Nebius at $191 With a $40B Backlog and a 7x Capex-to-Revenue Ratio

Nebius Group traded around $191 on Tuesday, up more than 3% on the session after a reiterated Buy rating and a Street-high $286 price target, implying roughly 49% upside. That advance interrupts a sharp decline: the stock has fallen about 36% from its June peak near $300 and dropped 16% in the past week alone.

Against that, year-to-date performance remains between 111% and 122% depending on the measurement date, which makes NBIS one of the best performers in the AI infrastructure trade. The stock also surged 27% during one week following a multi-year AI cloud agreement worth more than $1 billion with Reflection AI, an Nvidia-backed startup.

The company reports second-quarter results before the market opens on Wednesday, August 12. Consensus places revenue between $569.89 million and $574.65 million, an increase of 442% to 447% from $105.1 million in the year-ago quarter. Loss estimates span $0.67 to $0.86 per share on a GAAP basis, against $0.38 in the prior-year quarter, with non-GAAP consensus at $0.70.

The first quarter set the benchmark and it was exceptional. Revenue reached $399 million, up 684% year over year, and adjusted EBITDA swung to a $129.5 million profit from a $53.7 million loss a year earlier. Headline net income of $621 million was flattered by a large non-cash gain on revaluation of equity investments; the underlying adjusted net loss was approximately $100 million.

That distinction between reported and underlying results is where the analytical work sits. A company that reports $621 million of net income while losing $100 million on an adjusted basis is a company whose income statement requires reconstruction before it can be assessed.

The forward commitments are enormous. Nebius has raised its 2026 capital expenditure target to $20 billion to $25 billion from a prior $16 billion range, expanded target AI compute capacity above 4 gigawatts for 2026, and maintained full-year revenue guidance of $3.0 billion to $3.4 billion at an adjusted EBITDA margin near 40%. It targets annualized run-rate revenue of $7 billion to $9 billion by year-end.

Michael Burry has disclosed a short position. CoreWeave reports after Tuesday's close, which provides a sector read hours before Nebius opens the books.

$25 Billion Of Capex Against $3.4 Billion Of Revenue

The capital intensity is the number that defines this equity, and the ratio is unlike anything in conventional technology investing.

The 2026 capital expenditure target of $20 billion to $25 billion sits against full-year revenue guidance of $3.0 billion to $3.4 billion. At the midpoints, that is $22.5 billion of spending against $3.2 billion of revenue, a ratio of 7.0 times.

For context, a hyperscaler typically spends 15% to 25% of revenue on capital projects. Nebius is spending 700%. That is not a technology company profile. It is a utility or a greenfield industrial developer, and it should be valued accordingly.

The revision matters as much as the level. The target moved to $20 billion to $25 billion from a prior range starting at $16 billion, which means the company raised its spending plan by roughly 40% to 56% at the top end within a single guidance cycle. Raising capex that aggressively signals either that demand has outrun the prior plan or that construction costs have escalated.

The compute target explains the spending. Nebius has expanded its target AI compute capacity to more than 4 gigawatts for 2026, and the Pennsylvania AI factory alone represents 1.2 gigawatts. At prevailing costs for GPU-dense data center capacity including hardware, roughly $20 billion to $25 billion is the arithmetic requirement for that footprint.

The financing question follows directly. A company generating $3.2 billion of revenue at a 40% adjusted EBITDA margin produces roughly $1.28 billion of adjusted EBITDA. Funding $22.5 billion of capital projects from $1.28 billion of operating cash flow requires more than $21 billion of external capital across a single year.

That is the entire bear case in one line, and it is the reason the stock fell 36% from its June high while revenue quadrupled.

The offsetting consideration is the asset-light shift. Expectations point toward Nebius moving to a more asset-light model, which could support growth without a proportional rise in capital spending. Confirmation of that pivot Wednesday would be more consequential for the equity than any revenue figure.

Q1 Delivered $399 Million And 684% Growth With Positive EBITDA

The prior quarter is the standard against which Wednesday's print will be judged, and it cleared a bar that most of the sector cannot.

