Innovation Studies

All seven FFplus beneficiaries are not allowed to take part in the open call:

• Universität Stuttgart
• scapos AG
• Arctur
•  CINECA
• CESGA
• Teratec
• Cyfronet

Additionally, the main participants from the Open Call-1 innovation studies are only allowed to participate and apply for an extension of their innovation study if their month 7 impact report was evaluated as belonging to the top 70% of innovation studies in the first open call. The main participants of the other 30% of the innovation studies from Call-1 are not allowed to submit a new application in Open Call-2.

The SME definition of the European Commission will be used: in particular, this means that an SME is an independent enterprise with less than 250 employees and a €50 M annual turnover.

Before the conclusion of the funding agreements, the SME will be required to document their status as an SME in line with European Commission definitions and their self-assessment of adherence to EU guidelines for trustworthy AI. 

Yes, assuming that they also qualify as SME using the EC SME definition and are able to explain the expected business impact, whereby it is understood that they would not be in a position to report on an established history of business operation and an extensive customer base.

The FFplus Type-2 open calls are open to proposals that address business challenges from European SMEs and start-ups highly competent in generative AI, professional software development, and data processing but lacking large-scale HPC resources.

A main participant is an SME or a Start-up and supporting participants are organisations assisting the main participant to complete activities foreseen for the innovation study.

The expectation is that the consortium for the new innovations study includes all necessary partners for the execution of the innovation study.

While single-member consortia are eligible in principle, the proposal would need to clearly demonstrate that the sole partner SME or start-up has all necessary competences and skills to carry out the innovation study.

The total number of consortium partners (main participant and supporting participants) is limited to three (3).

Please note, that participation as a main participant in the FFplus innovation studies and in the business experiments (covered by the open call for proposals for Business Experiments addressing the uptake of HPC by SMEs) is mutually exclusive. However, it is possible that an SME is the main participant in a Type 2 innovation study and a supporting participant in Type 1 business experiment(s) – subject to the overall funding limits for Type 1 and Type 2. (See also “Can supporting participants participate in Type-1 business experiments and Type-2 innovation studies?”)

For FFplus Type-2 open calls this is not a problem if the company has a clearly defined role in the consortium and the necessary skills and competences to carry out the proposed work.

The main participant is allowed to take part in only one innovation study per tranche. While an SME or start-up could be the main participant of several proposals, we strongly recommend concentrating on a single proposal. 

However, it is expressly foreseen that the SMEs/start-ups may participate in more than one tranche of innovation studies; that is, if developments and results of their initial innovation study are evaluated successfully, then they would be eligible to submit a proposal for an extension of the developments in a subsequent open call. More information about this is available in the call announcement.

There are no formal restrictions on the number of proposals in which a supporting participant participates. However, the funding restriction for supporting participants -- a maximum aggregated funding of € 150,000 for all innovation studies from this open call (i.e. under call identifier FFplus_Call-2-Type-2) and a maximum aggregated funding of € 300,000 over all innovation studies in which they take part -- must of course be considered.

Supporting participants can participate in both Type-1 experiments and Type-2 innovation studies if their roles are well justified. Of course, they need to adhere to the funding restrictions for both types: 150,000 € for Type-1 business experiments (for all FFplus open calls) and 150,000 € for Type-2 innovation studies in this open call (i.e. under call identifier FFplus_Call-2-Type-2). In addition, it should be noted that for Type-2 innovation studies there is a global limit of 300,000 € for supporting participants over all Type-2 calls. (See also “Are there any rules regarding the structure of the proposal consortium?”)

For proposals in which the need for a consortium of participants is justified, the key criteria for the construction of the consortium is the competence to carry out the tasks of the work plan. Since the international collaborative aspect within FFplus is achieved at the level of the project and all its sub-projects, there is no added value assigned to a proposal with international participants compared with participants from a single country.

For supporting partners only technical/engineering activities are eligible for funding. This means that coordination activities are not eligible for funding for supporting partners.

First, the SME has to check whether the ownership, or part ownership, of the holding company negatively impacts their status as an SME (according to the EC definition). Additionally, they need to check whether they are eligible to receive Digital Europe funding, i.e. whether the company's head office is within an EU member state or in a country associated with the Digital Europe programme (and the country’s association agreement must precede the submission date of this call).

An SME based in a country eligible to receive Digital Europe funding – and having the competence and resources to perform the planned work in that country – is eligible to receive FFplus funding even if fully-owned by a parent company from a country ineligible to receive Digital Europe funding.

