Nebius buys Inferize, a startup founded in January that cuts the wait while AI models load

Nebius, the Amsterdam-based GPU cloud, announced on 1 October that it has acquired Inferize, which it describes as “an inference optimization company whose technology shortens the time needed to launch and scale large AI models.”

The problem Inferize works on is the cold start: the delay while a large model loads onto GPUs before it can answer requests. If that delay is short, a cloud can add and remove capacity as demand changes instead of keeping GPUs loaded and idle. Nebius says Inferize's technology and team have joined Nebius Token Factory, its platform for serving models in production.

Nebius says Inferize was founded in January 2026 and had a working prototype within three months. “The terms of the transaction were not disclosed.” Calcalist reported a price range, but Nebius gives none, so no figure is repeated here.

Danila Shtan, Nebius's chief technology officer, said: “Running inference well takes more than fast GPUs and optimized models. The whole system needs to respond when demand changes, including how quickly additional capacity is ready to serve customers.”

This is a separate event from Nebius's 50 MW data-centre lease with AIB Data Centers, which the Switchboard ran on 1 October. Together they show the GPU clouds spending on both sides of inference: more capacity, and software that wastes less of it.

• Confirmed Nebius acquired Inferize; its technology and team joined Nebius Token Factory; terms were not disclosed. Nebius (EQS release)
• Claimed Inferize was founded in January 2026 and had a working prototype within three months. Nebius (EQS release)