First-quarter revenue reached $399 million, up 684% from the year-ago period. Adjusted EBITDA reached $129.5 million against a $53.7 million loss a year earlier, a swing of $183.2 million. That produced an adjusted EBITDA margin of 32.5%.

Positive adjusted EBITDA at a company growing 684% is the single most important differentiator in this sector. CoreWeave delivered $1.2 billion of adjusted EBITDA on $2.08 billion of revenue in its first quarter, a 57.7% margin, but against a GAAP loss of $740 million. Nebius produced a smaller absolute figure at a lower margin with a substantially smaller underlying loss.

The reconciliation of the headline number is where the caution belongs. Net income of $621 million was inflated by a non-cash gain on revaluation of equity investments. Stripping that out leaves an adjusted net loss near $100 million, which is the figure that describes the operating business.

That gap of roughly $721 million between reported and underlying results means the income statement cannot be read at face value. The equity stakes generating those revaluation gains are real assets, and a holding in ClickHouse represents genuine value, but marking them up does not fund a data center.

The sequential progression to the second quarter implies deceleration on percentage terms and acceleration on absolute terms. From $399 million to $574 million is 43.9% sequential growth. Year-over-year growth falls from 684% to roughly 445%, which is arithmetic rather than deterioration since the comparison base has risen.

The full-year guidance of $3.0 billion to $3.4 billion requires substantial second-half acceleration. First-half revenue at $399 million plus $574 million equals $973 million, which means the second half must deliver $2.03 billion to $2.43 billion. That is 109% to 150% sequential growth across two quarters.

The previous earnings report sent shares nearly 16% higher in a single session, which establishes the reaction precedent.

The $7 Billion To $9 Billion ARR Target Is The Real Test

The annualized run-rate revenue guidance is the metric that carries the valuation, and the gap between it and current revenue is the largest single leap in the sector.

Nebius targets annualized run-rate revenue between $7 billion and $9 billion by the end of 2026. Second-quarter revenue of $574 million annualizes to $2.30 billion. Reaching $7 billion requires the exit run rate to be 3.0 times the second-quarter annualized figure. Reaching $9 billion requires 3.9 times.

That trajectory has to be delivered across two quarters. From a $2.30 billion annualized base in the second quarter to $7 billion by December requires quarterly revenue reaching $1.75 billion by the fourth quarter, which is 205% above the second-quarter level. The $9 billion target requires $2.25 billion quarterly, or 292% above.

Those figures are achievable only if contracted capacity comes online in large discrete blocks rather than incrementally. That is precisely how GPU data centers commission: a facility either delivers 200 megawatts or delivers nothing, and revenue steps rather than ramps.

The relationship between ARR and reported revenue is what makes the target credible or not. A run rate measured in December reflects capacity operating in that month, while full-year revenue reflects capacity averaged across twelve months. A company that commissions most of its capacity in the fourth quarter reports modest annual revenue against an enormous exit run rate.

That structure explains why full-year guidance of $3.0 billion to $3.4 billion coexists with a $7 billion to $9 billion exit ARR. The two are consistent if the second half delivers the bulk of new capacity.

The pricing environment supports the revenue side. Demand for AI compute capacity is strong with tight supply helping maintain firm pricing, and the company sold out its first-quarter capacity despite higher prices. Selling out at higher prices is the cleanest available evidence that the constraint is supply rather than demand.

Investors will focus on how tight supply and firm pricing affect the 2026 revenue guide, and any raise to the $3.0 billion to $3.4 billion range would validate the ARR path.

The Vineland Delay Is The Specific Risk To The Narrative

The most concrete bearish development ahead of the print is a construction timeline, and it produced a substantial target reduction.

A Hold rating was reiterated with the price target cut to $175 from $250, with the reasoning that delays at the key Vineland facility challenge the narrative and that the build-out is not expected to be completed this year.

That is a $75 reduction, or 30%, on a single facility's schedule. The magnitude reflects how concentrated the ARR target is in a small number of large sites. If Vineland does not commission in 2026, the capacity that was supposed to carry the fourth-quarter run rate does not exist, and the $7 billion to $9 billion figure becomes arithmetically unreachable.

