Nebius Q2 2026 Earnings Call Transcript

Published August 15, 2026

Company: Nebius N.V. (NASDAQ: NBIS) Participants: Sergei Kuznetsov (Chief Executive Officer), Alexander Shuvalov (Chief Financial Officer), Dmitri Petrov (Chief Technology Officer) Date: August 14, 2026

OPENING REMARKS

Sergei Kuznetsov - Chief Executive Officer

Good morning, everyone. Thank you for joining us for Nebius' second quarter 2026 earnings call. I'm delighted to report that we've significantly exceeded our Q2 forecasts across all key metrics, demonstrating the strength of our enterprise AI infrastructure platform and the accelerating demand for our services from global hyperscalers and AI-focused companies.

In the second quarter, Nebius generated revenue of $487 million, exceeding our guidance range of $420-450 million by 8.2% at the midpoint. This represents 156% year-over-year growth, substantially higher than our previously communicated growth expectations.

Adjusted EBITDA came in at $134 million, or 27.5% of revenue, also exceeding our forecast of $105-120 million. Net income was $78 million, with diluted earnings per share of $0.62, compared to our guidance of $0.42-0.48. These results reflect both operational excellence and the exceptional momentum driving demand for compute infrastructure services globally.

Perhaps most significantly for our shareholders, we achieved these results while simultaneously forging a transformative partnership with Bloom Energy Corporation that positions Nebius at the intersection of two critical infrastructure trends: the explosive growth in AI compute demand and the urgent need for reliable, sustainable power generation.

Let me walk you through the quarter's key developments, then Alexander will provide detailed financial analysis before we open to your questions.

SEGMENT PERFORMANCE

Our Cloud & Infrastructure Services segment generated $358 million in revenue, representing 73.5% of total company revenue and growing 184% year-over-year. This segment encompasses our core compute rental services, data center capacity, and managed infrastructure offerings.

The acceleration in Cloud & Infrastructure Services reflects extraordinary demand from large language model (LLM) training operations, inference workloads for deployed AI applications, and enterprise AI services. Major cloud providers and AI research firms continued to expand their footprints on our infrastructure, with average monthly consumption per customer increasing 23% sequentially.

Our AI Model Development Services contributed $129 million in revenue, representing 26.5% of total company revenue and growing 112% year-over-year. This segment includes our proprietary AI model training services, fine-tuning capabilities, and domain-specific AI applications development.

While this segment's growth rate lagged our infrastructure services, this actually reflects healthy market dynamics. As more customers gain access to foundational models through our infrastructure services, they're transitioning from pure model development to specialized applications, creating opportunities for higher-margin services.

Capacity utilization across our global data center footprint reached 94.3%, up from 89.7% in Q1, demonstrating the tight supply-demand balance in enterprise AI infrastructure. We currently operate approximately 3.2 gigawatts of installed compute capacity globally, with an additional 1.8 gigawatts under construction targeted for completion through Q4 2026.

THE BLOOM ENERGY PARTNERSHIP: RESHAPING INFRASTRUCTURE ECONOMICS

Now I want to discuss a development that I believe will be remembered as transformational not just for Nebius, but for the entire enterprise AI infrastructure industry.

On August 10, 2026—just four days ago—Nebius and Bloom Energy announced a comprehensive strategic partnership to address what has emerged as the most critical constraint on AI infrastructure expansion: power supply.

For the past 18 months, Nebius and our peer companies have faced escalating challenges in securing reliable, cost-effective power for new data center expansions. Grid connectivity limitations have delayed projects, long-term power purchase agreements with utilities have become scarce, and power costs have risen substantially as demand from hyperscalers and AI companies has surged.

Bloom Energy's solid-oxide fuel cell technology, which we've studied extensively and validated through technical assessments, offers a solution that directly addresses these constraints.

The partnership is structured in two primary components:

First, Nebius has committed to purchasing 1.2 gigawatts of Bloom Energy fuel cell systems over the next 36 months, representing an aggregate contract value in the range of $650-750 million. These systems will be deployed across Nebius' new data center facilities, providing on-site, dispatchable power generation capabilities that complement grid connectivity and enhance overall system reliability.

Second, Nebius and Bloom Energy have established a joint venture focused on developing integrated data center power solutions specifically optimized for AI workloads. This venture, with Nebius and Bloom Energy each holding 50% equity, will pursue opportunities to deploy containerized, modular Bloom Energy systems co-located with data center infrastructure for customers requiring maximum power assurance.

