevant staff and the project lead are named. |
| Is the timeline realistic? | Spending has started; the application will come later. | Pre-registration, e-signature and any incentive certificate are in the plan. |
This scorecard is not an official form; it shows early that the word "AI" is not an R&D justification on its own. Connecting a ready-made language model API to existing software is usually routine integration; using a model to solve a measurable, domain-specific problem that known methods cannot solve may be R&D. For writing scope with acceptance criteria, see our guide to scoping an AI project.
Hypothetical example: visual quality inspection on a production line
This example is hypothetical and does not describe a real company. A mid-sized metal parts manufacturer wants to catch surface defects with cameras. Inspection is visual today, and a trial with an off-the-shelf vision system missed fine scratches on reflective surfaces. The company has four needs, and each leads to a different door:
| Need | Candidate programme | Condition to check |
|---|---|---|
| Develop the defect detection method with its own engineers | TÜBİTAK 1501, or 1507 if among its first projects | Is the R&D weight in the company's team, and is there an employee with a relevant degree? |
| Have an AI SME develop the method | TÜBİTAK 1707; 1711 if it fits smart manufacturing and a university lab joins | The manufacturer applies; 1711 opens once a year |
| Buy cameras, lighting, industrial PCs and servers | KOSGEB SME Digital Transformation | Is there a maturity report, and is the equipment new? |
| Use cloud compute for model training | KOSGEB AI Credit | Without a Technology Venture Badge this route is closed |
The first decision is who does the R&D. If the company has no engineer with vision experience and will hand the work entirely to a supplier, applying to 1501 accepts the critical-finding risk from the start; 1707 is designed for this arrangement. Hardware is a separate financing decision, with the maturity report to finish first. Check each programme's rules on whether one cost can be supported twice. For the technical preparation, our article on computer vision for quality control covers the steps from camera placement to threshold selection.
Which mistakes sink an application?
The most common mistake is shaping the project to fit a programme instead of choosing the programme that fits the project. Practical warnings from the call texts:
- Presenting finished work as R&D: A project built on a system that already works can meet the finding that the R&D was completed before applying.
- Outsourcing everything: In 1501 and 1507, an applicant with no or limited R&D contribution can be rejected.
- Padding the budget with hardware: Investment-heavy projects and costs that do not contribute to R&D are critical findings.
- Leaving prerequisites to the last day: TEYDEB requires organisation-level pre-registration and a valid e-signature, and the 1501 call text requires the pre-registration documents to reach TÜBİTAK at least one working day before the call closes.
- Relying on an old source: Rates and priority areas change between calls; the 1711 page notes that bonus points were removed for this call.
How should you follow the current call?
Rely on the programme's official page and that period's call text; summaries, including this one, are only a starting point. 1501 and 1507 open twice a year and 1711 once a year; three 1707 periods were planned for 2026. On KOSGEB pages, check the directive's revision date. While reading, note who can apply, which costs are eligible and what leads to outright rejection. For the cost lines that make up the whole project, see our article on what an AI project costs.
What Hatson Tech does and does not do here
To be clear: Hatson Tech does not prepare KOSGEB or TÜBİTAK applications, does not advise on them, and makes no statement or commitment on whether a project will be funded. This article answers a question buyers often ask, from official sources.
Our field is the project itself: which work is worth doing with AI, whether the data is good enough and how a pilot will be measured. For organisations that want those questions settled first, AI consulting is a separate, paid engagement whose output is a written priority list, data assessment and pilot scope, usable however the work is financed. For software projects, the first meeting and discovery are free, and we can look at the technical side of your project together there.