EX-99.2
3
tm2622968d1_ex99-2.htm
EXHIBIT 99.2
Exhibit 99.2 Nebius Group
Letter to shareholders
Q2 2026
August 12, 2026
Q2 2026: Demand inflecting
Other businesses and investments
Strategic equity investments
Nearly 3x size of autonomous
fleet YTD with 1M+ miles in 2026
Surpassed 60,000 commercial
rides on Uber platform
Launched AI Systems
Engineering programs
and expanded its enterprise
B2B offering
Other businesses
Reported valuation of $15B
in January 2026 funding round
Data solutions business
backed by Bezos Expeditions
Building the full-stack AI cloud
Nebius AI cloud “Aether 3.6”
Introduced Nebius Echo — an AI agent
to help developers operate and manage
workloads
Enhancing the platform to support greater
volumes as customer commitments scale
Token Factory production
inference workloads
increased >3x in Q2
Eigen AI and Clarifai teams joined
Nebius this quarter
New open-weight models in Token Factory:
Kimi K3 GLM 5.2 Nemotron Ultra
Open ecosystem
Demand, proven in signed contracts
Financial highlights
First asset-backed
financing
at attractive terms (SOFR + 2.50%)
~$775M secured in July
Q2 Nebius group revenue
Up 454% YoY to $582M
Nebius AI cloud
adjusted EBITDA margin 50%
Q4’2025 Q1’2026 Q2’2026
45%
24%
Strong cash position ending Q2
Including $2.3 billion in positive
operating cash flow
$8B cash
Repeatable framework
2026 FY
guidance
reiterated
quarter over quarter
for deals closed in Q2
4x
for deals closed in Q2
Сapacity coming online through 2027
All numbers are approximate
ACV per MW is stepping up(1)
Q2 deals
>$20M
Q3 short-term
capacity deals
>$40M
2026 base
$12M
(2)(3) Total contract
value grew
Payback
period
1 year
10 months
Landmark AI cloud wins in Q2
Capacity footprint Raising capacity guidance for YE’26
Existing sites
Sites >100MW
Sites 1GW
New sites added in 2026 Finland
Israel
Iceland
UK
France
New Jersey
Missouri
Oklahoma Alabama
Minnesota
Kansas City Pennsylvania
Spain
2026 contracted power
guidance
>2.5 GW
>3 GW
> 4GW
>1 GW
Aug’25 Nov’25 Feb’26 May’26 Current
5GW
Estonia
4 large deals >$1B average deal size
across the four deals
50-60% of capex self-financed
such as Reflection and Cohere by prepayments
(1) ACV — Annual contract value.
(2) Methodology is on a revenue recognition basis, excluding any prepayment.
(3) Estimated payback period for new deals signed in Q2, based on forecast costs and contracted future capacity, including capacity not yet built.
3
This was the quarter the market validated our strategy.
Demand for AI capacity continues to grow exponentially,
and we are converting that demand into contracted, profitable
growth.
Q2 was our strongest commercial quarter yet
• Deals saw an average yield of more than $20 million per
megawatt.
• We closed four landmark deals for our AI cloud business, for
an average total contract value (TCV) of more than $1 billion
each, and a yield of $20-25 million per megawatt.
• TCV of Q2 wins grew nearly 4x quarter-over-quarter, with TCV
from new customers growing more than 9x.
• New pricing initiatives in early Q3 — such as our first-ever
auction and short-term capacity deals — are showing prom-ise. We see a price opportunity in the $40-50 million per MW
range and we signed our first one this week.
• Overall deal economics strengthened, with 70% of deals
including prepayments, covering 50-60% of the associat-ed capex.
• Together, the expected payback period for the associat-ed capex and related operating costs for Q2 deals is 1 year
and 10 months, down from our two-to-three year payback
period previously.
• We could sell our entire 2027 capacity on these terms today.
We are deliberately not doing so because we see higher val-ue in retaining some capacity for immediate customer needs.
A contract mix built to capture a rising market
We address customer needs with three types of deals. Each has
a different duration, pricing, and role in our business.
Our contract mix can capture this rising market on both ends:
longer, higher-priced contracts lock in future revenue, while short-er contracts can capture near-term upside.
• Shorter-duration contracts, typically between three and six
months, is a new effort, for customers with an acute, time-bound need — priced at a significant premium. One such deal
has been signed in Q3.
• Mid-term contracts — our core business — now with an aver-age duration of one to three years with the world’s most
ambitious AI companies.
