NBIS Nebius

Research status: Primary-source verified through the 24 August 2026 financing close; entry/add diligence refreshed 30 August 2026 using the $209.18 close on 28 August 2026.

Nebius continues to show extraordinary AI-cloud growth, but financing and per-share economics remain the central underwriting issue. Q2 group revenue rose 454% to $582.3M; AI-cloud revenue rose 514% to $575M; AI-cloud ARR reached $3.0B and adjusted EBITDA margin was about 50%. Management reaffirmed $3.0–3.4B 2026 revenue, $7–9B year-end ARR and about 40% adjusted EBITDA margin.

The August financing closed at $5.75B gross principal: $3.45B of 0.50% notes due 2030 and $2.30B of 4.50% notes due 2034. Expected net proceeds were about $5.68B. Separately, $800M of older converts are being exchanged for approximately 15.8M Class A shares. The financing materially reduces near-term funding risk for contracted capacity, but increases gross debt claims, coupon/accretion burden and dilution complexity.

At the current $209.18 price, the existing PowerFund scenario values imply approximately 39.8% annualized probability-weighted return over 24 months and 25.0% annualized over five years, but the bear case remains extremely severe at about -64%. The large modeled upside does not by itself justify adding because the scenario dispersion is dominated by whether the newly raised capital creates value per share after dilution, coupons, principal accretion and replacement capex.

Diligence conclusion: hold the current small starter; do not add yet. Require actual post-financing balance-sheet/share-count disclosure and evidence that newly financed capacity converts into signed revenue, EBITDA and cash at attractive per-share returns before increasing exposure.

Investment case

Nebius is building a full-stack AI cloud spanning GPU infrastructure, orchestration, training and inference, with additional optionality from Avride, TripleTen and strategic investments. Demand for scarce AI capacity is being converted into large signed agreements, customer prepayments and attractive management-reported payback periods. A mix of owned, colocated and asset-light capacity may allow the platform to scale faster than traditional clouds.

> The PowerFund thesis is that Nebius can become a large, profitable specialist AI cloud rather than a leveraged commodity GPU lessor. The thesis only works per share if contracted capacity converts into revenue and cash at attractive returns while financing, dilution, depreciation and replacement capex do not consume the value created.

Verified operating baseline

• Q2 group revenue: $582.3M, +454%; AI-cloud revenue $575M, +514%.
• AI-cloud ARR: $3.0B, +56% sequentially; AI cloud was approximately 98% of group revenue.
• AI-cloud adjusted EBITDA: approximately $285.7M, margin 49.7%; group adjusted EBITDA $236.2M.
• Four Q2 cloud agreements averaged >$1B TCV and roughly $20–25M annual contract value per MW.
• Approximately 70% of Q2 deals included prepayments covering 50–60% of associated capex; management estimated ~1 year 10 months payback for those deals.
• 2026 guidance: $3.0–3.4B revenue, $7–9B year-end ARR, ~40% adjusted EBITDA margin, $20–25B capex.
• Q2 capex: approximately $5.7B; management expects >$9B customer prepayments during 2026.
• 30 June balance sheet: $8.04B cash, ~$8.55B current/non-current debt, plus restricted cash; shares outstanding 271.86M.
• Through 30 June, Nebius sold 12.7M shares through its ATM at weighted average $223.60, raising about $2.8B.
• Management raised year-end 2026 contracted-power target to 5 GW and plans to deploy >1 GW per year from 2027.

August 2026 financing — final settlement economics

The financing closed on 24 August 2026 with full exercise of both purchaser options:

• 2030 Notes: $3.45B original principal, 0.50% coupon, initial conversion price ~$313.46.
• 2034 Notes: $2.30B original principal, 4.50% coupon, initial conversion price ~$324.65.
• Total new original principal: $5.75B.
• Expected net proceeds with full option exercise: approximately $5.68B.
• Annual cash coupon on the new notes: approximately $120.8M before any conversions/redemptions.
• The 2030 notes accrete to 110% of original principal at maturity and the 2034 notes to 125%, adding roughly $920M of aggregate principal accretion if held to maturity.
• $800M of older 2029/2031 notes are exchanged for ~15.8M Class A shares, removing roughly $20M/year of old cash coupons while increasing basic shares.
• Pro-forma basic shares after the announced exchange are roughly 287.7M using the 30 June share count, before subsequent ATM/SBC activity.
• At initial conversion rates, the new August notes represent approximately 18.1M shares of potential stock settlement, or ~6.3% of that pro-forma basic share count, before make-whole adjustments. Actual dilution depends on stock price and settlement method.

The financing should not be judged as automatically dilutive or accretive. Initially it adds cash roughly alongside debt, so the key economic question is what return Nebius earns on the $5.68B net proceeds relative to coupon/accretion, future dilution and replacement capex.

KPI ladder

PowerFund valuation scenarios

Reference: $219.13 close on 21 August 2026. This mark is stale for 24 August and must be refreshed before a buy/add decision. The scenario values below remain broad PowerFund estimates, not company guidance; they already assume substantial future dilution and case-specific net debt, so the full exercise of the $750M option changes funding certainty more than the broad valuation range.

24 months

Probability-weighted working value: approximately $409/share. Against the stale $219.13 reference, that implies very high modeled upside, but the dispersion is extreme and the financing/capacity conversion still needs proof.

60 months

The wide range matters more than the weighted value. Financing now provides more runway to reach the upside cases, but also raises the hurdle for proving per-share value creation.

