FAQ
Mimer AI Factory brings together industry, academia, and high-performance computing infrastructure to support AI development and innovation in Sweden. This FAQ section addresses common questions about our activities, collaboration models, available resources, and how to engage with the initiative. For questions not covered here, you are welcome to contact us directly.
General questions
What is the mission of Mimer AI Factory?
Our mission is to lower the barrier between Swedish organizations and frontier AI by providing the compute, expertise, and ecosystem needed to develop, train, and deploy AI that serves Swedish society and competitiveness.
Is Mimer AI Factory a company?
No, we are a project coordinated by the National Academic Infrastructure for Supercomputing in Sweden (NAISS), through Linköping University, and the Research Institutes of Sweden (RISE).
What is an AI Factory?
AI Factories are dynamic ecosystems that will build around AI-optimised supercomputers, offering computing resources and support services to the European industry, as well as to the European scientific users for the development of large AI models to take advantage of AI technology capabilities in the European Union, and for the development of skills and knowledge in the domain of AI.
How are you funded?
We are fully public funded by the European High Performance Computing Joint Undertaking (EuroHPC-JU) through the Horizon 2020 programme, together with the Swedish national state through Vinnova and the Swedish Research Council (Vetenskapsrådet).
Under which conditions do you operate and for how long will you do this?
Our funding is secured until March 2028, although we expect that we can extend our activities for longer than this.
How should Mimer be acknowledged?
If the project results in a publication or presentation, Mimer AI Factory should be acknowledged using the following sentence:
We acknowledge Mimer AI Factory, funded by Swedish Research Council, Vinnova, and EuroHPC (Grant ID: 101234349), for providing AI expertise for this project.
If Mimer AI Factory staff made a substantial contribution to contents of a publication or presentation that qualifies them as co-authors, according to recommendations for the conduct, reporting, editing, and publication of scholarly work such as the ICMJE in medical journals, they should be made co-authors. Digital copies of publications or presentations should be sent to Mimer AI Factory upon acceptance.
Partner opportunities
Does Mimer AI Factory partner with other organisations?
Yes, get in touch with us through our contact form! We have a very streamlined process for partners, especially if they are European cloud service providers.
Can Mimer AI Factory be part of a consortium for a research project?
No, as we are not a company. However, we can provide letters of support to a research project from our AI Factory.
Compute access
Does Mimer AI Factory have a supercomputer I can use?
Not yet. As of May 2026, our supercomputer is procured and it is expected to be up and running by early 2027. Meanwhile, Swedish companies can use for free the existing supercomputers funded by the EuroHPC-JU.
How do I apply for Compute Access?
Applications for compute, for all types of organizations from across Europe, are managed through a centralized portal. There, you can select the most suitable compute access call and apply through an online form. We offer a support service where we can help you identify the most suitable call, advise on key pieces of information to provide, and review your application form before you submit it, feel free to get in touch with us regarding it. We also offer online guides to help you through the process.
Drop us a line if you managed to gain compute access thanks to our materials. It means a lot to us!
Playground, Fast Lane, AI for Science... What are these?
When you apply for European computing resources, you need to choose one of the “calls” that has you as the appropriate target.
- Playground Call: This has 5 000 GPU hours, intended primarily for SMEs/startups that are in a lower maturity level of their product. The duration of the resources is up to 3 months.
- Fast Lane: This is intended for companies that has a higher level of maturity in their products and are ready to scale, offering up to 50 000 GPU hours. The duration of the resources is up to 3 months.
- AI for Science: This call is aimed to public sector and academia, or for industrial applications that have projects funded by the European Union. It offers up to 90 000 GPU hours.
- Large Scale: For companies that are in need of over 50 000 GPU hours. This call is also ranked with others that apply on the same deadline, unlike Fast Lane and Playground Access calls which are evaluated only on technical eligibility. The duration of the resources is up to 1 year.
Can allocations on your supercomputers be extended?
No, once the lifetime of an allocation is exhausted, you have to apply for access again. There is a grace period under which your data will be kept on the supercomputer, so that you can move it to a new project.
What supercomputers are offered and where are they located?
A wide range of supercomputers is available, with different hardware, capabilities and locations across Europe, depending on the call you choose. Several supercomputers are in various stages of deployment at the moment, so expect frequent additions to the list. Not all systems are available in all access calls though. We can support you in identifying the best combination of access call and supercomputer based on your needs.
Is this intended only for SMEs? Can large companies, academia and public sector also apply?
Yes, large companies can use the resources for free regardless of the call but they need to do open science, which means to publish some of the outcomes obtained from using the supercomputers. SMEs and startups are exempt of this requirement.
What are these supercomputers for?
The systems are meant for product development and innovation; therefore, after you are done developing your product, you need to look for a cloud service provider to deploy it.
Do I have to pay something?
No, we are publicly funded and therefore everything offered is at no cost.