• Long-term contracts with investment-grade customers,
which help us finance our buildout faster and more efficient-ly. The $775 million secured facility we raised in July was
on the back of one of these agreements — and with $40 bil-lion in customer commitments, we will do more of this.
Dear
shareholders, Arkady Volozh,
Founder and CEO
“We closed our largest
AI Cloud deals on our strongest
terms to date, at prices that
represent a step-change
in the economics of our business.”
Capacity build-out accelerating, expecting
to deploy >1 GW per year in 2027
• We continue to secure power ahead of our expectations,
and today we are raising our year-end contracted power tar-get again to 5 GW.
• We expanded our footprint with additional sites and pro-gressed on construction at our owned AI factories in the US.
• The pace in which we bring capacity to market will accelerate
as we plan to deploy more than 1 GW per year — and we plan
to do so starting in 2027.
• We are innovating in how we build. This quarter we introduced
an asset-light partnership model that addresses the two con-straints in this industry: capital and capacity.
Software platform for the full AI lifecycle
• Inference workloads on Token Factory more than tripled in Q2,
increasingly running on open models.
• Training and inference are converging. Unlike standalone
inference providers, Nebius serves the full workload lifecycle
on one platform that will drive better total cost of ownership
and utilization of compute.
• The acquisitions of Eigen AI and Clarifai brought indus-try-leading inference optimization in-house.
• We shipped Nebius AI Cloud 3.6 (“Aether 3.6”), scaling
the core platform alongside our growing global footprint.
Financial model proving itself as we scale
• Q2 group revenue of $582.3 million, was up 454%
year-over-year. Nebius AI cloud revenue grew 514% year-over-year to $575 million, ARR reached $3.0 billion.
• We also continued to demonstrate the operating leverage
inherent in our business: our AI cloud business delivered
an adjusted EBITDA margin of 50% in the quarter.
• We expect over $9 billion in customer prepayments
in 2026. We are reaffirming our full-year 2026 guidance
across all metrics.
Arkady Volozh
(4) Annualized run-rate (ARR) is calculated by taking Nebius AI cloud revenue from the last month of the quarter multiplied by 12.
4
Group Q2 update
Customers:
Deal economics strengthened
across every dimension
Q2 was our strongest commercial quarter yet:
• Annual contract value per MW repriced higher. Our core AI
cloud business won deals averaging above $20 million in rev-enue per MW, driven by increasing prices for new-generation
GPUs and more than 30% higher pricing on older-generation
GPUs versus Q1.
• Deal sizes and new logos set records. We closed four land-mark deals, which averaged more than $1 billion in TCV each.
• TCV of deals closed in Q2 grew nearly 4x quarter-over-quar-ter, with TCV from new customers growing more than 9x.
• Roughly 70% of deals closed included customer prepay-ments — an all-time high.
• Prepayment has become the market standard for securing
capacity in a supply-constrained environment.
• Deal terms shortened the expected payback period (3)
on these deals to 1 year and 10 months, from our historical
two-to-three-year range.
These economics will flow through to revenue as the underlying
capacity comes online. The majority of these deals were signed
against capacity arriving in late 2026, and will contribute primar-ily to 2027 revenue.
Our contract mix captures this rising market on both ends: longer,
higher-priced contracts lock in future revenue, while shorter con-tracts can capture near-term upside.
In Q3, we closed our first short-term capacity deal. This deal
represents dedicated capacity for a customer with an acute, time-bound need for compute. The capacity is expected to go live
in Q4. Such deals are typically multi-month engagements, priced
at a premium, for customers who need to move fast and know
exactly what they need.
We also conducted our first capacity auction pilot, seeing early
success by securing the highest price we have cleared for NVIDIA
Blackwell chips to date.
Landmark customer agreements
The strong deals we closed in the quarter includ-ed four landmark deals for our AI cloud.
This success in our core market validates our positioning
and our ability to continue capturing more market share.
The average TCV for these deals was greater than $1 billion,
with annual contract value of $20-25 million per MW.
• Reflection selected Nebius for a multi-year agreement to train
and run its open-source models on our platform, accelerating
open-source AI development while supporting model safety
and the broader developer ecosystem.
• Cohere is partnering with Nebius to support the growth of its
agentic AI solutions for enterprises, grounded in flexible
next-generation NVIDIA architectures, technical collabora-tion, and our track record as a long-term strategic partner.
