Neocloud Brief · neocloud stocks · what is a neocloud · neocloud ETF

NBIS — Nebius: the full-stack neocloud priced for perfection

The Yandex spinoff building GPU cloud plus the software on top — +454% revenue growth, ~40% EBITDA margins, a $17.4B Microsoft deal, and a funding need big enough to scare even its fans.

Ticker NBIS · Nasdaq Price $222.13 Mkt cap ~$57.9B 52-wk $73.52–299.86 Snapshot Sep 18, 2026

01The 30-second read

Fast read

Nebius is the closest thing the neocloud group has to a proven full-stack cloud: Q2 2026 revenue of $582.3M, up 454% YoY, AI-cloud revenue up 514% to $575M, AI-cloud ARR at $3.0B, and EBITDA margin around 40% — with a $7–9B year-end run-rate target off a $3.0–3.4B full-year guide.

The problem is ambition: $20–25B capex, a $17.4B Microsoft deal and $3B Meta deal anchoring demand, but ~$250B of estimated cumulative funding need for 5 GW by 2030. Burry is short near $212; one broker cut its target on Vineland construction delays. The only one of the four with positive TTM GAAP EPS — and the most expensive.

$582.3MQ2 revenue · +454% YoY
$3.0BAI-cloud ARR
~40%adj. EBITDA margin
$20–25B2026 capex plan

02What Nebius actually is

Nebius is the Amsterdam-based AI cloud spun out of Yandex (previously Yandex N.V.), rebuilt as a full-stack AI infrastructure company: GPU cloud by the megawatt, plus the platform software, models, and developer tools on top, and a few side bets (TripleTen edtech, Avride autonomous delivery).

The differentiator versus CoreWeave and IREN is the stack: it owns the software layer on top of the metal, which is why margins run near 40% and why it can claim a full-stack position rather than pure rental. It carries a strategic alliance with Nvidia, with two spectacular contracts: $17.4B with Microsoft and $3B with Meta.

Management targets 5 GW of contracted power by year-end 2026, and new AI contracts typically bring 50–60% upfront cash coverage of capex — the key mechanism that makes a $20–25B buildout fundable.

03The bull case

1 · Growth is the realest in the group

  • +454% YoY revenue; AI-cloud +514%; ARR $3.0B; $7–9B year-end run-rate target
  • The only neocloud where “hockey stick” is past tense

2 · GAAP-profitable (barely, but yes)

  • TTM GAAP EPS +$0.33 — the only positive EPS in the group
  • EBITDA margin ~40% and moving toward real profit; reported P/E ~1,300× reflects how early this is

3 · Contracts de-risk the capex

  • $17.4B Microsoft + $3B Meta anchor demand
  • 50–60% upfront cash coverage on new contracts; July saw a $775M ABS facility at mid-single-digit rates

4 · The Street is broadly constructive

  • Consensus Moderate Buy, avg target $260.20 (TipRanks-reported ~$280); Baird $340 high
  • ~+17% to consensus; Truist initiated Buy (Sep 2026)

04The bear case

1 · Priced for flawless execution

  • At ~$222, the $57.9B cap assumes the $7–9B run-rate, ~40% margins, and an unstalling funding machine
  • No room for a miss — the stock already fell from ~$300 once

2 · The funding need is staggering

  • Analysts estimate ~$250B cumulative need for the 5 GW/2030 plan
  • $20–25B capex vs. $3.0–3.4B revenue guide — buildout outruns revenue by an order of magnitude

3 · Execution cracks are visible

  • DA Davidson cut to $175 from $250 on Vineland, NJ construction delays — potentially threatening 2026 guidance
  • Delivery is the whole thesis, and delivery just wobbled

4 · High-profile skepticism

  • Michael Burry disclosed a short near $212; Piper Sandler is Neutral at $224
  • A crowded, expensive stock has a long way to fall if sentiment turns

05The numbers

MetricValueNote
Q2 2026 revenue$582.3M+454% YoY
Q2 AI-cloud revenue$575M+514% YoY
AI-cloud ARR$3.0BAt Q2 end
TTM revenue$1.36B
TTM GAAP EPS+$0.33Only positive EPS in group
Adj. EBITDA margin~40%Guided ~40% for 2026
FY26 revenue guide$3.0–3.4BBase ~$530M in 2025
Year-end run-rate$7–9BAnnualized target
Capex plan$20–25BQ2 capex ~$5.7B
Microsoft deal$17.4BGPU capacity
Meta deal$3B
Backlog (reported)~$46BThird-party estimate
Power target5 GWContracted by year-end
ConsensusModerate BuyAvg $260.20; Baird $340 high
Implied upside~+17%DA Davidson cut to $175

06What could change the story

Bull triggers

  • Nov 10 print confirms the $7–9B run-rate, with the Microsoft deal near full run-rate
  • Vineland delays resolve without hitting 2026 guidance — the “execution edge” returns
  • More 50%-upfront contracts; each de-risks the next tranche of funding
  • Adjusted EBITDA moving solidly into profit on the guided ~40% margin

Bear triggers

  • Vineland (or any site) delays push 2026 guidance down — the DA Davidson scenario
  • Funding terms deteriorate: cheap contracted-cash-flow debt stops being available
  • A Microsoft/Meta milestone slips — the backlog is concentrated in few customers
  • The ~$250B cumulative funding estimate becomes the market’s framing

07What’s next

NOV 10
2026

Q3 earnings, after close. The key check: progress to the $7–9B run-rate, Microsoft/Meta delivery, and Vineland updates.

Q4
2026

Run-rate proof. Whether the year ends near the $7–9B annualized target determines the 2027 setup.

2027

Vineland + 5 GW. Construction resolution and the continued march to 5 GW contracted power.

08FAQ

Is Nebius profitable?

Nominally yes on GAAP EPS (+$0.33 TTM) — the only one of the four — with adjusted EBITDA margin around 40%. But revenue is tiny relative to the buildout, and the reported P/E (~1,300×) reflects how early this is.

What does Nebius actually do?

Full-stack AI cloud: GPU infrastructure by the megawatt plus the platform software on top, spun out of Yandex, anchored by a $17.4B Microsoft GPU deal and a $3B Meta deal.

Why is NBIS so expensive?

The market prices the run-rate trajectory: +454% growth, a $7–9B year-end target, ~40% EBITDA margin, and 50–60% upfront-funded contracts. Highest-quality story in the group — and the richest price.

Is NBIS a buy?

Consensus Moderate Buy, avg target $260.20 (Baird $340 high) — but Burry is short, one broker cut to $175 on construction delays, and the funding need is enormous. High conviction on both sides. This page is research, not a recommendation.

Sources & method