The same assessment acknowledged that Nebius has built a strong reputation as a quality neocloud with an experienced management team and the ability to rapidly scale contracted gigawatt capacity, a better capital structure than peers including CoreWeave, and solid ancillary businesses led by the ClickHouse stake.

That combination is what produces the valuation premium. Nebius trades at a 21% premium on backlog relative to CoreWeave and a 172% premium to Oracle's AI cloud business.

A 21% premium to CoreWeave on backlog is defensible if the capital structure is genuinely better, since backlog is worth more when the company financing it faces lower distress risk. A 172% premium to Oracle is a different proposition, since Oracle operates at scale with investment-grade credit and diversified revenue.

Investors remain focused on execution as the company expands its data center footprint, which is the correct framing. Demand is contracted, pricing is firm, and capacity is sold out. The variable is whether concrete gets poured and GPUs get racked on schedule.

Management commentary on data center expansion, GPU deployments, and construction at the 1.2-gigawatt Pennsylvania AI factory is where Wednesday's report will be won or lost. Any confirmation that Vineland slips into 2027 removes the ARR target regardless of what the revenue line shows.

A $40 Billion Backlog From Microsoft And Meta Anchors Demand

The contracted revenue base is the asset supporting the entire capital program, and its counterparties are the strongest available.

Nebius carries a contract backlog of roughly $40 billion anchored by Microsoft and Meta Platforms. Against a market capitalization implied by a $191 share price, that backlog is a multiple of the equity value.

Counterparty quality is what distinguishes this backlog from a pipeline. Microsoft and Meta are investment-grade entities with capital budgets measured in tens of billions annually, and both rose in Monday's session as cloud demand signals held. Microsoft gained 1.21% to $506.06 and Meta added 0.48% to $594.92.

The Reflection AI agreement adds a different customer profile. That multi-year AI cloud contract worth more than $1 billion comes from an Nvidia-backed startup, which introduces venture-funded counterparty risk alongside the hyperscaler base. It also produced a 27% weekly share price gain, which measures how much the market values incremental contract announcements.

The asset-backed financing structure combined with the backlog supports ongoing capital-intensive growth even after the company shifted to a net debt position in the second quarter. That shift from net cash to net debt is a threshold event: the balance sheet has moved from funding growth internally to funding it externally.

Nvidia holds a 9.3% stake in Nebius, which enhances technology access and customer security. That stake is tied to warrants from March and cannot be sold before September 2026, which removes it as a supply overhang through the current quarter.

The strategic value of that holding runs beyond the equity. A GPU supplier with an ownership position in a customer has an incentive to prioritize allocation, which is the scarcest resource in this sector. Access to Vera Rubin generation hardware, which Nebius launched at its Finland data center, follows from that relationship.

Capacity expansion beyond the current footprint is under evaluation. Potential data center deployment in India would extend beyond the existing North America, Europe, and Israel presence, and hiring activity suggests the company may exceed its 2030 guidance of 5 gigawatts, possibly reaching above 7 gigawatts.

Michael Burry's Short Puts Leverage At The Center

The disclosed short position is the most visible bear thesis on this name, and its specifics are relevant rather than merely sentimental.

Michael Burry disclosed short positions against Nebius and Oracle, with the reasoning centered on high leverage and capital spending needs amid rapid growth. NBIS dropped for two consecutive days following the disclosure, falling to $187 and 37% below its high.

The thesis is structurally coherent. A company spending $20 billion to $25 billion against $3.2 billion of revenue, having shifted to net debt, funding construction with asset-backed facilities, and depending on facilities that have already slipped, carries the risk profile the short is targeting.

The pairing with Oracle sharpens it. Both companies are financing AI infrastructure buildouts with debt against contracted revenue from a small number of hyperscaler counterparties. If that model breaks, it breaks for both simultaneously, which makes the pair a coherent expression of a single view about AI capital intensity.

