Combined average of Moat (AI Resilience), Growth, and Valuation scores.
Moat Score
Yandex spinoff turned NVIDIA-backed AI compute pure-play with full-stack infrastructure, NVIDIA partnership, and large hyperscaler contracts — but thinly covered, highly volatile, and unproven at scale.
Nebius's potential edge is vertical integration plus NVIDIA equity backing, but the moat is unproven:
- NVIDIA Strategic Partner: NVIDIA's $2B equity investment positions Nebius as a preferred GPU cloud partner alongside CoreWeave — providing early architecture access (Blackwell, Vera Rubin) and validation.
- Full-Stack Infrastructure Heritage: Inherited from Yandex (formerly Yandex N.V., renamed August 2024), Nebius brings a mature cloud platform, datacenter ops experience, and developer tooling — a differentiator vs. pure GPU rental shops.
- Mega-Contract Customer Anchor: $27B Meta contract and $19.4B Microsoft contract provide multi-year revenue visibility (~$50B backlog), but expose Nebius to the same hyperscaler concentration risk as CoreWeave. That risk sharpened on July 1 2026 when Meta announced 'Meta Compute,' its own GPU rental service undercutting AWS/Azure by 20–30% — Meta is Nebius's single largest customer and is now also a direct competitor for the enterprise GPU-rental business Nebius targets.
Moat Verdict
Nebius's moat is narrow and unproven: NVIDIA equity backing plus inherited Yandex infrastructure heritage provide early advantages, but most of the equity moats are N/A, destroyed or weakened. The business is fundamentally a hypergrowth, leveraged, concentrated GPU cloud with execution risk on a $16–20B capex plan. Investment thesis rests entirely on growth (88 score) and the durability of NVIDIA preferred-partner status. Suitable only as a small speculative position sized to risk; high volatility and thin analyst coverage amplify drawdown potential.
35.0 resilient · 35.0 vulnerable · 80/20 = 35.0 · = 35
Open a moat to read its note.
Nebius sells raw GPU cloud capacity and developer tooling to AI labs and hyperscalers, not consumer UI experiences.
Nebius runs CUDA workloads on NVIDIA hardware with a thin orchestration layer; no proprietary business-logic moat.
Nebius does not derive moat from public data access.
Inherited Yandex datacenter and cloud engineering talent provides early advantage, but talent retention and growth in Western markets is unproven and competitive.
Nebius bundles GPU compute + cloud platform + developer tools (post-Eigen AI), but customers can replicate or migrate to AWS/Azure/GCP — the bundle is convenience-driven, not structural lock-in.
Customer AI workloads are private; Nebius does not derive proprietary data moat from the compute it sells.
Yandex separation in 2024 removed Russian regulatory entanglement; no Western regulatory protection exists.
GPU compute is fungible across providers; no network effect between Nebius customers.
Multi-year capacity contracts (Meta $27B, Microsoft $19.4B) embed Nebius operationally for the contract life, but contract renewal is open competition with CoreWeave and hyperscalers.
Nebius is not the system of record for any customer's AI workload metadata or training history.
A GPU cloud competing against hyperscalers that buy silicon, power and capital at larger scale and lower cost; no unit-cost lead.
Sells a fungible product or capacity: buyers pay the market price, not a brand premium.
Moat in 5–10 years
Nebius's moat holds at its low level, on the same read as CoreWeave. Embedding, scale, bundling and talent are already weakened because GPU capacity is fungible and renewals are open competition with hyperscalers and other neoclouds, and the Meta and Microsoft contracts keep it embedded for their term rather than beyond. Nothing in view moves a pillar further in either direction, and the capex and concentration risk sits in growth.
No moat trending · 35 holds
Combined average of Moat (AI Resilience), Growth, and Valuation scores.
Moat Score
Yandex spinoff turned NVIDIA-backed AI compute pure-play with full-stack infrastructure, NVIDIA partnership, and large hyperscaler contracts — but thinly covered, highly volatile, and unproven at scale.
Growth Score
Q2 2026 (reported August 12) delivered group revenue of $582.3M, +454% YoY and +46% QoQ, with the Nebius AI segment at $575M (+514% YoY, 98% of group). Annualised run-rate revenue reached $3.0B, and management reaffirmed FY2026 guidance of $3.0–3.4B revenue (against FY2025's $530M) and $7–9B of year-end ARR. Pricing is firming: new deals averaged above $20M of revenue per MW, older-generation GPUs repriced more than 30% higher than in Q1, and four Q2 deals averaged more than $1B each with 50–60% of related capex prepaid. Both contracted Meta tranches are delivered and in servicing, and Microsoft's first tranche was delivered on time. Group adjusted EBITDA margin rose to 41% (AI cloud 50%). The offset is the build: ~$5.7B of Q2 capex inside a $20–25B FY2026 plan, 800 MW–1 GW of connected power targeted by year-end, and capacity that must land on schedule for contracted revenue to be recognised.