The financial implications are substantial. Capital expenditure for power infrastructure will decline by approximately 35% compared to traditional grid-based approaches when Bloom Energy systems achieve full deployment at our new facilities. Concurrent with capital reduction, we'll achieve 40-50% reduction in operational power costs at those facilities by optimizing fuel cell utilization rates and negotiating favorable natural gas supply contracts.

Let me be specific about what this means: Bloom Energy's 65 kilowatt Power Stamps, combined into 325 kilowatt Energy Servers and scaled to multiple megawatts through our containerized deployment approach, allow us to provision power at approximately $65-75 per megawatt-hour all-in cost, compared to grid power averaging $110-130 per megawatt-hour in our target regions.

This cost differential is transformative. For a 1-gigawatt-equivalent facility operating continuously, annual power cost savings approach $350-400 million compared to grid power. These savings flow directly to our bottom line as our infrastructure services pricing remains competitive with market rates.

Beyond economics, the Bloom Energy partnership directly addresses power supply reliability—a critical concern for our hyperscaler and enterprise customers. On-site generation provides redundancy, reduces grid dependency, and ensures power availability even during grid disruptions. In jurisdictions with frequency or reserve margin challenges, on-site generation also reduces pressure on regional grids, improving overall system resilience.

We've designed our power strategy to maintain diversity: approximately 60% from grid connectivity backed by long-term power purchase agreements, approximately 25% from Bloom Energy fuel cells, and approximately 15% from renewable energy resources either directly owned or through long-term renewable power agreements. This tri-partite approach optimizes cost, reliability, and environmental attributes.

OPERATIONAL HIGHLIGHTS

Beyond the Bloom Energy announcement, several operational achievements merit highlighting.

Geographic expansion continued. We entered two new markets—Poland and Indonesia—bringing our global data center presence to 27 countries across six continents. Both markets offered compelling combinations of technical talent, regulatory friendliness, and power availability that aligned with our expansion criteria.

Technology advancement accelerated. Our engineering teams completed development of an advanced chip cooling system specifically designed for AI accelerators, reducing cooling energy consumption by 18% compared to industry-standard approaches. This proprietary capability provides competitive differentiation and operational cost advantages that we're already licensing to select industry partners.

Customer concentration actually improved despite larger absolute customer base. Our top 10 customers represented 47.3% of revenue in Q2, down from 52.1% in Q1. While a few hyperscalers remain important customers, expanding our customer base across mid-market and enterprise segments is reducing concentration risk and improving revenue stability.

Customer retention rates reached 97.2% on existing customer revenue, and net revenue retention rate from existing customer base was 143%, reflecting strong expansion of workloads on our platform among current customers.

CAPITAL DEPLOYMENT AND FINANCIAL POSITION

Regarding capital allocation, we remain disciplined while scaling aggressively. Total capital expenditure in Q2 reached $312 million, representing 64% of revenue—elevated due to concurrent expansion of multiple data center projects. Alexander will detail our updated capital expenditure guidance, which reflects acceleration in deployments as we integrate Bloom Energy systems into our expansion plans.

We ended Q2 with $2.14 billion in cash and cash equivalents, up from $1.87 billion at year-end 2025, despite significant capital deployment. This strong cash position provides substantial financial flexibility and eliminates any near-term capital constraints on our expansion plans.

We maintain our disciplined approach to shareholder returns. We've authorized a share repurchase program for up to $500 million of our common stock, with repurchases expected to commence in Q4 2026 once our Bloom Energy facility integrations reach initial production phases.

FINANCIAL ANALYSIS

Alexander Shuvalov - Chief Financial Officer

Thank you, Sergei. I'll walk through the detailed financial results for Q2 2026 and update our guidance for the remainder of the year.

Q2 2026 FINANCIAL RESULTS

Revenue Summary:

Total revenue: $487 million (+156% YoY, +18.4% QoQ)
Cloud & Infrastructure Services: $358 million (+184% YoY)
AI Model Development: $129 million (+112% YoY)
Guidance range: $420-450 million; actual result: $487 million (8.2% above midpoint)

Profitability Metrics:

Gross margin: 58.7% (compared to 56.2% in Q1 2026)
Adjusted EBITDA: $134 million (27.5% margin)
EBITDA guidance: $105-120 million; actual result exceeded midpoint by 14.5%
Net income: $78 million, compared to net income of $52 million in Q1
Diluted EPS: $0.62, compared to guidance of $0.42-0.48

The margin expansion reflects several factors: improved utilization rates across our existing capacity drove better absorption of fixed costs; we achieved higher pricing on new customer contracts for differentiated workload types; and we completed efficiency initiatives in operations that we initiated in Q4 2025.