• Another US-based AI neolab chose our platform for fron-tier model development to accelerate product innovation
in response to growing enterprise demand and to support
their scaling revenue.
• A large US-based quantitative trading firm represents one
of our first large-scale customer wins in financial services.
This firm chose Nebius for our consistent speed and relia-bility at scale, as compute performance is critical to trading
execution.
These large deals are only a handful of the wins we saw across
the range of industries we serve. Customers require infrastruc-ture and tooling that keeps pace with their speed of development
and supports their competitive advantage, and are choosing
Nebius as their infrastructure backbone.
Some of these key vertical wins include:
• AMI, co-founded by Yann LeCun, is developing world mod-els that learn abstract representations of real-world sensor
data and make predictions in representation space. AMI uses
Nebius compute and storage infrastructure to support its AI
research and model training workloads.
• Basecamp Research trained EDEN, a 28-billion-parameter
biological foundation model, on 9.7 trillion tokens from over
a million newly discovered species, entirely on Nebius, at 20x
faster training speed compared to their previous internal
environments. EDEN exhibits state-of-the-art performance
(3) Estimated payback period for new deals signed in Q2, based on forecast costs and contracted future capacity, including capacity not yet built.
5
and designs programmable therapeutic molecules validated
across different modalities.
• Prima Mente scaled its Alzheimer’s and Parkinson’s epige-netic foundation model from 1 billion to 68 billion parameters
on a dedicated Nebius cluster, achieving 89-97% diagnostic
accuracy from blood samples.
• Higgsfield, one of Nebius’s first AI cloud customers, has
expanded its usage on the platform by >20x since their first
contract. This summer they premiered a 95-minute feature
film made entirely with AI on Nebius infrastructure — an indus-try first that compressed a two-year process into weeks,
at a fraction of the cost of traditional filmmaking.
Capacity: Executing our global
capacity expansion
The second quarter was once again defined by strong execution
and significant progress toward our capacity expansion targets.
We continue to bring capacity online to support near-term growth
while securing land and power commitments for 2027 and beyond.
Capacity underpinning near-term growth
In Q2, we continued to deliver on our capacity commitments
across both AI cloud customers and strategic long-term contracts.
• We brought additional capacity online across sites in the US
and Europe during the second half of the second quarter.
• To date, we’ve delivered all capacity tranches to Microsoft
under the contract, including two tranches over the past
month.
In May, we announced a long-term partnership with Bloom Energy
to deploy behind-the-meter fuel cells this year. We are excited
about this partnership as it adds fuel-cell technology to the range
of options available to us for future deployments.
We have already received our first NVIDIA Vera Rubin NVL72
systems and are using these early units to validate compute, net-working, and orchestration together as a single system. Testing
the integration of this new architecture across our full stack will
position us to deploy Vera Rubin as a production-ready offering
for our customers once shipments scale. We are already engag-ing with customers for Vera Rubin access.
• To support agentic AI orchestration, tool-calling and memory
workloads, we are complementing our NVIDIA GPU fleet with
more general-purpose CPU compute, including early adop-tion of NVIDIA Vera CPUs.
Building the foundation for future growth
We continue to secure power ahead of our expectations. We now
anticipate ending 2026 with 5 GW of contracted power(5), up from
the +4 GW we indicated last quarter.
The pace at which we bring capacity to market will accelerate
as we plan to deploy more than 1 GW per year, starting in 2027.
We expanded our contracted capacity footprint in Q2, with addi-tional sites in the UK, Estonia and Finland. Construction is also
underway at some of our owned AI factories in the US.
We also continued to progress the buildout supporting our second
Meta agreement, with capacity on track to come online in early
2027.
New asset-light path to capacity
This quarter we introduced an asset-light partnership model
to supplement the capacity we bring online ourselves.
Capacity remains a binding constraint on AI adoption, and bring-ing capacity online at a rapid pace is central to meeting customer
demand.
To help unlock this, we launched a model that allows partners
to deploy Nebius’s full-stack AI cloud platform in their own AI
data centers.
Nebius contributes:
• Our full-stack AI cloud software,
• Our systems architecture and reference designs, and
• Our global go-to-market organization that brings the demand.
Under this model, partners get fully-owned AI infrastructure
assets, designed to Nebius standards, and a fast route to serve
the AI cloud market; Nebius converts partner-financed capacity
into high-margin revenue with minimal capital outlay.
This complements our owned and colocated portfolio while
reducing the capital burden of adding capacity for customers.