Valuation Score
At ~$243.48 (September 24, 2026 close, +7.4% on the day after BNP Paribas Exane upgraded to Outperform with a $399 target) NBIS sits ~13% above the reset base ($215) and ~21% of the way to the $350 bull, ~68% above the $145 bear — piecewise 61. The September 5 ladder ($80 / $170 / $280) was built on FY2026 alone; the stock is now priced off 2027, so the ladder is re-anchored on the ~$12.2B FY2027 revenue consensus (BofA $12.7B). At ~$62B market cap (~253M shares) NBIS trades at ~19× the reaffirmed FY2026 revenue guide ($3.0–$3.4B) and ~5.0× FY2027 consensus, and ~1.2× the ~$50B contracted backlog. Q2 revenue was $582M (+454%) with a 41% adjusted EBITDA margin and a $3B annualized run-rate exiting June; FY2026 guidance (run-rate $7–$9B exiting the year, ~40% adjusted EBITDA margin, $20–$25B capex) was reaffirmed. The discount in the base (4.5× vs 5.0× at spot) is for dilution: capex of that size will be funded partly with equity and converts, and the stock remains a speculative, asset-financed position.
The Emerging GPU Cloud Moat
Nebius's potential edge is vertical integration plus NVIDIA equity backing, but the moat is unproven:
- NVIDIA Strategic Partner: NVIDIA's $2B equity investment positions Nebius as a preferred GPU cloud partner alongside CoreWeave — providing early architecture access (Blackwell, Vera Rubin) and validation.
- Full-Stack Infrastructure Heritage: Inherited from Yandex (formerly Yandex N.V., renamed August 2024), Nebius brings a mature cloud platform, datacenter ops experience, and developer tooling — a differentiator vs. pure GPU rental shops.
- Mega-Contract Customer Anchor: $27B Meta contract and $19.4B Microsoft contract provide multi-year revenue visibility (~$50B backlog), but expose Nebius to the same hyperscaler concentration risk as CoreWeave. That risk sharpened on July 1 2026 when Meta announced 'Meta Compute,' its own GPU rental service undercutting AWS/Azure by 20–30% — Meta is Nebius's single largest customer and is now also a direct competitor for the enterprise GPU-rental business Nebius targets.
Moat Verdict
Nebius's moat is narrow and unproven: NVIDIA equity backing plus inherited Yandex infrastructure heritage provide early advantages, but most of the equity moats are N/A, destroyed or weakened. The business is fundamentally a hypergrowth, leveraged, concentrated GPU cloud with execution risk on a $16–20B capex plan. Investment thesis rests entirely on growth (88 score) and the durability of NVIDIA preferred-partner status. Suitable only as a small speculative position sized to risk; high volatility and thin analyst coverage amplify drawdown potential.
35.0 resilient · 35.0 vulnerable · 80/20 = 35.0 · = 35
Open a moat to read its note.
Nebius sells raw GPU cloud capacity and developer tooling to AI labs and hyperscalers, not consumer UI experiences.
Nebius runs CUDA workloads on NVIDIA hardware with a thin orchestration layer; no proprietary business-logic moat.
Nebius does not derive moat from public data access.
Inherited Yandex datacenter and cloud engineering talent provides early advantage, but talent retention and growth in Western markets is unproven and competitive.
Nebius bundles GPU compute + cloud platform + developer tools (post-Eigen AI), but customers can replicate or migrate to AWS/Azure/GCP — the bundle is convenience-driven, not structural lock-in.
Customer AI workloads are private; Nebius does not derive proprietary data moat from the compute it sells.
Yandex separation in 2024 removed Russian regulatory entanglement; no Western regulatory protection exists.
GPU compute is fungible across providers; no network effect between Nebius customers.
Multi-year capacity contracts (Meta $27B, Microsoft $19.4B) embed Nebius operationally for the contract life, but contract renewal is open competition with CoreWeave and hyperscalers.
Nebius is not the system of record for any customer's AI workload metadata or training history.
A GPU cloud competing against hyperscalers that buy silicon, power and capital at larger scale and lower cost; no unit-cost lead.
Sells a fungible product or capacity: buyers pay the market price, not a brand premium.
Moat in 5–10 years
Nebius's moat holds at its low level, on the same read as CoreWeave. Embedding, scale, bundling and talent are already weakened because GPU capacity is fungible and renewals are open competition with hyperscalers and other neoclouds, and the Meta and Microsoft contracts keep it embedded for their term rather than beyond. Nothing in view moves a pillar further in either direction, and the capex and concentration risk sits in growth.
No moat trending · 35 holds
Growth Analysis
Growth Drivers
Key Risk
If $20–25B of FY2026 capex slips on power/datacenter availability, NVIDIA Vera Rubin allocation timing, or convertible debt service capacity, Nebius misses the $3.0–3.4B revenue guide and the $7–9B ARR exit, and the equity re-rates sharply on dilution and a broken hypergrowth narrative. A newer variant: Meta (up to $27B, the largest customer) launched its own competing GPU rental service ('Meta Compute,' July 1 2026, priced 20–30% below AWS/Azure) — raising the question of whether Meta expands third-party commitments at renewal or increasingly serves its own rental ambitions internally.
