Nebius Group N.V. (NBIS)

The panel's take

Verdict: WAIT · Conviction: MEDIUM · Last price: USD 236.33 (as of 2026-07-01)

Nebius is the panel's cleanest "real business, wrong price, fresh crack in the thesis" call. The growth is genuine and enormous: Q1 revenue jumped 684% year-over-year to $399M, AI-cloud ARR reached $1.92B (+54% QoQ), and the company carries a $50B contracted backlog anchored by Microsoft ($17.4B through 2031) and Meta (up to $27B), with Nvidia holding a ~$2B strategic stake and priority GPU access. But the stock still trades at roughly 60-70x trailing sales (~20x guided 2026 revenue) — pricing near-flawless multi-year execution — while the underlying business is loss-making (the headline GAAP profit came from a one-time ~$780M non-cash revaluation gain; the adjusted net loss widened to ~$100M) and consumes capital by design: 2026 capex is guided to $20-25B against ~$3B of revenue, roughly seven times sales. Then, on July 1, the reason for the ~13% drop that brought it back to $236: Bloomberg reported that Meta — its largest customer — is building "Meta Compute" to resell its own excess GPU capacity, i.e. the anchor tenant turning into a competitor. Four of six frameworks are bearish, with fair-value estimates clustering from ~$110 to ~$250, mostly below the price; only the disruptive-innovation lens is a firm bull, and the momentum lens is riding it but trimming. With a legitimate AI-infrastructure story, no margin of safety, and a live customer-concentration threat that management hasn't yet answered, the synthesis is WAIT — reassess after clarity on Meta and the next quarter.

Key levels

Key resistance starts at ~$262 (a near-term shelf), then ~$285, up to the $300 all-time high ($299.86) that must be reclaimed to negate the blow-off read. Key support sits at ~$222 (the round-number shelf near recent lows), then ~$190-200 (the prior breakout-retest zone — the level that would question the uptrend if lost), with deeper structural support near ~$155 (roughly a 50% retracement of the parabolic advance). The analyst consensus target spans an extraordinarily wide USD 120 to 380, averaging near USD 237 — essentially the current price, meaning consensus has been overtaken by the selloff rather than re-rated for the Meta news; the panel's own fair values ($110-250) sit at and below that, and target cuts are likely as the Meta Compute overhang is digested.

What legendary investors think

We ran Nebius past a panel of six legendary investors' frameworks.

Where they agree — and where they clash

Common ground

• The growth and backlog are real: revenue +684%, ARR $1.92B (+54% QoQ), a ~$50B contracted backlog (Microsoft + Meta), an Nvidia strategic stake, and 4GW of contracted power.
• The core business is loss-making and capital-hungry: the GAAP profit was a one-time revaluation gain, the adjusted net loss widened to ~$100M, and 2026 capex ($20-25B) is ~7x revenue.
• The valuation prices perfection: ~60-70x trailing sales (~20x forward), a large premium to CoreWeave (~7-8x), with fair-value estimates mostly below the price.
• The Yandex lineage is no longer the risk — it's been cleanly resolved; the risk migrated to capital intensity and customer concentration.

The real debate

• Moat or middleman? The growth lens sees a picks-and-shovels AI-compute leader with an unmatched backlog; the value and forensic lenses see a commodity GPU-reseller whose only edge — getting chips faster than hyperscalers — evaporates when a hyperscaler decides to compete.
• Is the Meta news a crack or a catastrophe? Bulls say Meta Compute may target a different segment and the $27B contract stands; bears say the largest customer becoming a competitor is "the thesis unwinding in real time."
• Does the financing hold? The build-out needs continuous debt/equity to fund $20-25B of capex; bulls trust backlog conversion, bears flag ~zero FCF, rising dilution, and insider selling into strength.

The question it comes down to: Is Nebius an early, Nvidia-backed leader in a multi-year AI-infrastructure build-out whose $50B backlog justifies owning the volatility — or a capital-consumptive GPU-reseller priced for perfection, with fair value well below the quote and its biggest customer now building the very product it buys?

The numbers

Metric	Value
Price / Market cap	USD 236.33 / ~61B
P/S (TTM / fwd FY26)	~60–70x / ~19–22x (vs CoreWeave ~7–8x)
Revenue (Q1 2026)	USD 399M (+684% YoY); AI-cloud ARR $1.92B (+54% QoQ)
Profitability	Adj. net loss ~-$100M; GAAP profit was a one-time ~$780M non-cash revaluation gain; AI-cloud adj. EBITDA margin ~45%
2026 capex (guided)	USD 20–25B (~7x revenue)
Contracted backlog	~USD 50B (Microsoft ~$17.4B to 2031; Meta up to $27B)
Balance sheet	Cash ~$1.7B (≈$9.3B incl. recent raises); FCF ~zero; ~254M shares (+~30% in 18 mo)

Figures as of Q1 2026 (reported mid-May 2026) / July 1, 2026; sourced from Nebius IR/SEC 6-K, Futurum, Motley Fool, Bloomberg, MacroTrends, StockAnalysis, TipRanks. Headline GAAP net income (~$621M) was driven almost entirely by a $780.6M non-cash equity-revaluation gain (the Nvidia/other holdings), not operations — the adjusted net loss ($100M) is the representative figure. FY2026 guidance: revenue $3.0-3.4B, ARR target $7-9B exiting 2026, ~40% adjusted EBITDA margin. The ~7x capex-to-revenue ratio and ongoing dilution are the swing variables; the July 1 "Meta Compute" report (largest customer building a competing resale business) is the live catalyst behind the selloff.

The bottom line

Nebius is the panel's sharpest test of a real business against a demanding price and a freshly opened risk. The bull case is legitimate and, on the top line, spectacular: 684% revenue growth, ARR compounding 54% a quarter, a ~$50B backlog anchored by Microsoft and Meta, Nvidia as both supplier and shareholder, and 4GW of contracted power — the picks-and-shovels profile the disruptive-innovation lens is happy to underwrite toward a multi-year $500-700 outcome. But five of six frameworks see the same problems the bull acknowledges: the business is loss-making once you strip out a one-time revaluation gain, it burns capital by design ($20-25B of 2026 capex against ~$3B of revenue, funded by debt, dilution, and customer prepayments), it trades at ~20x forward sales with fair-value estimates clustering from ~$110 to ~$250 — mostly below the price — and it just absorbed a genuine blow to its moat thesis, with its largest customer reportedly building the very service it sells. Insiders have sold 17 times with zero buys since January, and short interest sits near 24% of the float. What would tip the call bullish is clarity that Meta Compute is additive rather than substitutive, continued backlog growth, and capex financed without heavy dilution — most of which the next quarter and management's response will start to answer. What keeps it a wait is that, until then, buyers are paying a priced-for-perfection multiple for a capital-hungry, loss-making reseller whose biggest customer just became a competitor — a real story worth revisiting, at a price and with a clarity that don't yet exist.