Token Factory
Production inference workloads more than tripled in Q2 2026, increasingly on open-weight models.
The numbers
What it is
Managed inference service that serves open-weight models such as Kimi K3, GLM-5.2 and NVIDIA Nemotron Ultra, and bills per token. Eigen AI and Clarifai, acquired in 2026, add inference optimization.
Why it matters to the stock. Usage-based inference revenue is a second growth path next to capacity rental. It also raises utilization of the GPU fleet.
News
Nebius acquires Inferize to strengthen Token Factory inference stack
Nebius said on October 1, 2026 that it had acquired Inferize, a young inference-optimization company founded in January 2026.
Palantir names Nebius its preferred sovereign AI infrastructure partner
Palantir and Nebius announced a partnership under which Nebius compute and inference endpoints will sit inside Palantir's enterprise perimeter, letting eligible Palantir customers use Nebius cloud and inference while keeping control of their data and models.
Nebius completes Eigen AI acquisition after regulatory approvals
Nebius completed its acquisition of Eigen AI on June 10, 2026, after the required regulatory approvals and customary closing conditions were met.
Nebius hires Clarifai's core team and licenses its inference technology
Nebius hired the core engineering and research team of Clarifai, led by founder Matthew Zeiler, who becomes SVP of Research.
Nebius agrees to buy inference optimizer Eigen AI for about $643M
Nebius agreed to acquire Eigen AI, an inference and model-optimization firm, for about $643 million, subject to adjustments.