Yes, this is not a problem.

The main participant cannot be a public entity, administration, university or research centre because it has to be an SME or start-up. However, public entities (e.g. hospitals) or organisations belonging to public administrations can be supporting participants if their role in the proposed sub-project is well justified. Universities or research centres can also support participants. But note that for supporting participants, only technical/engineering activities are eligible for funding.

No, there should only be one main participant.

No, if the main participant SME is well qualified to perform the work on its own, there is no need for a supporting participant.

If your company participated as a supporting participant in an innovation study from Call-1, this is no problem. However, the funding restriction for supporting participants -- a maximum aggregated funding of € 150,000 for all innovation studies from this open call (i.e. under call identifier FFplus_Call-2-Type-2) and a maximum aggregated funding of € 300,000 over all innovation studies in which they take part -- must of course be considered.

If your company participated as a main participant, the answer to the question depends on the evaluation result of your month 7 impact report: only if the report was evaluated as belonging to the top 70% of innovation studies are you allowed to apply again.

Meetings organised by FFplus, which are mandatory for innovation study partners or coordinators, will in general be held online. Each innovation study should participate in a kick-off meeting for the tranche of innovation studies organised by the FFplus project, which all participants are expected to attend (please note that this will be an online meeting).

Each innovation study will hold a kick-off meeting for the study itself, which all participants are expected to attend. Further innovation study-internal meetings, whose attendance and frequency depend on the work plan, are anticipated. As a guide, two such innovation study-internal meetings are to be expected. If these meetings would lead to substantial travel costs, online meetings should be preferred.

Furthermore, a representative of the innovation study is expected to attend one FFplus review meeting.

The target is for the new innovation studies to commence at the beginning of September 2026. We expect to be able to communicate the results of the evaluation of proposals by July/August 2026. However, the exact dates will depend on the number of proposals received and the resulting workload of the evaluation.

Proposals submitted with a Part B whose length (excluding the cover page, but including the provision of scientific literature references) exceeds the 13-page limit will be rejected without further evaluation.

Please note that during past open calls, there have been some issues connected with the export of “google doc” documents to the final pdfs. This sometimes led to blank pages or page breaks being inserted, which created a final pdf document exceeding the page limit. If you use an online tool for writing the proposal, please be careful to check the page count before submitting.

According to our current planning, the deadline for the next tranche of Open Calls for innovation studies will be (approximately) in September 2027. The call will be published two or three months earlier.

The innovation studies will receive support from the project with a range of actions relating to interactions with the project and also relating to potential collaborations with other sub-projects. Furthermore, direct support for each individual innovation study will be provided relating to gaining access to EuroHPC JU-provided computing resources and technical consultation relating to the effective execution of the innovation study work plan. 

There is a strict page limit for Part B of the proposal (13 pages for Type-2 innovation studies). This obviously also poses a limit on the level of information you can provide. It is important, that the external experts evaluating the proposals can understand who is doing which work and with which amount of effort. Please provide all information that is requested in the template/exemplar for Part B of the proposal.

No specific amount of effort is expected. However, contributions to the creation of the success stories and promotional material are required.

Part A only contains standardised administrative information and cannot include a letter of support. In Part B it is okay (however, in most cases not necessary) to include a letter of support if the proposal still meets the page limit (13 pages including any scientific references for innovation studies).

Yes and the information is to be included in Part A of the proposal.

For supporting participants, only engineering/technical activities are eligible for funding. Activities such as business consultancy, marketing initiatives, administrative tasks, and other non-engineering/non-technical activities are not eligible for funding.

Travel must be well justified in terms of the necessity for performance of the proposed innovation study work plan.

Yes, the work of SME owners not receiving a salary may be declared as personnel costs, if they fulfil the general eligibility conditions and are calculated as unit costs. For more details please check the rules of the Digital Europe Programme],  given in the model grant agreement.

Yes, for all participants if they are well justified and conform with the guidelines specified in the budget modules table included in the call announcement.

The lump sum funding mechanism used in parts of the current EC programmes is not applicable to the funding agreements that will be used for the FFplus sub-projects. The payments to the consortia (which might be single organisation “consortia”) will be made in multiple steps, with a first payment at the start of the sub-project.

There will be a need for reporting to the FFplus coordinator, HLRS (part of the University of Stuttgart), on costs arising.

For more details, check the funding agreement, which will be used for the AI Innovation Studies.

The details will be covered in the funding agreements to be concluded between the University of Stuttgart (HLRS) and the 3rd parties, which is available in the Funding Agreement. 