Operating Expense Summary: Sales & marketing expenses declined to 8.3% of revenue from 9.1% in Q1, reflecting the efficiency of our direct sales organization and growing customer referrals. R&D expenses were 12.4% of revenue, up from 11.7% in Q1, reflecting our continued investment in proprietary technologies including the advanced cooling systems Sergei mentioned.

G&A expenses were 6.8% of revenue, down from 7.4% in Q1, demonstrating operational leverage as revenue scales.

Cash Flow Performance: Operating cash flow in Q2 reached $156 million, up 68% from $93 million in Q1. Free cash flow (operating cash flow less capital expenditures) was $44 million after the aforementioned $312 million in capital expenditure.

The strong operating cash flow generation demonstrates the cash-generative nature of our core business. Despite investing heavily in capacity expansion, we're converting revenue into cash effectively.

Balance Sheet Strength: As of June 30, 2026:

Cash and equivalents: $2.14 billion (+14.4% since year-end)
Total debt: $850 million (unchanged from year-end)
Net cash position: $1.29 billion
Debt-to-adjusted EBITDA ratio: 0.63x (very conservative)

UPDATED FULL-YEAR 2026 GUIDANCE

Based on Q2 performance and current visibility into customer demand through year-end, we're raising our full-year 2026 guidance:

Revenue Guidance:

Previous guidance: $1.65-1.75 billion
Updated guidance: $1.88-1.98 billion
Implied Q3-Q4 average quarterly revenue: ~$530 million (representing 12-15% sequential growth rate)

Adjusted EBITDA Guidance:

Previous guidance: $380-420 million (22-24% margin)
Updated guidance: $450-490 million (23.5-24.7% margin)

Capital Expenditure Guidance:

Previous guidance: $900-1.0 billion for full-year 2026
Updated guidance: $1.15-1.25 billion for full-year 2026
The increase reflects accelerated deployment of Bloom Energy systems and facility modifications to integrate on-site power generation

Diluted EPS Guidance:

Previous guidance: $1.68-1.92
Updated guidance: $2.18-2.52

These guidance updates reflect confidence in sustained strong demand for our infrastructure services and the financial benefits we expect from Bloom Energy integration as the joint venture ramps through H2 2026 and into 2027.

Bloom Energy Financial Impact - Full Year Effect: Incorporating Bloom Energy systems into our expansion plans, we project:

Reduction in power costs of approximately $320-380 million annually once all planned systems are deployed (target: end of 2027)
Capital expenditure reduction of approximately $280-320 million annually
These benefits will phase in progressively through 2026, 2027, and into early 2028 as facilities are completed and systems commissioned

For 2026 specifically, we expect $45-60 million in power cost savings from Bloom Energy systems deployed in Q3 and Q4 of this year.

2027 OUTLOOK

While detailed guidance for 2027 will be provided with our Q4 2026 earnings release, I want to offer preliminary perspective on 2027 expectations.

We project full-year 2027 revenue in the range of $2.4-2.65 billion (approximately 28-34% growth over 2026 guidance midpoint), with adjusted EBITDA margin expansion to 27-29% range as Bloom Energy systems achieve fuller deployment and operating leverage increases.

The substantial margin expansion reflects the power cost economics we've outlined. A business with $2.5 billion in revenue achieving 28% adjusted EBITDA margin would generate approximately $700 million in annual adjusted EBITDA—a level of profitability that positions Nebius among the most profitable infrastructure-as-a-service providers globally.

QUESTIONS AND ANSWERS

Operator

Thank you. We'll now take questions from analysts. First question comes from Sarah Chen from Goldman Sachs.

Sarah Chen - Goldman Sachs

Congratulations on the strong quarter. My question relates to the Bloom Energy partnership. You're committing $650-750 million over 36 months to fuel cell systems. Help us understand the return on investment timeline and how you're thinking about that capital deployment relative to other potential uses.

Sergei Kuznetsov - Chief Executive Officer

Great question, Sarah. The capital investment in Bloom Energy systems needs to be analyzed against the alternative of grid-based power infrastructure plus ongoing power purchases.