The market has partially validated the concern independently. Nvidia fell 2.88% Monday as a potential $500 billion structured-finance package tied to hyperscaler infrastructure expansion drew scrutiny. Intel dropped 4.06% on a $15 billion equity offering. Capital intensity is being penalized across the AI supply chain regardless of demand.

The counterargument rests on the capital structure comparison. Nebius has been assessed as having a better capital structure than CoreWeave, which has raised more than $30 billion in debt and equity year to date against a $43 billion market capitalization and carries Ba2/BB+ ratings with facilities priced at SOFR plus 550 basis points.

The ancillary assets provide a second offset. The ClickHouse stake and other equity investments generated the revaluation gain that produced $621 million of reported net income in the first quarter. Those holdings are monetizable in a way that a data center under construction is not.

The rate environment is the variable that determines which side is right. The US 10-year at 4.726% approaching a seven-month high raises the cost of every dollar of asset-backed financing, and money markets price 22 basis points of Federal Reserve tightening by the end of 2026, up from 17 basis points on Friday.

A $175 To $286 Target Range Spans 63%

The distribution of institutional views quantifies how little consensus exists on this business.

Wall Street carries a Moderate Buy consensus based on seven Buys and four Holds, with an average target of $241 implying roughly 26% to 31% upside depending on the measurement date. A broader survey of 18 analysts produces a Buy rating with a $255.44 average target and 33.31% implied upside. One aggregated fair value estimate sits at $270.06.

The extremes are what matter. The Street-high target sits at $286, implying about 49% upside. The low sits at $175, implying roughly 8% downside. That range spans 63% of the low value, which is an extraordinary dispersion for a company with $40 billion of contracted backlog from named investment-grade counterparties.

The reason for the spread is that the two sides are modelling different questions. The bulls model demand, pricing, and the ARR trajectory. The bears model construction schedules, financing costs, and the gap between capital spending and revenue. Both are correct about their respective variables.

Coverage was initiated recently with the view that the asset-light expansion model supports long-term growth while investors remain focused on execution. That framing captures the split precisely: the model works if it executes.

The target revisions have moved in both directions within days. One firm cut to $175 from $250 on the Vineland delay. Another reiterated $286. Those two actions occurred in the same week on the same company ahead of the same report, which means the market has no anchor.

Five key themes have been identified as investor focus areas, with pricing, demand, and the 2026 ARR outlook leading. That prioritization is correct: the revenue line for the second quarter is already known within a narrow band, and the guidance is what carries the stock.

For a trader, the practical implication is that the target dispersion translates directly into expected volatility. A stock where the institutional range spans 63% will move substantially on any guidance revision.

The Stock Fell 36% From $300 While Revenue Quadrupled

The price action across two months is the clearest measure of how the market is pricing capital intensity.

NBIS peaked near $300 in June and traded to roughly $187 to $191, a decline of 36% to 38%. Across the same period, revenue growth ran at 684% year over year in the first quarter with second-quarter consensus at 445%.

That divergence between operating performance and equity performance is not a valuation correction. It is a change in the multiple the market applies to contracted growth funded by debt.

The year-to-date figure keeps the longer arc constructive. The stock remains up 111% to 122% for 2026, which means the June peak represented an advance of roughly 250% from January before the retracement. A 36% drawdown from a 250% advance is a normal consolidation.

The 16% decline in the past week is the more concerning datapoint. That move occurred in the run-up to earnings rather than in response to news, which indicates positioning reduction ahead of a binary event rather than a fundamental reassessment.

Tuesday's 3% recovery on a reiterated Buy and a $286 target demonstrates how sensitive the stock is to sentiment at these levels. A single reiteration moving a stock 3% ahead of its own earnings report signals thin positioning and low conviction on both sides.

The report lands with the stock no longer priced for perfection, which is the meaningful change from June. At $300 the equity required flawless execution on a $25 billion capital program. At $191 it requires competent execution, which is a materially lower bar.

The peer comparison places NBIS ahead. The stock has outperformed its neocloud peers, with CoreWeave up 19% to 30% year to date depending on measurement and trading roughly 41.9% below its own high.

Options positioning provides no clean read since the stock reports before the open rather than after the close, which means the reaction occurs in a single session rather than in after-hours trading.