For the FFplus project, indirect costs or overheads cannot be funded. This is stated explicitly in the call announcement and applies to all participants (main and supporting). The personnel costs need to be calculated in line with the Digital Europe model grant agreement, which includes some salary-related costs for the employer, but not the indirect costs such as office and electricity. The reference document (annotated grant agreement) can be found here. 

The FFplus funding agreements for the sub-projects (templates here) will use cost claims based on actual costs. This is covered on P.47 of the annotated grant agreement linked above.

For supporting participants only technical activities are eligible to be funded (referred to in the call text as “engineering/technical activities”). There is no explicit limitation on the type of staff that could be used, but obviously a technical role would actually require technical staff. For the main participant SME, all staff necessary to perform the work, including technical management, are eligible.

Yes, more than 50% of the funding should be allocated to the main participant. If this is not the case, the proposers need to clearly and convincingly justify the budget distribution to the participants.

No, FFplus is entirely funded under the Digital Europe Programme, managed by the EuroHPC Joint Undertaking. In that way, the funding, including the Financial Support to Third Parties, is centrally managed by the European Commission, not by national authorities. The de minimis aid regulation only applies to aid granted by Member States or through state resources. For that reason, FFplus is not subject to de minimis limits or reporting requirements.

Yes, there are no special access rules for the FFplus innovation studies

For the sub-projects to be submitted to this open call, it is expected that the AI Factories access scheme is the most appropriate choice and provides a range of options (such as timing of submissions, expected response times for access evaluation, compute volumes and periods of use). 

In order to avoid potential delays for the execution phase of innovation studies selected for funding, it is recommended that proposers familiarise themselves with the available access schemes.

There is no guarantee, but experience from Call 1 shows that problems with access to EuroHPC JU machines are few and eventually could be solved. The HPC National Competence Centres may be able to aid proposers with the selection of appropriate resources and access schemes and the related application process.

This is permissible when suitably justified. It should be noted that the provision and cost of charged HPC resources by a supporting partner must be based on actual costs and not commercial rates. These costs have to be included in the experiment’s budget as “other direct cost” of the partner providing the HPC resources to the experiment consortium.

The expectation that the EuroHPC systems be used comes from the JU and relates to a number of aspects: access to those systems is provided free-of-charge; those large systems have been procured for use by the EU R&D&I ecosystem; private, in-house systems might not have the same scale of resources.

However, it is not excluded that other systems be used for the innovation studies - which might be commercial systems or systems offered by national research programmes. The proposals must explain and justify their choices and strategy. Of course, in case of use of commercial systems, the costs would need to be included in the budget for the experiment.

In general, the expectation for the FFplus proposals is that the expected volume of necessary computing resources is clearly explained and linked to the proposed work plan. Additionally, the approach for system access has to be explained: either using EuroHPC resources (free-of-charge) or (linked to the business challenge being addressed) the need for use of charged/ commercial services (which always have to be carefully justified). If commercial/charged services are used, then the corresponding costs must be included in the proposal budget (in Part B of the proposal).

Unfortunately, the FFplus project is not linked to the Access system and policies for the EuroHPC JU platforms, other than the fact that the FFplus project is funded by the JU’s work programme and that we have been given the expectation that the large-scale compute EuroHPC resources should be used. Using the AI Factories access schemes should provide quite quick turn-around times meaning that you should be able to request computing times after receiving the information about the selection for funding from FFplus and not lose (much) time in the innovation study.

We have not specified any bounds on the required compute resources, but there is a key requirement that large-scale resources be used. Our advice would be to compare your needs with the access offerings of the EuroHPC JU systems described here. 

Another possibility might be to think of a multiple-step approach, e.g., starting with access with more restricted resource limits (e.g. Fast Lane or Playground Access to AI Factories) and then later (if granted) using the Large-Scale Access to AI Factories. 

Using this scheme, one could ask – for example – for up to 2.4M GPU node hours on LUMI. Each GPU node has 4 AMD MI250X, which yields 8M GPU hours. However, it is not clear to us that the LUMI hosting site would be able to offer so many resources in the time-frame needed for the innovation study. So, an alternative might be to think about the use of reduced resources.

Your proposal should give and justify the probable range of expected computing time. There will be no penalties if you err, but the requested compute resources should be such that your proposal has a high chance to succeed (i.e. likely be oriented towards the higher end - no budget extension could be granted in case of budget overrun e.g. for commercial compute resources, should you choose to use such).

