AI funding 2026: Austria, Germany and Switzerland

This selection was checked against official sources on 26 September 2026. It is a starting point for business projects, not a complete funding database or an award commitment. The responsible institutions decide individual applications. Check the conditions again before committing to work.

The table summarises the status or next check for the eight programmes covered in this article, based on the official sources checked on 26 September 2026.

A consulting grant supports a defined advisory activity. That does not establish that software development, licences and maintenance will also be paid for. An implementation programme may cover different costs and require a demonstrable operational change. A loan can improve financing terms but generally remains repayable. Research funding requires a suitable innovation or research objective.

Start with the activity that needs financing. “We want to use AI” does not say whether you intend to investigate a process, connect existing software or develop something technically novel. Describe the intended result in one sentence, for example: “Incoming supplier invoices will enter a traceable approval workflow.” Compare programmes only once that result is clear.

The table compares the four funding instruments by their main decision question and a common misunderstanding.

This distinction also improves discussions with suppliers. A quote should show which activity belongs to each category. A single “AI package” line makes both the application and the later project review harder to assess.

KMU.DIGITAL 4.0: The official website says the budget is fully exhausted and applications for consulting and implementation support are no longer possible. It also states that the programme will not continue from 2027. Previous grant rates should therefore not be advertised as currently available savings. KMU.DIGITAL status.

Carinthia's AI implementation support: The WKO page, dated 9 September 2026, also reports an exhausted budget and no further applications. Its older description of supported services remains visible beneath that notice. The status notice is the relevant information for a new decision. WKO Carinthia.

aws digitalisation support for trustworthy AI: The programme page describes completed calls and says there is currently no further call. Treat it as historical programme information rather than an open application opportunity. aws programme page.

Vienna Business Agency digitalisation support: The published terms include a 50% rate, a €50,000 maximum grant, a €10,000 minimum project and deadlines of 30 September and 31 December 2026. It targets established SMEs operating in Vienna for at least one year. Selection is competitive; exclusions include internal staff costs and standard office software. Official programme overview.

A business located outside Vienna cannot assume eligibility for the Vienna programme. Equally, an AI idea in Vienna is not automatically eligible. Prepare your location, legal form, operating history, project size and intended improvement before speaking to the funding team. A supplier's proximity to the funding region does not establish the applicant's eligibility.

The FFG AI Mission Austria overview describes an umbrella initiative spanning research and business implementation. That page alone does not establish a current call suitable for your company. A research proposal needs a separate review of the particular call, timetable, partners and criteria.

BAFA: consulting support

BAFA's business-consulting programme supports individual conceptual advice, including organisational and process matters. The federal funding database also describes exclusions, including advice directed towards selling the consultant's own goods or services. Do not simply relabel an implementation package as eligible consulting. Official programme description.

BAFA's funding guide lists up to €3,500 of eligible consulting costs and regional rates of 50% or 80%, with maximum grants of €1,750 or €2,800. Confirm current conditions, consultant eligibility and the application procedure with the responsible institution. BAFA funding guide.

KfW 511/512: digitalisation loan and supplementary grants

KfW's ERP digitalisation loan 511/512 is a financing product. It distinguishes basic, LevelUp and HighEnd digitalisation. For the latter two levels, KfW describes supplementary grants of 3% and 5% of the disbursed loan, capped at €200,000. Apply through the financing partner before the project starts, including the supplementary grant where relevant. KfW product information.

Repayments, interest, security and payment dates remain central to the business decision. A cheaper loan cannot make an undefined automation project economical. Continue to budget for operating the finished system even when implementation is financed. Programmes in individual German states may also be relevant, but require their own review against the original conditions.

The Innosuisse innovation cheque supports a preliminary study with a Swiss research partner. Its published conditions include up to CHF15,000 towards the research partner's costs, a Swiss registered office, fewer than 250 full-time positions and an identified research partner. Applications can generally be submitted at any time; approval remains an individual decision. Innosuisse innovation cheque.

Not a routine purchase. An ordinary chatbot purchase does not automatically constitute a suitable preliminary study. Define the unresolved question: what remains unknown, why research is needed, what the partner will investigate and which result will inform a potential follow-on project. “We need a new website with AI” is not a sufficient research description.

Synthetic example. A synthetic example could investigate whether a previously unresolved document class can be processed under specified quality requirements. The study might evaluate data quality, candidate methods and feasibility. Innosuisse determines whether the proposal qualifies. Subsequent product development, integration and maintenance need their own plan and financing.

Without an application. The distinction is useful even when no application is ultimately submitted. If a supplier can already implement the task using a supported standard feature, paying for an elaborate research framing may add little value. If feasibility really is uncertain, making that uncertainty explicit helps the business avoid treating an experiment as a promised production outcome.

