Nebius snaps up Clarifai’s compute orchestration tech and talent to enhance AI inference

Dutch artificial intelligence infrastructure giant Nebius Group N.V. said today it’s recruiting the core engineering team from AI orchestration software firm Clarifai Inc. in an effort to boost its managed inference services.

As part of the deal, Nebius is also snapping up Clarifai’s portfolio of patents and licensing its inference and compute orchestration technology, the companies said.

Nebius did not disclose how much it’s paying for the “acqui-hire,” but the move allows it to get its hands on some significant talent that will bolster its own research teams. That talent includes Clarifai founder and Chief Executive Matthew Zeiler, a pioneering researcher in the machine learning world, who will join Nebius as its new senior vice president of research. It also includes a select group of veteran researchers and engineers who’ve collectively spent decades building production-grade AI infrastructure systems.

Clarifai sold a full-stack AI development platform that’s used to create intelligent applications that leverage both structured and unstructured data as a source of knowledge. With its platform, developers can train the AI models that power their applications using very specific datasets, thanks to features including its vast data lake, an automated data labeling tool and a search tool for indexing that information.

Clarifai also offered a compute orchestration service that enables companies to organize all of their AI resources – including their own on-premises servers and those rented from public clouds – within a single, centralized portal that makes everything easier to manage and optimize based on performance and cost requirements.

Those compute orchestration capabilities pair nicely with Nebius, which has emerged as one of the leading so-called “neoclouds,” or pure-play AI cloud infrastructure providers, alongside the likes of CoreWeave Inc. While Amazon Web Services Inc., Microsoft Corp. and Google Cloud offer general-purpose infrastructure, neoclouds like Nebius are laser-focused on the specialized requirements of large language models.

Clarifai’s technology will be used to enhance Nebius’s new Token Factory inference service, which provides dedicated infrastructure for running trained models in production. Running models reliably and at scale is difficult and expensive, but Token Factory helps to simplify this with a vertically integrated stack featuring hardware and software that’s optimized for the token economy.

Nebius has been racing to build out its Token Factory. Last month it spent $643 million to acquire the inference software startup Eigen AI Inc., and Clarifai’s tech is another piece of the puzzle. While Eigen AI provides model-level optimization, essentially making the AI models more streamlined, Clarifai’s team brings the system-level expertise for inference.

By integrating Clarifai’s technology into the Token Factory, Nebius says, it will be able to offer a full-stack inference platform that enables superior token efficiency, reducing the cost per word or image generated, while supporting more advanced features like multimodal and agentic reasoning and long-term context memory. AI agents running on Nebius’ infrastructure will be able to remember previous interaction and process both text and visual data simultaneously without increasing latency, the company said.

Nebius co-founder and Chief Business Officer Roman Chernin stressed that in order to deliver efficient inference at scale, the company has to make sure that model optimization, system design and compute orchestration all work in unison. “The integration of Clarifai’s advanced system-building capabilities and proven team will further strengthen Nebius Token Factory, offering customers the infrastructure they need to run models reliably and cost-effectively in production,” he said.

Photo: Nebius