AI cloud service provider NEBIUS (NBIS.US) is introducing a new product designed to offer access to open-source models and the computing power required to run them—part of its strategy to expand its market share in the AI sector. The product, named "Token Factory," focuses on inference workloads, the process of executing trained AI models and applications. It allows customers to choose from leading open-source models, including OpenAI’s GPT-oss, Meta’s (META.US) Llama, and DeepSeek, while providing secure computing resources to deploy their applications.

Token Factory competes with cloud offerings from Amazon (AMZN.US) and Microsoft (MSFT.US), the latter of which recently signed a $19.4 billion deal with NEBIUS for AI computing capacity. Startups like Fireworks and Baseten also provide similar services.

NEBIUS, spun off from Russian internet company Yandex last year and now based in the Netherlands, has emerged as a notable "new cloud services" provider. The company sells AI cloud resources from data centers in the U.S., Europe, and Israel, where it operates one of the country’s first publicly available clusters of NVIDIA’s (NVDA.US) latest-generation AI chips.

For AI infrastructure providers like NEBIUS, selling software services on top of cloud offerings can yield higher profits. However, Roman Chernin, NEBIUS’s co-founder and Chief Business Officer, emphasized that the company prioritizes attracting more customers through a diversified product lineup over immediate profit gains.

"Owning infrastructure alone isn’t enough. We aim to be a major player, not just a utility provider," Chernin said in an interview. He sees an opportunity in serving the rapidly evolving AI market, where some developers are reconsidering reliance on closed-source or proprietary models from top AI labs.

Building on closed systems may limit customization flexibility and increase costs, Chernin noted. "Developers are shifting from monolithic, closed ecosystems to more diversified solutions. We’re creating a scalable, reliable platform that lets customers seamlessly transition from initial setups to full-scale deployments."

Early adopters of Token Factory include Amsterdam-based tech firm Prosus and AI video platform Higgsfield. Hugging Face is also utilizing NEBIUS’s infrastructure for inference and collaborating to feature Token Factory in its inference service marketplace.