• Delivery of $7–9B year-end 2026 ARR and approximately 40% adjusted EBITDA margin.
• Late-2026 capacity for the four large Q2 agreements coming online on schedule and contributing primarily to 2027 revenue.
• Collection of more than $9B of expected 2026 customer prepayments and maintenance of 50–60% capex coverage on new deals.
• Execution of the 5 GW contracted-power pipeline and deployment of more than 1 GW annually from 2027.
• Asset-backed and asset-light financing reducing the need for corporate debt and equity dilution.
• Diversification beyond the largest strategic customers and continued premium pricing for short-duration capacity.
• Token Factory/inference and full-stack software increasing utilisation, customer retention and value beyond GPU rental.
• Transparent evidence that Avride and strategic equity stakes create per-share value without distracting capital from the cloud.

• Extreme capital intensity: Q2 capex was approximately $5.7B and 2026 capex guidance is $20–25B.
• Financing stack materially larger: the August deal closed at $5.75B, not $5.0B, after full exercise of the $750M purchaser options.
• Coupon/accretion burden: annual cash coupons on the new notes are about $120.8M; stated principal accretion could add roughly $920M by maturity if the notes remain outstanding.
• Conversion/dilution complexity: initial conversion prices are ~$313.46 and ~$324.65; the two new series correspond to ~18.1M shares at the initial conversion rates if fully stock-settled, before make-whole adjustments.
• Existing-note exchange: $800M of older converts are exchanged for ~15.8M Class A shares, roughly 5.8% of the 30 June basic share count, though much of their economic dilution was already embedded in the old converts.
• Full financing stack is complex: March converts, secured debt, customer prepayments, ATM shares, August converts, leases and asset-light arrangements must be assessed together.
• Customer concentration and execution: several large contracts make deployment delays, renegotiations or utilisation misses material.
• Scarcity pricing: current AI-cloud margins and revenue/MW economics may normalise as supply grows.
• Adjusted economics: depreciation, SBC, interest, principal accretion and replacement capex remain weaker than headline adjusted EBITDA.
• Technology obsolescence: GPUs/networking equipment can lose economic value quickly.
• Valuation: large modeled upside does not justify adding without current price data and proof that incremental financing increases value per share.

Warning — investigate and freeze additions

• Quarterly ARR growth falls below approximately 25% sequentially before the 2027 capacity ramp is substantially delivered.
• AI-cloud adjusted EBITDA margin falls below 40%, or capex rises without proportional signed ACV/prepayment support.
• Major capacity milestones slip by more than one quarter or connected capacity materially trails guidance.
• Equity dilution exceeds approximately 5% over 12 months without a demonstrable increase in per-share base-case value.
• EV/forward ARR exceeds approximately 12x while ARR guidance and contract economics stop improving.

Reduce — normally trim 25–50%

• Year-end ARR or revenue guidance is cut by more than 15%.
• AI-cloud adjusted EBITDA margin remains below 30% for two quarters.
• Prepayments cover less than approximately 35% of associated capex on new large agreements, increasing funding dependence.
• A major customer delays, renegotiates or cancels contracted capacity, or customer commitments prove materially less binding than assumed.
• Net debt, convertibles and dilution grow faster than enterprise value created per share.
• Utilisation or pricing for commissioned capacity weakens while competing supply comes online.

Invalidate — exit unless a documented exception is approved

• Contracted capacity repeatedly fails to convert into connected, revenue-producing infrastructure.
• Large customer losses or renegotiations reduce normalised revenue/EBITDA power by more than 20%.
• The AI-cloud operation becomes adjusted-EBITDA negative after the current ramp or cannot fund maintenance/replacement capex through a cycle.
• Capital markets, customer prepayments or asset-backed finance become unavailable on economic terms, preventing contracted delivery.
• Evidence shows the business is primarily a commodity GPU lessor with no durable software, utilisation or cost advantage.
• Governance, disclosure or accounting concerns make customer commitments, margins, capex economics or per-share dilution unreliable.

Nebius competes with AWS, Azure and Google Cloud at the platform level and with CoreWeave, Oracle, Crusoe, Lambda and other specialist capacity providers. Its strengths are speed of deployment, a purpose-built AI software stack, access to current NVIDIA systems, flexible contracts and demonstrated ability to win large customers.

Its disadvantages are smaller scale, weaker diversification, greater financing dependence and less proven enterprise distribution than hyperscalers. The durable moat must come from orchestration software, utilisation, performance per dollar, customer integration and financing efficiency—not merely temporary GPU scarcity. The asset-light model could improve capital efficiency, but it remains early and must be measured separately from owned/colocated economics.

1. Reconcile the full post-settlement financing stack from the next filing: March 2026 converts, July secured debt, August $5.75B converts, remaining older converts, customer prepayments, ATM equity, leases and asset-light financing.
2. Confirm the settlement of the $800M old-note exchange into ~15.8M shares and reconcile the actual updated basic/diluted share count.
3. Track deployment of the ~$5.68B net proceeds by GPUs, data centers, land/power, working capital and other uses.
4. Tie each major financing source to contracted customer economics: TCV, prepayment, capex, MW, revenue start, expected EBITDA and payback.
5. Track connected, commissioned and utilised MW plus revenue/EBITDA per MW by cohort.
6. Recalculate scenarios after the next updated balance-sheet/share-count disclosure.

Position rule: hold current small starter; no add solely because the stock is down or because the financing closed. Addability requires evidence that incremental financing increases base-case value per share.