Build a short register for your project rather than collecting attractive grant percentages. Each programme entry should include:

• Its name, responsible institution and direct official source.
• The date checked and actual application status.
• Location, company-size and applicant requirements.
• Supported activities, excluded costs and scope limitations.
• Deadline, permitted start and necessary documents.
• Unresolved questions, contact and next review date.

Statuses. Use distinct statuses such as “closed,” “potentially suitable, requirements unresolved,” “application submitted” and “award received.” This prevents a promising possibility from becoming a supposedly guaranteed budget item. Reading an article, receiving a submission acknowledgement and obtaining an award are separate events.

Approved conditions. For an active project, also retain the applicable approved conditions. A public page may change later; the project record should still show which costs and activities were accepted. Discuss material changes with the responsible institution before shifting the scope.

Size and ownership. Keep this register small enough to maintain. Five options with clear eligibility questions are more useful than fifty programme names without dates or owners. Assign someone to remove exhausted options and recheck those on which a forthcoming commitment depends. The register should support a decision, not become another unmaintained document.

Assume a company plans a document workflow with €20,000 of external project costs. These figures are invented arithmetic assumptions, not a statement that the costs qualify under any particular programme:

Grant. For this example alone, assume a 40% grant against the €15,000 base. The possible grant would be €6,000, not €8,000. External costs remaining after that grant would be €14,000. Internal work, tax treatment and additional expenses may also matter; they are not valued in this example.

Cash flow. If the business pays the supplier first and receives reimbursement later, it may still need €20,000 of cash in the meantime. A lower final contribution does not remove that payment requirement. Add an expected payment date to every invoice and model delayed reimbursement as a separate scenario.

Without funding. Without funding, the external cost remains €20,000. Assess whether the project is still worthwhile, whether scope should change or whether a smaller initial implementation makes more economic sense. Base the decision on measurable process improvement. Funding is a financing benefit, not evidence that an application creates business value.

Less support than expected. The same discipline applies when a project receives less support than expected. Recalculate the remaining contribution using the approved cost base rather than only replacing the headline rate. A reduced eligible base and a delayed payment can affect the project in different ways, even when the advertised maximum percentage remains unchanged.

Many programmes require an application before the project begins. What counts as a start must be established for the particular programme. Resolve that question before a binding order, advance payment or commencement of the supported activity. Do not assume that an unpaid invoice means the application can still be submitted in time.

For internal preparation, we recommend this sequence before commissioning funded work:

1. Describe the process and result.
2. Separate the costs.
3. Identify suitable programmes.
4. Clarify outstanding conditions in writing.
5. Then establish the permitted next step.

A loan requires early involvement of the financing partner. A preliminary research study requires a suitable research partner.

Ask the responsible institution specific questions before the project starts:

• May the quote already be accepted?
• Which preparatory activities are permitted?
• Are recurring licences covered?
• What evidence is required for external services?
• Does changing the supplier need approval?

Put the resulting sequence into the project plan rather than relying on a remembered telephone conversation.

An application deadline should not force a vague technical commitment. If there is insufficient time to describe the deliverables or confirm the permitted start, rushing an order may create a larger problem than missing a particular round. Use the programme's actual procedure and the institution's answer to decide what can responsibly proceed.

Separate lines. Request separate lines for advice, configuration, integrations, training and ongoing operation. Define the expected output for each line: a documented field mapping, a tested exception list or an operational handover, for example. “AI implementation on a time basis” says little about what was delivered.

Work packages. Associate invoices and acceptance records with the same work packages. If a planned calendar integration becomes a CRM integration, make that change visible. It may be technically sensible while still requiring a discussion about the approved grant. Clear allocation also helps prevent the same cost from being improperly included in multiple applications.

Responsibility. Assign responsibility for funding records and for the project's actual operating benefit. One person can perform both roles, but the two outcomes should be assessed separately. A successfully reimbursed project does not prove that its workflow is reliable. A working system does not replace required funding evidence.

Handover. At handover, check that the team can identify the funded deliverables, the assets it owns, the accounts it must maintain and the costs that continue after the supported period. This closes a common gap between a project budget and the practical responsibility for running the result.

Choose a specific process from the AI automation guide. Build the cost model without a grant, then add a separate funding scenario. Compare software, an internal team and an agency, since not every requirement needs custom development.

modernagents can describe the technical activities and integrations needed for that review. This does not imply an award or a claimed consultant accreditation for any particular programme. The funding institution assesses the application, costs and conditions. A clear project scope makes that assessment substantially easier to prepare.