Domyn plans to launch open-source frontier AI model within a year

Europe has spent the better part of the AI boom watching from the sidelines while American and Chinese companies built the models that now power everything from customer service bots to drug discovery. Italy’s Domyn wants to change that, and it’s putting a timeline on the ambition: one year.

The Milan-based startup announced plans to release a fully open-source frontier AI model anticipated to exceed 400 billion parameters. The model will be trained from scratch and designed to run on local servers, a detail that matters enormously for the regulated industries Domyn is targeting.

From iGenius to billion-dollar contender

Domyn wasn’t always called Domyn. Founded in 2016 in Milan, the company previously operated under the name iGenius before rebranding. Under CEO Uljan Sharka, the company has grown its valuation to over $1 billion, backed in part by G42, the Abu Dhabi-based AI and cloud computing firm.

The company currently offers two models: Domyn Small, which has 10 billion parameters, and the larger Domyn model. It also has a collaboration with NVIDIA for supercomputing capabilities. But the planned frontier model represents a massive leap in scale, moving from the tens of billions of parameters to over 400 billion.

The company’s ambitions extend well beyond a single model release. Domyn has publicly discussed plans to raise over 1 billion euros for AI initiatives and invest up to $10 billion over three years in AI infrastructure.

The EU Grand Challenge and Europe’s sovereignty play

The company leads the Europa consortium, which won the European Commission’s Frontier AI Grand Challenge. That challenge is specifically designed to produce a sovereign AI system capable of functioning in all 24 official EU languages.

Domyn’s focus on regulated sectors like finance and government isn’t accidental. These are industries where data residency, model transparency, and infrastructure ownership aren’t nice-to-haves. They’re legal requirements. The broader European push for AI sovereignty also includes initiatives like EuroHPC, which aims to provide public access to supercomputing resources across the continent.

What this means for investors and the AI landscape

The $10 billion investment target over three years, if realized, would represent one of the largest AI infrastructure commitments in European history. Training a 400-billion-parameter model from scratch requires not just capital but also access to enormous quantities of compute, high-quality training data, and top-tier research talent.

The backing from G42 provides both capital and potential access to computing resources in the Middle East, where data center capacity is expanding rapidly. The EU Grand Challenge win provides political legitimacy and likely opens doors to government contracts across the bloc. And the open-source commitment could attract a developer community that accelerates the model’s improvement after launch.

As AI governance frameworks tighten globally, and particularly within the EU under the AI Act, companies that build for compliance from day one may find themselves with a significant competitive moat.

Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.