CoreWeave's Print Tonight Sets The Sector Read

The sequencing of the two reports creates a specific information dynamic that traders can use.

CoreWeave reports after Tuesday's close with consensus revenue at $2.55 billion to $2.56 billion against $1.21 billion a year ago, growth near 111%, and a loss of $1.21 per share against $0.27. Options imply a 13.21% move, or $11.65 on a share price near $89, down from 15.5% priced on Friday.

Nebius reports before Wednesday's open. Between those two events sits roughly fifteen hours, and the July US Consumer Price Index arrives at 8:30 a.m. Eastern Time on Wednesday, minutes before or after the Nebius release depending on exact timing.

That compression creates three overlapping catalysts. CoreWeave's results provide a demand read for the neocloud market, since both companies operate in the same segment and sell to overlapping customers. The CPI print sets the rate environment for both. Nebius's own numbers land in the middle of it.

The sector read matters because the two companies share the critical variables. Both depend on GPU capacity converting into contracted revenue. Both finance construction with asset-backed debt. Both carry backlogs anchored by hyperscaler counterparties. CoreWeave's backlog stood at $99.4 billion at the end of the first quarter against Nebius's approximately $40 billion.

CoreWeave's report could affect sentiment toward Nebius directly, which means a Nebius holder carries CoreWeave exposure overnight without any position in it.

The differentiation between the two is the capital structure and the margin profile. CoreWeave delivered a 57.7% adjusted EBITDA margin against a $740 million GAAP loss. Nebius delivered 32.5% against a roughly $100 million adjusted net loss. CoreWeave has raised more than $30 billion year to date. Nebius shifted to net debt only in the second quarter.

CoreWeave closed a $2.6 billion oversubscribed facility on August 10, the day before its report, which removed the financing question from that print. Nebius has no equivalent recent confirmation.

The 4-Gigawatt Target And The 1.2-Gigawatt Pennsylvania Factory

The physical capacity plan is the constraint that determines whether the financial targets are reachable.

Nebius has expanded its target AI compute capacity to more than 4 gigawatts for 2026, with 2030 guidance at 5 gigawatts and hiring activity suggesting a possible path above 7 gigawatts. The Pennsylvania AI factory at 1.2 gigawatts represents 30% of the 2026 target in a single site.

Building 4 gigawatts of GPU-dense capacity within a year is an industrial undertaking with few precedents. Power interconnection, transformer procurement, cooling systems, and grid capacity each carry lead times measured in quarters, and the bottleneck is typically utility interconnection rather than construction.

That is the mechanism behind the Vineland delay. A facility does not slip because concrete cures slowly. It slips because a transformer arrives late or an interconnection agreement takes an extra quarter, and neither is within the company's control.

The Finland data center provides the counterexample. Nebius launched its next-generation Vera Rubin AI compute rack there, expanding computing capacity on schedule. European operations have proceeded while the US facilities have encountered delays, which suggests the constraint is jurisdiction-specific rather than operational.

The India evaluation would add a third region beyond North America, Europe, and Israel. Geographic diversification addresses the interconnection bottleneck by moving capacity to markets with available power and faster permitting, at the cost of introducing political, currency, and operational risk.

The relationship between gigawatts and revenue is roughly linear at current pricing. A gigawatt of GPU capacity generates revenue in the low single-digit billions annually depending on utilization and contract terms, which means 4 gigawatts fully commissioned supports the $7 billion to $9 billion ARR target with room.

The gap is commissioning timing rather than total capacity. Capacity that comes online in December contributes one month to 2026 revenue and twelve months to the exit run rate, which is why the ARR figure and the revenue guidance diverge so widely.

Updates on GPU deployments and Pennsylvania construction are the specific disclosures that matter Wednesday.

A 40% EBITDA Margin Target Against 32.5% Delivered

The margin trajectory is the third leg of the investment case and the one with the least visibility.

Full-year 2026 guidance calls for an adjusted EBITDA margin near 40%. The first quarter delivered 32.5%, with adjusted EBITDA of $129.5 million on $399 million of revenue. Reaching 40% for the full year requires the remaining quarters to average above 42%.