Score Derivation
95.0 base + 1.3 trajectory + 4 margin − 15 risk = 85
Base 95 (40–60% CAGR, midpoint 50% — saturated above 30%, so cutting the old 100–250% to a sustainable rate costs nothing) + 1.3 trajectory (core AI cloud accelerating; hyperscaler contracts now in servicing and the developer platform both stable: (1 − 0) / 3 × 4) + 4 expanding margin (group adjusted EBITDA margin up to 41% in Q2 2026 from Q1; AI cloud 50%) − 15 severe risk, held with CRWV on the shared neocloud financing risk. The Meta and Microsoft lines move from accelerating to stable because both are delivered and billing — run-rate revenue, not a ramp. The formula computed 84 on the old inputs = 85
Price Scenarios (12–24 Months)
Price vs Targets
Dashed lines mark the 12–24 month bear, base and bull targets. Daily closes.
Valuation Multiples
| Trailing P/E (GAAP) | N/M |
| Forward P/E (NTM) | N/M |
| Price / Sales (FY26 guide) | ~19× |
| Price / Sales (FY27 consensus) | ~5.0× |
| Market cap / Backlog | ~1.2× |
Trailing multiples are meaningless for a business going from $582M a quarter to a $7–$9B exit run-rate. On the FY2027 consensus (~$12.2B) NBIS trades at ~5.0× sales, versus CoreWeave at ~4× FY2026 guided sales in its August review — the premium is for faster growth from a smaller base and a 41% adjusted EBITDA margin, against a backlog (~$50B) less than half CoreWeave's. The reset ladder prices FY2027 revenue at 3× (bear), 4.5× (base, discounted for dilution from $20–$25B capex) and 4× a ~$22B end-2027 ARR (bull, BNP's case). Price has run through the average Street target set before the upgrades; the new high target is $399.
Approximate figures as of September 25, 2026.
Where We Are vs Targets
Loading live price…
Capacity delivery slips on power and datacenter timing, FY2027 revenue undershoots consensus, capex funding forces dilutive equity and convert issuance, and the multiple compresses toward 3× forward sales — CoreWeave-like leverage without CoreWeave's backlog.
- FY2026 revenue lands at the $3.0B low end of guide and the exit run-rate misses the $7–$9B range as capacity comes online late
- Funding $20–$25B of capex requires further equity and convertible issuance at lower prices, diluting existing holders
- NVIDIA reallocates Vera Rubin capacity preference toward CoreWeave or hyperscalers, or Meta scales its own compute rental faster and does not expand its up-to-$27B Nebius contract
- 3× the ~$12.2B FY2027 revenue consensus → ~$36.6B market cap ÷ ~253M shares → ~$145/share (~40% below $243.48)
FY2026 lands in the $3.0–$3.4B guide with ~40% adjusted EBITDA margin, the exit run-rate reaches $7–$9B, FY2027 revenue meets the ~$12.2B consensus, and the market pays 4.5× — a discount to spot for the dilution that the capex plan implies.
- FY2026 revenue lands at ~$3.2B (guide midpoint) with adjusted EBITDA margin near the ~40% guide (Q2 printed 41%)
- The year-end run-rate reaches the $7–$9B guide as the landmark Q2 deals begin contributing in late 2026, and contracted power reaches the raised 5 GW year-end target
- FY2027 revenue meets the ~$12.2B consensus; capex of $20–$25B executes, funded with a mix of debt, converts and equity
- 4.5× ~$12.2B FY2027 revenue → ~$54.9B market cap ÷ ~253M shares → ~$215/share
Capacity ramps faster than plan, sovereign and enterprise demand keeps new capacity sold out as it lands, and ARR approaches ~$22B by end-2027 (BNP's case) — the market pays 4× that exit run-rate.
- Capacity keeps selling out as fast as it comes online (as management said on the Q2 call), and new capacity of more than 1 GW a year from 2027 converts to revenue on schedule
- Eigen AI integration plus developer-platform momentum lift non-hyperscaler revenue mix, de-risking customer concentration
- Adjusted EBITDA margin holds above 40% at scale and a path to GAAP profitability is visible by 2027
- 4× ~$22B end-2027 ARR → ~$88B market cap ÷ ~253M shares → ~$350/share (below BNP's $399 target)
NBIS: common questions
Is Nebius Group stock a buy?
We rate Nebius Group Avoid, with a composite score of 54/100: moat 35, growth 85 and valuation 61. The valuation score is struck against our bear, base and bull targets and moves with the price.
What is Nebius Group's moat?
Nebius Group scores 35/100 on moat. Yandex spinoff turned NVIDIA-backed AI compute pure-play with full-stack infrastructure, NVIDIA partnership, and large hyperscaler contracts — but thinly covered, highly volatile, and unproven at scale.
How fast is Nebius Group expected to grow?
We estimate 40–60% a year over the next three to five years, for a growth score of 85/100.
What is Nebius Group's price target?
Our base case is $215, with $145 in the bear case and $350 in the bull case.