Land and facility
Power
Hardware
Cloud platform
Service and software
GTM
Asset light
Hardware
Cloud platform
Service and software
GTM
Colocation
Hardware
Cloud platform
Service and software
GTM
Owned data centers
System
architecture
by Nebius
Land and facility
Power
Land and facility
Power
Nebius
Partner
(5) Contracted power, secured by contracted land and power commitments.
6
Cloud: Advancing an open
ecosystem across every layer
of the stack
As enterprises put AI into production, inference is becoming one
of the fastest-growing layers of AI infrastructure, and increasing-ly runs on open models for their control, flexibility, and lower cost
at scale. Token Factory, our managed inference platform, pow-ers this transition:
Production inference workloads more than tripled in Q2 as cus-tomers deployed multiple applications and expanded existing
ones. A growing share are agentic (such as search, coding,
and customer-facing agents), where a single task drives many
model calls, so consumption scales with the complexity
of the work automated, rather than with user count.
Training and inference are converging. We see that inference
and training use cases are not isolated: Those training models
require production inference, while those inferencing open-weight models often evolve toward fine-tuning and post-training
as they optimize for quality, latency, and cost.
• Unlike standalone inference providers, Nebius serves the full
workload lifecycle on one platform that will drive better total
cost of ownership and utilization of compute.
• A growing share of inference demand is generated by train-ing itself — reinforcement learning roll-outs, evaluations,
and synthetic data generation.
• A digital health Token Factory customer was able to move
their AI mental-wellbeing product to a more capable model
and cut worst-case response times by orders of magnitude.
Model performance. Day-one support for the most advanced
models like Kimi K3, GLM-5.2, MiniMax 3 and NVIDIA Nemotron
3 Ultra.
• Nebius delivers both accuracy and speed, validated by inde-pendent benchmarks.
• In the most recent Endpoint Accuracy Index from Artificial
Analysis, Nebius achieved full accuracy parity on GLM-5.2
while delivering among the fastest output speeds of any
provider. The benchmark measures how much of a model’s
accuracy each inference provider preserves.
• The Eigen AI and Clarifai integrations deepened our model
optimization, inference systems, and deployment capabilities.
Agentic infrastructure. Q2 was Tavily’s first full quarter with-in Nebius, pairing Token Factory’s inference with the real-time
web access that agents need for factual accuracy — reasoning
and retrieval on one platform.
• Tavily’s developer community has grown to more than 2.5 mil-lion developers (from 1 million in February) serving Fortune
500 enterprises and powering production agents.
• In Q2, Tavily launched keyless pay-per-search built for
autonomous agent consumption, and achieved ISO 27001
certification for enterprise deployments.
• As model interfaces standardize, differentiation shifts to cost,
latency, reliability, quality in production and to the agentic,
multi-step workloads defining the next phase of enterprise
AI. Token Factory is built for both.
Nebius AI Cloud “Aether 3.6”: Strengthening the core
platform
In Q2 we shipped Nebius AI Cloud 3.6 (“Aether 3.6”), strengthen-ing the developer experience, governance controls, and storage
performance teams need to run AI in production.
As our capacity expands and our customer commitments grow,
our cloud platform needs to scale with it. We are enhancing our
platform to support greater volume as we close larger custom-er contracts. We are also deploying our software stack to new
regions as we bring on more data center sites globally.
As part of this roadmap, we released the following features as part
of Aether 3.6:
• Developer experience. The release introduced Nebius Echo,
our own AI agent for natural-language infrastructure control,
built into Nebius AI Cloud and running on open-source models
served by Token Factory. Echo lets customers manage their
environment in plain language, asking questions grounded
in our documentation, checking live resource status and con-ducting simple operations such as instance creation without
commands or additional setup. In the future, Echo will provide
automated infrastructure diagnostics and handle complex,
multi-step deployments end-to-end.
• Security and governance. New controls support sensi-tive workloads in regulated environments, including a Key
Management Service with customer-managed encryption
keys, Workload Identity Federation for credential-free
authentication, Budgets to help FinOps teams set spend-ing limits, and a Bring Your Own Image option for hardened
base images.
• Storage. Our storage platform saw several enhancements this
quarter. Object storage now delivers 30% higher read speed.