Margin expansion in this business comes from utilization rather than pricing. A data center's fixed costs are set at commissioning, and every incremental GPU-hour sold above the breakeven point contributes at high marginal margin. Selling out capacity at higher prices, as occurred in the first quarter, is the mechanism.

The offsetting force is that newly commissioned capacity starts at low utilization. A facility coming online in the third quarter carries full depreciation and partial revenue, which compresses margin in the quarter it commissions and expands it thereafter.

That timing effect means a company adding 4 gigawatts across a year will show margin volatility rather than a smooth path. The 40% target is a full-year average that could be reached with quarters ranging from 25% to 50%.

Whether the company can maintain profitability while expanding rapidly is the stated focus, and management's margin outlook will be closely watched. That framing places the guidance ahead of the reported figure in importance.

The depreciation question underlies everything. GPU-dense infrastructure depreciates on schedules between three and six years depending on the assumption about hardware useful life, and the choice between those schedules moves reported EBITDA and net income substantially. A company deploying $22.5 billion of assets carries depreciation measured in billions annually once the fleet is complete.

Adjusted EBITDA excludes that depreciation entirely, which is why the metric can be positive while the adjusted net loss runs near $100 million. Both figures are accurate and they describe different questions: whether the assets generate cash, and whether they generate returns above their cost.

At a 4.726% ten-year and asset-backed financing priced above that, the second question is the one that determines equity value.

Nebius Stock Price Forecast: Levels, Scenarios And Invalidation

The base case places Nebius between $165 and $230 in the reaction to Wednesday's report, with the stock's 36% retracement from $300 having removed the perfection premium and the $175 to $286 target range framing the distribution.

The bullish path requires three confirmations. First, revenue at or above the $574 million consensus with the 2026 guidance of $3.0 billion to $3.4 billion maintained or raised, which validates the second-half acceleration the full-year figure requires. Second, reaffirmation of the $7 billion to $9 billion exit ARR target with specific commentary on capacity commissioning timing, since that target is the valuation anchor. Third, confirmation that Vineland and the 1.2-gigawatt Pennsylvania facility remain on a 2026 timeline, or credible substitution of alternative capacity. Clearing all three targets $220 and then the $241 consensus average, with the $286 Street high requiring a guidance raise rather than a reiteration.

The bearish path requires two. Confirmation that Vineland slips into 2027, which removes the capacity underpinning the exit ARR target, or an adjusted EBITDA margin materially below the 32.5% first-quarter level that contradicts the 40% full-year guidance. Either outcome delivers the stock toward $175, and a combination of both opens the $150 area with the June-to-August downtrend resuming toward the levels the short thesis targets.

Invalidation for the bullish case is a close below $175. Invalidation for the bearish case is a close above $230.

The medium-term case rests on contracted demand at a scale the capital program can monetize. A $40 billion backlog anchored by Microsoft and Meta, a $1 billion-plus Reflection AI agreement, first-quarter capacity sold out at higher prices, revenue growth of 684% with adjusted EBITDA swinging positive by $183.2 million year over year, Nvidia holding 9.3% with warrants locked until September 2026, and a capital structure assessed as better than CoreWeave's all support the premium.

The medium-term constraint is the arithmetic of $20 billion to $25 billion of capital spending against $3.0 billion to $3.4 billion of revenue, a 7.0 times ratio that requires more than $21 billion of external funding in a single year. The company has shifted to net debt. The Vineland build-out is delayed. A disclosed short position targets exactly that leverage. Valuation carries a 21% premium to CoreWeave on backlog and 172% to Oracle's AI cloud. The 10-year at 4.726% raises the cost of every dollar of asset-backed financing.

The trade into Wednesday is the $187 to $191 area. CoreWeave's print tonight sets the sector tone, the July CPI lands the same morning, and Nebius reports into both. Above $230, the retracement from $300 resolves upward and $241 becomes reachable. Below $175, the short thesis gains the execution evidence it has been waiting for.

That's TradingNEWS