AI Funding Austria 2026: KMU.DIGITAL, FFG, aws and the Research Premium for AI Projects

For an AI project in Austria there are four useful routes in 2026. The shorthand: aws has call-based AI implementation grants, FFG funds eligible R&D, the research premium (Forschungsprämie) credits 14 percent of eligible R&D expenditure, and KMU.DIGITAL covers consulting plus small implementation. Availability and eligibility differ by instrument. As checked on 2 September 2026, the official KMU.DIGITAL page reports its budget as exhausted, the aws AI-Start and AI-Adoption calls are closed, FFG's Basisprogramm and Innovationsscheck are open, and AI Ökosysteme closes on 6 October 2026.

This is an editorial overview, not eligibility or tax advice. Rates are not interchangeable: aws AI-Adoption's qualifying 80 percent is a non-repayable grant, the FFG Basisprogramm normally uses a grant-and-loan mix, and the research premium is a tax credit. Check the current program document, call status, aid basis, eligible costs, and cumulative state-aid limits before planning.

The four routes at a glance

What fits what, with the verified key figures. All details as of mid-2026, double-check before you apply.

aws: for adopting AI

aws runs the call-based AI Mission Austria implementation instruments. The current AI-Start: Green page describes a first AI project focused on green technologies or sustainability, implemented with a cooperation partner, at up to 50 percent and EUR 15,000. AI-Adoption targets innovative trustworthy-AI projects, including regulatory and standards preparation, and its last listed call closed on 28 February 2025. It is a non-repayable grant, not a loan. The exact rate depends on the aid basis and company or project category.

FFG: for research and new builds

FFG is the research track. It is relevant when there is genuine development with novelty in it, meaning your own model, a novel method or a real technological barrier you overcome. The Basisprogramm is open in scope and accepts rolling submissions, up to 50 percent for small companies as a mix of grant and loan. The Innovationsscheck is the low-barrier entry point: up to EUR 10,000 in grant, but mandatorily with a research partner. If you do collaborative frontier AI research, look at AI Ökosysteme 2026 under the AIM AT umbrella, with a submission deadline of 6 October 2026. Important for drawing the line: FFG does not fund the mere adoption of existing AI, that is what aws is for.

Research premium (Forschungsprämie): 14 percent on genuine R&D

The research premium credits 14 percent of eligible R&D expenditure for a financial year. For in-house R&D, an annual FFG opinion is requested free through FinanzOnline, while the tax office decides the cost base. Eligible categories can include R&D payroll, direct expenditure, investment and financing costs, attributable overhead, and a notional entrepreneur's salary. Routine work or a standard-software rollout generally fails the Frascati test. The 2026 reform concerns the tax-accounting basis and treatment of direct investments, including a sustainability period and later corrections, not a blanket expansion for near-market work. Get project-specific tax advice.

KMU.DIGITAL: consulting and small implementation

KMU.DIGITAL is not an AI-specific program, but its digital consulting and implementation modules can fit an eligible project. Consulting is funded at 80 percent up to EUR 400 per status analysis and 50 percent up to EUR 1,000 per strategy consultation, within the program's overall cap. Implementation offers up to 30 percent and EUR 6,000 and requires a completed funded consultation. On 2 September 2026, the official page said the KMU.DIGITAL 4.0 budget was fully exhausted and applications were no longer possible.

Stacking and the double-funding ban

The basic rule is simple: the same euro of expenditure may not be funded twice. But different cost items of a project may come from different pots. In practice that means the research premium can be combined with a grant, but the subsidized cost base is reduced before the 14 percent are calculated. You do not claim the premium on the part that was already subsidized. On top of that, the EU state-aid maximum intensities cap the total rate per project, so stacking eventually hits a ceiling. Which combination is permissible in your case is something you settle with FFG, aws and your tax advisor. How the building blocks stack in detail is shown in our piece on stacking aws, FFG and the research premium.

Which program for which project

1. Adopting or integrating existing AI tools: check the exact aws call scope when a new window opens. AI-Start: Green requires a first sustainability-focused implementation with a cooperation partner; AI-Adoption requires an innovative trustworthy-AI project and is not a generic software-purchase rebate.
2. Building something new, genuine R&D (your own model, a novel method): FFG Basisprogramm (rolling) or AI Ökosysteme (call until 6 October 2026).
3. A first small R&D step with a research partner: FFG Innovationsscheck.
4. A tax refund on ongoing R&D: the research premium, provided genuine R&D under Frascati exists.
5. Positioning and strategy before the project: KMU.DIGITAL consulting, provided budget is open again.

"Justify the project first, then look for funding, never the other way around. If an AI project only adds up with grant money, it does not add up. And anyone who sells you the funding as the main argument rarely sells the best implementation."

The honest part

Funding is real and it pays off, but it has side effects that few people talk about. It invites over-scoping, because you inflate the project to fit a program instead of building the smallest possible thing that delivers value. The application effort is real: expert opinions, cost plans, cooperation partners, call windows and documentation can eat up the benefit on small grants. And "we will get you the funding" is not a quality criterion for an implementation partner. Whether an AI project is any good is decided by execution and impact, not by grant money. Funding is an accelerator, not a business model.

Which puts the weight back on the delivery side. Our guide to choosing a software agency covers what to ask instead of asking about grants, and AI software development in Austria sets out what we actually take responsibility for: the model layer, evaluations, guardrails and the product around them. Twinsoft AI is what that looks like delivered, with no funding story attached to it.