Shared Filesystem is 3x faster for small-file operations
Deepening customer collaboration
At our June 9 Inflection Event, we launched our Customer
Advisory Board, with AMI, Black Forest Labs, Cloudflare,
Cognition, Cohere, Core Automation, Higgsfield, Recraft, Revolut,
and Rhoda. This forum brings CEOs, CTOs, founders, and indus-try leaders together to sharpen our customer feedback loop,
and keep Nebius at the forefront of AI development.
We also opened the Nebius Builder Program in early preview,
which provides credits across Nebius AI Cloud, Token Factory,
Tavily, and Nebius Academy, with engineering office hours
and a builder community. We also launched Nebius Certifications
to give AI cloud professionals verifiable credentials.
Expanding executive talent
In July, we welcomed Lindsey Irvine as our new Chief Marketing
Officer. Lindsey brings deep experience leading global market-ing at high-growth technology companies as a multi-time CMO
at Square, Benchling and MuleSoft, with prior leadership roles
at Salesforce.
7
Other businesses
Avride
Autonomous vehicle operations achieved important goals in Q2
• The AV-capable fleet has nearly tripled since year-end 2025,
exceeding 200 vehicles in May 2026 with more than one mil-lion autonomous miles completed this year.
• Avride has completed over 60,000 commercial rides on Uber
in Dallas.
• The Dallas operating map has doubled since launch, with fur-ther expansion expected.
• The R&D fleet continues to collect mileage and data needed
to advance our AI technology stack and accelerate the tran-sition to No-Vehicle-Operator (NVO) operations.
Robodelivery operations launched in Arlington, Virginia,
and Miami through UberEats; new Chartwells Higher
Education partnership expands campus reach
• Robot deliveries more than tripled year-over-year in Q2
and exceeded 600,000 since inception.
• In Q2, Avride launched robodelivery operations in Arlington,
VA and Miami through UberEats.
• In July, Avride announced a Master Services Agreement with
Chartwells Higher Education, one of the largest on-cam-pus food service providers in the US. Bowling Green State
University is among the first campuses to launch, with more
joining this fall.
• Robodelivery operations are live at Indiana University
Bloomington.
• Momentum is building with vendor partners following a com-petitor’s exit from campus deliveries.
TripleTen
TripleTen, our edtech business, continues to develop
new offerings to meet evolving market opportunities
• TripleTen added AI Systems Engineering as a new program
in Q2, continuing to build out a comprehensive suite of AI
coursework.
• Tested new products in the US and expanded its enter-prise-focused B2B offering, with an emphasis on enterprise
AI capabilities and synergies across the Nebius ecosystem.
• Efficiency improvement and cost reductions have improved
segment financial performance.
8
Equity stakes
In addition to our non-core businesses, we own equity stakes
in both Toloka and ClickHouse, both of which were originally spun
out from Nebius Group.
Toloka
Toloka is a leading provider of scalable, high-quality data essen-tial for LLM, GenAI, and AI agent development.
To capture value across the entire AI development stack, Toloka
recently executed several strategic platform expansions:
• Toloka Arena: An independent suite of benchmarks that eval-uates agentic intelligence on private, non-contaminated tasks
to simulate real-world scenarios involving multi-turn tool use,
live databases, and strict business rules. With over 40 mod-els evaluated, Arena empowers enterprises to make confident
model selections and enables AI labs to license these propri-etary datasets to train and test their own models.
• Self-service platform: A major upgrade to the data-label-ling platform, accelerating time-to-scale for data pipelines
through API automation, agent-assisted setup, built-in qual-ity controls, and newly integrated synthetic data generation.
• Expanded catalog: A significantly broader catalog
of expert-validated datasets across high-demand domains,
including coding, STEM, reasoning, and RL Gyms.
• Physical AI and robotics: Accelerated investment in physical
AI, expanding robotics data offerings beyond egocentric vid-eo to include UMI-style data and teleoperation.
As of Q2’25 we no longer hold voting control of Toloka. However,
we maintain a significant equity stake, and are encouraged
by the growing investor interest in the AI data provider market.
ClickHouse
ClickHouse is an open-source database management system built
for real-time data processing and analytics.
In January 2026, it was reported that ClickHouse raised $400M
in a Series D financing at a valuation of approximately $15B.
Following this financing, Nebius Group continues to own a signif-icant minority equity stake in ClickHouse.
9
Financial update
Nebius Group once again executed against its financial goals. Q2
group revenue of $582.3 million, was up 454% year-over-year,
and up 46% compared to Q1. Year-over-year growth was driven
by capacity scaling and supported by higher pricing and improved
utilization from a year ago.