If I use these supercomputers, will all the data be exclusively mine?
Yes, neither the EuroHPC, Mimer AI Factory or the supercomputing centre takes any part or royalty of your intellectual property.
How do you protect the confidentiality of the data?
The systems are often hosted by public institutions, such as universities or research centres, and each individual system has its own measures to avoid data breach and relies on, for example, folders permissions where only the team are able to see the contents of the folder. It is up to the users to assess whether these measures are enough to protect their data.
Services
What kind of services do you offer and what is the cost?
Check our website to see all our services. They range from giving support to access supercomputers to data, consulting in our focus areas and even trustworthy AI!
Can you help us to develop our AI solutions?
Yes, in several cases we can help you to develop your AI solution for up to 1 month, whether you are in academia, in public sector or industry. We also expect you to put similar technical effort as us, and to have domain-specific knowledge, during the execution of the project.
Drop us a line if you managed to gain compute access thanks to our materials. It means a lot to us!
We are a consulting company developing AI Solutions for other companies. Can we use your services for a customer project?
You can request access to supercomputers on behalf of your customers. For these situations, support is given in a case-by-case basis.
Can I use AI to facilitate new material design or material synthesis optimization? If so, how should I prepare myself to be ready of adopting AI?
Yes! AI can meaningfully accelerate materials research by predicting properties (bandgap, formation energy, phase stability), suggesting new compositions worth investigating, optimizing synthesis parameters, and analyzing characterization images such as SEM or TEM automatically. However, AI works best as a complement to domain expertise, not a replacement. It is fundamentally a pattern recognition tool that learns from your existing data, which means its usefulness depends heavily on data quality and how well you define your problem. The key conceptual step is framing everything as an input-output relationship: your input X is what you can already measure or know, such as composition, crystal structure descriptors, synthesis temperature, precursor concentrations, or raw images, and your output Y is what you want to predict or optimize, such as a bandgap value, yield percentage, or phase purity. Therefore, the key is to collect meaningful data that provide X and Y.
To prepare yourself practically, data preparation deserves the most attention: a clean, consistently labeled dataset of even 200–500 samples with well-defined X and Y will outperform a fancy deep learning model trained on messy data. From there, you can experiment by starting from building the fundamentals first, validate everything honestly against held-out data, and then let your physical intuition guide which problems are actually worth solving with AI.
AI technology solutions
Do you offer pre-trained models for different use cases?
We plan to offer this in the future. Right now, we are looking into building tools and models that can help companies, so stay tuned!
Can you help getting started with open source LLMs?
Yes, this is a standard process that many companies reach out us for, including migrating from proprietary solutions to open source ones.
I want to tune an LLM for my use cases, can you help?
Yes, we help companies to perform fine-tuning or prompt engineering to their use case.
Do you offer advice on RAG solutions?
Yes, we help companies to develop and implement their own RAG solutions.
Do you offer no-code solutions/agents?
Not at the moment.
Do you offer expertise in AI for specific fields?
Yes! Several of our team members bring deep domain knowledge to AI applications across five focus areas: Materials Science, Autonomous Systems, Gaming, Life Sciences, and Trustworthy AI. They are ready to support any company or public sector organization that needs field-specific guidance: from research to real-world implementation.
That said, these focus areas don’t define the boundaries of what we can help with. If your needs fall outside them, reach out and let’s explore how MimerAI Factory can support you!
Training
Are training events open to industry participants?
Yes! Our training events are open to industry, public sector, and academia equally. Since they are publicly funded, there is no cost to attend.
Can we collaborate, host, or co-organize training where Mimer AI Factory contributes?
Yes, in several cases we can help you to develop your AI solution for up to 1 month, whether you are in academia, in public sector or industry. We also expect you to put similar technical effort as us, and to have domain-specific knowledge, during the execution of the project.
Drop us a line if you managed to gain compute access thanks to our materials. It means a lot to us!
Can I use my private email address when registering for a training event?
No. Registration requires an institutional or organizational email address (e.g., your company or university email). Personal email providers like Gmail, Outlook, or Yahoo are not accepted. Additionally, participation is limited to residents of Horizon Europe 2020 countries (see the official list). Other restrictions may also apply.
Our supercomputer
What are the specs of your planned supercomputer?
Our supercomputer will be based on the GB200 NVL4 system by Nvidia, where each node will be equipped with four Nvidia B200 GPUs (180GB HBM3e each) and two 72-core Nvidia Grace CPUs, with 960GB of LPDDR5 main system memory and Nvidia Nvlink high-speed interconnect across GPUs and CPUs. There will also be a small CPU-only partition for data processing workloads (using Nvidia Grace CPUs) and a partition with sovereign EU-designed AI accelerators.
Our contact form
Why do I need to use the contact form?
We receive many inquiries, and we have a streamlined process for getting in touch with clients. Talk with us and we will soon reach out to you!