Nebius Group (NASDAQ: NBIS) shares climbed in premarket trading Thursday after the company announced the acquisition of inference optimization firm Inferize to improve GPU utilization across AI workloads.
The deal brings Inferize’s technology and engineering team into Nebius Token Factory, the company’s managed production-inference platform designed for large-scale AI deployment.
Inferize’s core technology targets cold starts, the delays that occur when models load, new capacity comes online, or model weights are updated, leaving expensive GPUs sitting idle.
Those idle periods represent a significant infrastructure cost at scale, as cloud providers must hold spare GPU capacity in reserve to absorb sudden demand spikes even when that capacity is not actively processing requests.
Inferize’s approach is designed to make inference capacity more elastic, allowing Nebius to bring additional GPU resources online faster and align supply more closely with real-time demand.
The result, according to Nebius, could be meaningfully higher GPU utilization and lower infrastructure costs per AI token served across its platform.
NBIS shares gained as much as 1.9% in premarket trade following the announcement, though retail sentiment on Stocktwits shifted to “bearish” from “neutral” over the same period for the Nvidia (NASDAQ: NVDA)-backed company.
Inferize CEO Guy Bortnikov said the company’s technology is specifically designed to reduce the need to keep spare GPUs running purely to remain ready for sudden demand surges.
Inferize was founded in January 2026 and built a working prototype within just three months, with its engineers now set to join Nebius’s dedicated inference team.
The acquisition adds to a broader buildout of Token Factory, which previously incorporated model, kernel, and system-level optimization technology through the addition of teams from Eigen AI and Clarifai.
Nebius said Inferize’s work will extend across the full Token Factory platform rather than being limited to a single integration point, reflecting the company’s ambition to build a comprehensive production-grade inference stack.
Financial terms of the deal were not disclosed, and Nebius has not provided guidance on the expected timeline for Inferize’s technology to be fully integrated into its platform.