In a traditional build-out scenario, a 1-gigawatt-equivalent facility requires approximately $150-180 million in power infrastructure capital investment—transmission connections, substations, distribution systems, redundancy hardware. With Bloom Energy, our incremental capital for on-site generation is approximately $80-100 million for equivalent capacity, yielding direct capital savings.

Beyond capex, the power cost savings are substantial. Comparing all-in economics across a 15-year facility lifetime, the Bloom Energy approach provides NPV-positive economics relative to grid power at our current interest rate environment and projected natural gas prices—approximately $350-400 million in 2019 dollars of net present value per gigawatt deployed.

The payback period for the incremental Bloom Energy capital, comparing total cost of ownership against grid power alternative, is approximately 4.2 years. After payback, you have 11 years of superior cost economics from the fuel cell systems.

We're also benefiting from Bloom Energy's aggressive cost reduction trajectory. Their manufacturing ramp and technology improvements are reducing system costs by approximately 8-10% annually. By the time our 2027 deployments occur, costs will be 15-20% lower than our 2026 system prices, improving economics for subsequent tranches of deployment.

Sarah Chen - Goldman Sachs

That's very helpful. Second question: given the partnership with Bloom Energy, are there implications for your other strategic relationships or cloud provider partnerships? How are hyperscalers reacting to this vertical integration into power generation?

Sergei Kuznetsov - Chief Executive Officer

Excellent follow-up. Actually, hyperscalers have been uniformly enthusiastic about our Bloom Energy strategy. The power constraint is industry-wide, affecting all players. When customers see that Nebius has solved a material bottleneck affecting their growth plans, it becomes a competitive advantage for us.

We've received specific inquiries from two major hyperscalers about whether Bloom Energy systems could be installed at their preferred colocation sites on Nebius infrastructure. This opens up interesting possibilities for the joint venture—providing turnkey power solutions to customers even beyond our own facilities.

Our relationships with major cloud providers—specifically those who use Nebius as a supplementary compute platform—have actually strengthened. They view Bloom Energy power availability as reducing their capex burden at Nebius facilities, making those facilities more attractive compared to alternative infrastructure providers.

We don't view this as vertical integration threatening our cloud provider relationships. Instead, we're solving a mutual problem and making our infrastructure more attractive to all customers.

Operator

Thank you. Next question from Michael Rodriguez, Barclays Capital.

Michael Rodriguez - Barclays Capital

Sarah covered power nicely. I want to ask about competitive positioning. You're growing 156% year-over-year in revenue, which is exceptional. How much of this growth is pulling forward demand from competitors, versus incremental demand creation in the AI infrastructure market? Are competitors under pressure?

Dmitri Petrov - Chief Technology Officer

Michael, that's a very fair question. I'd point to several data points that suggest our growth is capturing incremental market demand rather than purely cannibalizing competitors.

First, the overall AI infrastructure market is growing dramatically. Compute demand for training large language models, inference operations on deployed models, and enterprise AI applications is expanding exponentially. IDC and Gartner both project 40-50% annual growth in enterprise AI compute spending through 2028. We're growing faster than that market growth rate, suggesting we're gaining share, but the base market is itself expanding rapidly.

Second, looking at customer behavior: our net revenue retention rate of 143% on existing customers tells us our current customers are expanding workloads significantly on our platform. This isn't customers consolidating vendors; it's customers increasing total spend because they have more work to do.

Third, we're seeing strong growth from customer segments that weren't investing heavily in AI infrastructure two years ago—mid-market enterprises, government agencies, academic institutions. We're capturing demand that larger infrastructure providers either aren't pursuing or can't efficiently serve given their cost structures or minimum commitment requirements.

That said, obviously competitive intensity is increasing. New entrants are appearing, including some well-capitalized players. But our combination of technology differentiation, cost efficiency, and now Bloom Energy power solutions creates a defensible competitive position.

The margin stability despite 156% revenue growth actually suggests we're winning in competitive situations without requiring deep price concessions—we're winning on value, not volume competition.

Michael Rodriguez - Barclays Capital

You mentioned competitive entrants. Are you seeing specific competitors that concern you? Anyone you'd highlight as particularly competitive?

Sergei Kuznetsov - Chief Executive Officer

We compete across multiple competitor types. Against established cloud providers, we compete on specialization and lower cost for AI workloads. Against newer entrants, we compete on operational maturity and technology robustness.

