Nebius Group NV, a rapidly rising player in the artificial intelligence infrastructure space, has unveiled a new platform aimed at democratizing access to powerful open-source AI models. The product – called Token Factory – offers developers the tools and computing power to run leading AI models efficiently, marking Nebius’ bold step into territory dominated by cloud giants like Microsoft Azure and Amazon Web Services (AWS).

A New AI Platform for the Open-Source Era

Token Factory focuses on inference workloads – the process of deploying and running AI models and applications after they’ve been trained. The platform allows clients to select from an expanding library of open models, including OpenAI’s GPT-oss, Meta’s Llama, and DeepSeek from China. Users can then deploy these models on Nebius’ globally distributed cloud infrastructure.

By combining flexibility and raw computing performance, Nebius aims to meet growing demand from companies eager to integrate AI but reluctant to depend solely on proprietary ecosystems.

Competing With Tech Titans

The move puts Nebius in direct competition with major cloud providers. Amazon and Microsoft both offer similar AI model-hosting and inference tools – but Nebius’ recent momentum gives it a unique edge.

Earlier this year, Microsoft struck a deal worth up to $19.4 billion to secure AI computing capacity from Nebius, underscoring the scale of the Dutch company’s ambitions. Nebius’ entry into open-model services could therefore serve as both a complement and a competitive challenge to its larger partners.

Other emerging AI infrastructure startups, such as Fireworks and Baseten, are also chasing this market, offering model-serving platforms for developers building AI applications.

Expanding Beyond Infrastructure

Founded in the Netherlands after its split from Russian tech company Yandex, Nebius has swiftly positioned itself as a major “neocloud” provider – a new breed of cloud companies focused on AI compute rather than general web hosting. The company operates data centers across the U.S., Europe, and Israel, where it recently opened one of the region’s first publicly accessible clusters equipped with Nvidia’s latest AI chips.

While many infrastructure companies are moving into higher-margin software services, Roman Chernin, Nebius’ co-founder and Chief Business Officer, insists the company’s motivation is customer-focused rather than purely financial.

“You have to be much more than just infrastructure,” Chernin said in an interview. “We want to be a large company, but we don’t want to be only a utility company.”

Empowering Developers With Flexibility

Chernin believes the market is shifting. Many AI developers are starting to question the trade-offs of depending exclusively on closed-source systems from the world’s leading AI labs. Proprietary models can offer state-of-the-art performance but limit flexibility for customization and fine-tuning – and come with steep usage costs.

“They start shifting from building everything on a closed ecosystem to a more diversified portfolio of models,” Chernin explained. “What we built is a scalable, reliable platform that lets clients seamlessly switch from whatever they started with to what they need at scale.”

Early Adopters and Partnerships

Nebius has already secured several prominent early customers. Dutch technology investment group Prosus NV and AI video company Higgsfield are using Token Factory for large-scale inference operations. Meanwhile, Hugging Face, the popular open AI community platform, is partnering with Nebius – both using its infrastructure and featuring Token Factory on its inference marketplace.

A New Challenger in AI Infrastructure

The launch of Token Factory cements Nebius’ position as one of the most ambitious new entrants in the global AI cloud race. As demand for scalable, affordable, and flexible AI computing grows, the company’s strategy to embrace open models – rather than depend on closed systems – could resonate with developers seeking independence from the major cloud monopolies.

Nebius’ message is clear: the future of AI infrastructure lies not just in raw computing power, but in the freedom to build, deploy, and innovate without boundaries.