The $5B Gamble: Can Nebius Solve AI Infrastructure's Biggest Bottleneck?

robot
Abstract generation in progress

Nebius (NBIS) is in a race against time—and its own success. Third-quarter demand was so strong that the company sold out all available GPU capacity instantly. Every new batch of compute power gets absorbed the moment it comes online. Now here's the problem: they can't scale fast enough.

The Crunch

Capacity constraints are literally leaving money on the table. To fix this, NBIS is betting big:

  • Power target by 2026: 2.5 GW (up from 1 GW announced just months ago)
  • Fully connected power by end-2026: 800 MW to 1 GW
  • CapEx hike: Bumped from $2B to $5B for 2025 alone

The math is simple: more power = more GPUs = more revenue. They're on track to hit $900M–$1.1B ARR by end of 2025, scaling to $7–$9B by end of 2026. But here's the catch—they revised 2025 revenue guidance down to $500–$550M (from $450–$630M), mainly due to the timing lag between infrastructure buildout and revenue generation.

The Competition is Heating Up

NBIS isn't alone in this arms race. Microsoft plans to boost AI capacity by over 80% in 2025 and double its data center footprint in two years. Azure revenue is expected to grow ~37%, though even with aggressive expansion, MSFT admits it'll be capacity-constrained through fiscal year-end.

CoreWeave is also scaling aggressively but just cut its 2025 revenue guidance to $5.05–$5.15B (from $5.15–$5.35B) due to delayed power cabinet deliveries from partners—a painful reminder that execution risks are real.

The Valuation Angle

NBIS stock has surged 144.2% in six months vs. 6.9% for the broader Internet Software industry. Yet it's trading at just 4.66X price-to-book vs. 39.95X for the sector. Earnings estimates have seen downward revisions over the past 60 days, and Zacks currently rates it a Sell (#4 Rank).

The Real Question

Can NBIS execute this $5B expansion plan while managing supply chain chaos, competing with tech giants, and delivering promised ARR targets? Capacity shortages are becoming the new normal in AI infrastructure—but whoever solves it first wins big.

This page may contain third-party content, which is provided for information purposes only (not representations/warranties) and should not be considered as an endorsement of its views by Gate, nor as financial or professional advice. See Disclaimer for details.
  • Reward
  • Comment
  • Repost
  • Share
Comment
Add a comment
Add a comment
No comments
  • Pinned