Q3 FY2025 Earnings Recap

Nebius Group N.V. (NBIS)

Reported Nov 11th 2025

News Summary

  • Nebius reported Q3 2025 revenue of $146M, up 355% YoY and 39% sequentially, beating consensus of $105M, with its core AI infrastructure segment growing 400% YoY at nearly 19% adjusted EBITDA margin — the stock rose roughly 4% on the day, and analyst sentiment from Goldman and D.A. Davidson was described as probing but broadly constructive on the growth trajectory.
  • Demand is running so far ahead of supply that CEO Arkady Volozh stated the $3B Meta deal was sized to available capacity rather than what Meta wanted, and CRO Marc Boroditsky described 'learning to say no to customers' after Q3 pipeline generation hit $4B in a single quarter — a direct read on how intense unmet GPU cloud demand is in the current market.
  • Nebius raised its 2026 ARR target to $7–9B with more than half already under contract, tripled its contracted power target from 1GW to 2.5GW, and raised full-year 2025 CapEx guidance from ~$2B to ~$5B, signaling aggressive infrastructure scaling to absorb demand it currently cannot serve.
  • Nebius is deploying NVIDIA Blackwell B200s in Israel, B300s in the UK, and will be first in Europe to deploy GB300s in Finland in December — all presold before launch — illustrating how rapidly next-generation NVIDIA silicon is being absorbed by hyperscaler-adjacent cloud providers.
  • The new Nebius Token Factory inference platform targets enterprise and vertical AI builders deploying open-source models including Llama, Qwen, DeepSeek, Mistral, and Nemotron at scale with guaranteed SLAs and transparent cost-per-token pricing, marking inference-as-a-service as a distinct and accelerating compute demand category.

Financial Highlights

MetricQ1 '24Q2 '24Q3 '24Q4 '24Q1 '25Q2 '25Q3 '25Q4 '25
Revenue$0.0B$0.0B$0.0B$0.0B$0.1B$0.1B$0.1B$0.2B
Rev Growth (YoY)120.5%157.1%766.0%465.7%389.5%769.7%237.4%▲ 500.8%
Gross Margin22.5%72.6%56.4%27.4%46.7%71.4%70.6%▼ 69.9%
Operating Margin-715.6%-792.8%-200.9%-393.9%-234.2%-105.8%-89.1%▼ -103.0%
EV/Sales626.6x295.0x80.4x109.2x68.1x122.0x192.0x94.0x

Earnings Call Highlights

Nebius delivered a blowout quarter defined by one dominant theme: demand is so far ahead of supply that the company is literally turning away customers and sizing mega-deals to whatever capacity it has available rather than what buyers want. Management's posture was confident to the point of aggression — announcing a $3B Meta deal on top of the prior $17-19B Microsoft contract, tripling contracted capacity guidance from 1GW to 2.5GW, and raising CapEx guidance from ~$2B to ~$5B in a single quarter. Analysts from Goldman and D.A. Davidson probed the sustainability and financing of this growth, and management's answers were disciplined but unapologetically bullish, with more than half of the $7-9B 2026 ARR target already booked. The call's tone was one of controlled urgency — a company racing to build infrastructure fast enough to absorb demand it cannot currently serve, while insisting margins and unit economics remain a governing constraint.

  • Nebius signed a ~$3B, five-year deal with Meta, and CEO Arkady Volozh stated explicitly that 'the size of the contract was limited to the amount of capacity that we had available' — meaning the deal was supply-constrained, not demand-constrained, a striking signal of the current GPU cloud pricing environment.
  • The company raised its 2026 contracted power target from 1GW to 2.5GW and set a physical build target of 800MW–1GW of connected capacity by end of 2026, with a corresponding ARR target of $7–9B, of which management says more than half is already under contract.
  • Chief Revenue Officer Marc Boroditsky disclosed that Q3 pipeline generation grew 70% quarter-on-quarter to $4B in a single quarter, but conversion was severely limited by capacity — he described 'learning to say no to customers' as a new professional skill, underscoring the intensity of unmet demand.
  • Nebius is deploying Blackwell B200s in Israel, B300s in the UK, and will be the first in Europe to deploy GB300s in Finland in December — all presold before launch — giving Cerebras a concrete read on how aggressively hyperscaler-adjacent clouds are absorbing next-gen NVIDIA silicon.
  • The new Nebius Token Factory inference platform targets the emerging wave of enterprise and vertical AI builders deploying open-source models (Llama, Qwen, DeepSeek, Mistral, Nemotron) at scale with guaranteed SLAs and transparent cost-per-token, signaling that inference-as-a-service is becoming a distinct and growing compute demand category.