Yes.

The announcement states: “The innovation studies must use large-scale European HPC resources (e.g., pre-exascale and exascale supercomputers) to develop and customise generative AI models such as foundation and large language models.”. We would then take the view that the development from scratch is not necessary. There is also in the document a statement about which activities are out of scope: “Model exploitation, deployment or operation are all out of scope of activities eligible to be funded by the FFplus project.” So, the emphasis is on creating the generative AI solution that addresses the SMEs business challenge or prospects – of course using large-scale HPC.

Check-List:

Is the proposed work, as described in the proposal, feasible in the technical and management sense? Are risks properly described and addressed?

Does the proposal give grounds to why generative AI serves as a solution to the problem, why the development of a new model is imperative and why the problem could not be solved sooner? Are potential obstacles identified along with approaches to overcome them?

Is the proposed approach explained in terms of the machine learning lifecycle: data preparation, model development/engineering and model evaluation¹. Furthermore:

• Is a detailed description provided and a demonstration of the availability of a suitable training data set?
• Are the characteristics of the models to be developed presented with sufficient detail, including type, size, hyperparameters, and architecture, and are their repercussions to training and exploitation outlined?
• Is the selection of performance metrics for model evaluation, scaling, and optimisation clearly outlined and justified? Are benchmarks to establish baselines described, and methods to ensure experiment reproducibility specified?
• Have potential risks been identified considering EU guidelines for trustworthy AI, including unfairness, bias, hallucinations, and model drift during the exploitation phase, and are means to address and mitigate them presented?

1 Model exploitation/deployment/operation is out of scope.

Check-List:

Is the work plan sufficiently clear and coherent, instilling confidence that the proposed work will be carried out effectively and will be directed towards achieving the objectives of the innovation study and the FFplus call?

Does the proposal include producing a pre-final results and potential impact report to be delivered by the end of the 7th month of the innovation study¹?

Comprehensive Data Management Plan: Is a data management plan presented that covers policies for data access, usage, sharing, retention, and disposal; outlines methods for protecting sensitive or personal data; and incorporates FAIR principles and their implementation when applicable?

Is the consortium as a whole well-qualified to carry out the proposed work? Is each consortium member, as presented in the proposal, qualified to carry out the work they are assigned? Is the assignment of that work clear?

Does the consortium contain the necessary partners with all the skills needed to carry out the proposed work? Are the roles of all partners clearly described, and does each partner have a significant and well-justified role? Are key personnel clearly identified and described? Are the contributions of supporting participants justified and also limited to technical/engineering activities (which are the only activities of supporting participants eligible for funding)? Are any types of partners missing?

Have appropriate resources (effort and budget) been allocated to members of the consortium in such a way that each of them has the required resources needed to carry out their part in the work effectively? Is the effort of each partner required for specific tasks clear? Has the major part of the requested budget (i.e. at least 50%) been allocated to the main participant? If this is not the case, is the budget distribution well justified?

Is the Resource Allocation (effort, budget, software licences or any subcontract) clearly justified? Does the proposal conclusively demonstrate how the allocated resources (personnel, IT/computing and any other resources) address and fill current gaps in the processes needed to implement the proposed action?

Is the resource allocation focused on technology developments, versus expenditures for supporting items, such as data or software purchases?

HPC resources: Does the proposal clearly explain the HPC resources (hardware, software, frameworks, and compute volumes) appropriate for the execution of the innovation study? Are the HPC resources needed defined, possibly using computing resources provided directly (free of charge) by the EuroHPC JU, e.g., through their AI Factories access scheme, or through national actions?

Will the development and customisation of generative AI models described in the proposal be expected to have the appropriate level of performance and parallel scalability to execute the most compute-intensive steps in the innovation study workflow? Does the proposal consider the performance characteristics of said developments and establish that their use is feasible on the proposed computing infrastructure or HPC services? Is that adequately described in the proposal?

[1]It is a requirement of all innovation studies that they deliver an intermediate report on the results achieved (and those expected by the end of the study) and the potential impact of those results on the SME’s business model and potentially by third parties using the results. It will be used to evaluate the eligibility to submit a proposal for extension/continuation of the innovation study to a subsequent open call for proposals.

1 It is a requirement of all innovation studies that they deliver an intermediate report on the results achieved (and those expected by the end of the study) and the potential impact of those results on the SME’s business model and potentially by third parties using the results. It will be used to evaluate the eligibility to submit a proposal for extension/continuation of the innovation study to a subsequent open call for proposals.