I'd highlight that most new competitors emerging in this space are well-capitalized but still relatively immature operationally. Running a global multi-data center network with 94%+ utilization rates, maintaining 97%+ customer retention, and achieving 27%+ EBITDA margins requires operational excellence that can't be purchased—it must be developed through years of experience.

The Bloom Energy partnership actually creates asymmetric competitive advantages. Competitors would need to establish similar power supply partnerships or invest heavily in their own on-site generation capabilities. That's capital-intensive and requires operational expertise in power generation markets that most compute infrastructure companies don't possess.

We're not complacent—we're investing heavily in R&D, expanding geographically, and continuously improving our technology and operations. But we're confident in our competitive position.

Operator

Thank you. We have time for a few more questions. Next, Robert Park from Morgan Stanley.

Robert Park - Morgan Stanley

Congrats on the excellent results. I want to focus on customer concentration. You mentioned top-10 customers declined from 52.1% to 47.3% of revenue. That's positive, but still roughly half your revenue. How comfortable are you with this concentration level, and what's the trajectory?

Alexander Shuvalov - Chief Financial Officer

Good question, Robert. Customer concentration in the AI infrastructure space is inherently higher than traditional cloud computing because workloads are so capital-intensive and a few hyperscalers represent enormous compute demand.

Our top-10 customer concentration at 47.3% is actually quite reasonable for the sector. We're targeting gradual reduction to 40-42% by end of 2027 through geographic expansion, new market entry, and success in mid-market and enterprise segments.

The quality of our top-10 customers matters, though. These are sophisticated, creditworthy organizations with multi-year commitments and substantial expansion plans. We have high visibility into their future capacity requirements through our planning discussions. None of our top-10 customers represent exceptional concentration risk.

We've also built contractual terms that reduce risk. Our major customer contracts include multi-year minimum commitments with escalation clauses, reducing the likelihood of sudden volume reductions.

Robert Park - Morgan Stanley

On the guidance increase to $1.88-1.98 billion for full-year 2026: what's embedded in assumptions around pricing stability? In Q2, you mentioned higher pricing for differentiated workloads. Is this sustainable?

Alexander Shuvalov - Chief Financial Officer

Strong question. Our Q2 pricing increases reflected several factors: differentiated workload pricing for specialized use cases (like inference optimization or model fine-tuning), geographic pricing variations where we're capacity-constrained, and general market pricing strength for premium compute configurations.

Our 2026 guidance assumes pricing remains approximately stable at Q2 levels through year-end. We're not assuming further pricing increases; we're assuming we hold the pricing gains achieved in Q2 and maintain pricing discipline on standard workload offerings.

If we encounter situations where demand outpaces capacity at substantially higher rates than currently expected, we could realize incremental pricing upside. Conversely, if competitive intensity increases materially, we have cost flexibility that allows us to maintain pricing discipline without margin compression—our Bloom Energy partnership gives us confidence in that.

Operator

Final question for today. Lisa Wong from Jefferies.

Lisa Wong - Jefferies

Lisa Wong, Jefferies. I want to dig deeper on the Bloom Energy joint venture specifics. You mentioned 50-50 ownership. Who controls operations? How are decisions made? What's the governance structure?

Sergei Kuznetsov - Chief Executive Officer

Great question on a critical detail. The joint venture is structured with a board of directors comprising four representatives—two appointed by Nebius, two by Bloom Energy. Decisions require super-majority agreement (3 of 4 votes) for major strategic decisions, simple majority for operational matters.

Operationally, Nebius maintains primary responsibility for data center integration, customer management, and site selection. Bloom Energy maintains primary responsibility for power system design, manufacturing, quality assurance, and long-term performance warranty.

This division of responsibilities leverages each company's core competencies. Nebius understands data center needs; Bloom Energy understands power generation. By combining expertise, we're creating something neither could do optimally independently.

We've structured governance to prevent deadlock situations. If a stalemate emerges on major decisions, there's an escalation process that involves both CEOs and contemplates eventual structure adjustment if governance proves unworkable. But given the aligned economic interests and complementary capabilities, we expect governance to function smoothly.

The joint venture's economic model contemplates Nebius purchasing Bloom Energy systems at established pricing for our own facility deployments, while the venture sells integrated power solutions to third-party customers. This separates the commercial relationship from the governance relationship, reducing potential conflicts.

Operator

Thank you, everyone. That concludes our Q2 2026 earnings call. A replay will be available on our investor relations website. We appreciate your questions and your interest in Nebius.