AI landscape update for March 2025

Continental investments, European and Finnish events

The US announced a new joint 500-billion private joint venture by OpenAI, Softbank and Oracle to build new AI datacenters. The venture is completely focused on hardware, and appears to include an earlier $100 billion investment by OpenAI/Microsoft.

The EU launched InvestAI, a €200 billion public-private fund to fund AI development. It includes new data centers that are now called “AI gigafactories”. The facilities can be used to train new models for medicine and science. Much of the funding goes to projects and companies as investments, so unlike the US, in Europe the approach is comprehensive instead of focusing on hardware only. Finland’s Kajaani Lumi by CSC is one of the AI Factories. It is a pleasure to see so many European countries joining us with the Lumi AIF.

Source: European Commission CC BY 4.0

There are also other AI datacenter projects in flight, for example this one for Nebius DC in Mäntsälä. Nebius is an AI cloud provider.

Mistral AI, a French startup, launched their ChatGPT competitor Le Chat smartphone app onto mobile app stores. Mistral AI also provides many models, with a summary available at https://builtin.com/articles/mistral-ai. It is also possible to run Mistral’s models for your services in AWS on top of AWS Bedrock platform.

Nitor is a Large Finnish Actor in AI, says a fresh report by AI Finland Business Finland.

OpenLLM Europe announced an aim to build the first multimodal multilingual european model – this should support smaller languages well. Their LLM projects in all European languages are listed in GitHub.

All in all, Europe is definitely a key player in the games rather than just “regulating”.

Artificial Intelligence Action Summit 2025 Paris

France hosted the Artificial Intelligence Action Summit, an invite-only event gathering companies, academia, organizations, and heads of states together. Our ecosystem company Renessai provides strategic AI consultancy, and Renessai’s Eija-Leena "Eikku" Koponen was invited to the event by the French Ministry for Foreign Affairs. 

Here are Eikku’s key takeaways of the event:

• With resource requirements on the downturn, the focus is now shifting to working smarter rather than harder with emphasis on efficiency and purposeful implementation

• Europe needs to transform from pushing AI to AI having pull

• Europeans should be a little more capable of taking risks

• While GenAI still consumes the media spotlight, the greatest wins may lie within the realm of classic machine learning, with robust and lower-risk solutions with better explainability – exactly what businesses need right now 

• Augment people instead of replacing them

• Transparency will only be reached when people understand AI concepts at least in general

• Europe needs to choose if it wants to compete or to collaborate

• Distribution of work can also be gains in quality

• As a person also working on technical solutions, it was funny to remind others that the problems in adopting and/or implementing AI generally don’t lie in the tech capabilities but on the softer side: skills, organisations and strategy 😇

• After this Summit I’m confident that we’ll find a European way – chaotic and diverse and with a few tidbits lost in translation for sure! 

The first chapters of the EU AI Act came into effect

The EU AI Act came into effect in August 2024, and the first obligations under the Act became applicable on February 2nd, 2025: Prohibitions on certain AI practices and requirements on AI literacy are now being applied.

Companies are required to educate their employees on AI literacy so that they can understand, critically assess, and effectively interact with AI technologies.

Agents

Agents are coming. Don’t make the mistake of equating large language models (LLMs) or the generative AI domain with the entirety of the AI ecosystem. While we can use models in the services we create with our clients (and that’s what they expect from us), the biggest change will be that in a few years it will be both humans and machines using the services we create. The machines will arise as “AI Agents;” programs that (semi-)independently complete a given task without supervision and without detailed instructions on how to accomplish said task.

The agents will run in the cloud as services. OpenAI launched their first agent, Operator, a general-purpose agent – basically a ChatGPT that can operate a web browser while you wait. Should you buy a $200/month Pro to use the Operator? No, it’s mid. But they gotta start somewhere.

The agents will run anywhere as applications. Check out Upsonic: ”GPT Computer Assistant Framework is an open-source framework for building AI agents that interact with computers like humans. GCA empowers developers to create intelligent automation using multiple LLMs, screen interaction, and voice capabilities.”

The agents will run on your phone. Apple’s vision is that they will not, though. The agents (and model training) will run in your home. That’s the vision of the chip manufacturer NVidia. The agents will run in your GitHub Copilot.

If you want to read more about AI agents, check out this collection of resources:

• AWS introduction on what AI agents are

• IBM introduction on what AI agents are

• Microsoft AI agents for beginners

• LangChain survey of 1300 professionals on how AI agents are being used (or not) today

This landscape update was collected via material shared by Nitoreans on our internal channels. Do let us know if this was helpful for you. We’re also more than happy to discuss your AI needs, AI strategy or what kind of experts could help you to succeed with sustainable digital development, so get in touch!