Command Palette
Search for a command to run...

Nebius Pushes Token Factory as Q1 Revenue Jumps 684%, Guidance Rises

businesscompany-earningsaiai-infrastructureai-inference-platformsai-compute-chips 3 posts · 2 accounts

Nebius is presenting Token Factory as the next leg of its AI infrastructure business, with co-founder Roman Chernin saying on the company’s first-quarter 2026 earnings call that the next bottleneck is running inference efficiently at scale rather than simply supplying GPUs. He described Token Factory as a production inference platform for fine-tuning, optimization, orchestration and deployment of open-source and custom models, positioning Nebius as more than a GPU rental provider.

The strategy comes as Nebius reported first-quarter revenue of $399 million, up 684% from a year earlier, while AI-specific revenue rose 841% to $390 million and accounted for 98% of total sales. Annual recurring revenue grew 674%, and the company raised full-year ARR guidance to $7 billion to $9 billion and revenue guidance to $3.0 billion to $3.4 billion. Contracted power capacity now exceeds 3.5 gigawatts, with a 4-gigawatt year-end target, and Nebius said Q1 acquisitions of Tavily, Eigen AI and Clarifai add search, model optimization and multimodal inference capabilities to the Token Factory stack.

From the sources (3 posts)

@stocksavvyshay

$NBIS co-founder Roman Chernin explained why the next AI bottleneck is running inference efficiently at scale. That’s why Nebius is building Token Factory around model optimization, orchestration & agentic AI deployment instead of acti

@milkroadai

RT @MilkRoadAI: Nebius will be a TRILLION dollar company and here is exactly why (Save this). Brad Gerstner's Altimeter just said on camer…

@milkroadai

Nebius will be a TRILLION dollar company. Most people look at Nebius and see a GPU rental business. Roman Chernin just explained on the Q1 2026 earnings call exactly why that framing is wrong and why it misses the most important thing hap

Preview built on a synthetic news corpus (16 weeks, Apr–Jul 2026). Impact calls are model reads, not price data.

About Archive