• Nebius AI cloud revenue was $574.9 million in Q2’26, a 514%
increase year-over-year from Q2’25. Our Nebius AI cloud
Revenue
In USD $ millions Three months ended June 30 Six months ended June 30
2025 2026 Change 2025 2026 Change
Revenues 105.1 582.3 454% 156.0 981.3 529%
business accounted for approximately 98% of total group rev-enue during the quarter.
• Annualized run-rate revenue (ARR)1 of $3.0 billion
as of the end of June was up 598% year-over-year and 56%
from the $1.9 billion reported at the end of March.
In USD $ millions Three months ended June 30 Six months ended June 30
Expense category 2025 2026 Change 2025 2026 Change
Cost of revenues 30.1 133.6 344% 54.8 237.4 333%
as a percentage of revenues 29% 23% 35% 24%
Product development 42.8 191.0 346% 79.3 258.4 226%
as a percentage of revenues 41% 33% 51% 26%
Sales, general and administrative 68.2 173.9 155% 129.1 317.7 146%
as a percentage of revenues 65% 30% 83% 32%
Depreciation and amortization 75.2 259.7 245% 124.3 471.7 279%
as a percentage of revenues 72% 45% 80% 48%
Total operating costs and expenses 216.3 758.2 251% 387.5 1,285.2 232%
as a percentage of revenues 206% 130% 248% 131%
Operating expense
10
Cost of revenue was $133.6 million in Q2’26, representing
an increase of 344% compared to the same period in 2025.
• The increase was due to the expansion of our Nebius AI cloud
business, with expenses incurred for co-location and operat-ing lease agreements as well as hiring to support our growing
operations.
• As a percentage of revenue, cost of revenue was 23%
in Q2’26, down from 29% in Q2’25, primarily reflecting oper-ating leverage as we scaled capacity.
Product development expenses were $191.0 million in Q2’26,
representing an increase of 346% compared to the same peri-od in 2025.
• The increase in product development expenses was primari-ly driven by hiring in our engineering and development teams
to build and enhance our product offerings, as well as $115.9
million of non-recurring expense pertaining to share-based
compensation expense and post-combination expenses rec-ognized in connection with the acquisition of Eigen AI.
• As a percentage of revenue, product development expenses
decreased to 33% from 41% in the prior year.
Sales, general and administrative expenses (“SG&A”) were
$173.9 million in Q2’26, representing an increase of 155% com-pared to $68.2 million in the same period in 2025.
• The increase was primarily driven by hiring to support
the growth of our business as well as consulting, legal
and other professional fees primarily incurred in connection
with acquisition-related activities and financing transactions.
• SG&A as a percentage of revenues declined to 30% from 65%
in Q2’25.
Depreciation and amortization expenses (“D&A”) were $259.7
million in Q2’26, representing an increase of 245% compared
to the same period in 2025.
• The primary driver of the dollar increase in D&A expenses was
the continued investments in GPU-related capital expendi-tures and related data center hardware for the Nebius AI
cloud business.
• We use a five-year useful life for our server and network
equipment based on usage patterns and current utilization
commitments, up from the four-year useful life used prior
to 2026.
• D&A as a percentage of revenue declined to 45% from 72%.
Adjusted EBITDA
Group adjusted EBITDA of $236.2 million in Q2
improved by $257.2 million year over year.
• The continued improvement in Adjusted EBITDA was driven
primarily by the strong growth in our Nebius AI cloud busi-ness, which generated adjusted EBITDA of $285.7 million
and adjusted EBITDA margin of 49.7%.
• We improved group adjusted EBITDA from Q1’26 by $106.7
million from $129.5 million to $236.2 million.
In USD $ millions Three months ended June 30 Six months ended June 30
2025 2026 Change 2025 2026 Change
Adjusted EBITDA / (loss) (21.0) 236.2 n/m (74.7) 365.7 n/m
as a percentage of revenues -20% 41% -48% 37%
Capital expenditures
In Q2’26, capital expenditures were approximately $5.7 billion, pri-marily driven by purchases of GPUs and GPU-related hardware,
and our data center expansion activities.
Capital requirements
We will continue to invest in capex throughout the year, and intend
to use a diversified range of funding sources including:
Cash generated from operations and upfront customer payments.
• We expect to receive over $9 billion in customer prepayments
in 2026.
• Upfront payments give us visibility into future cash flows
and flexibility on timing of additional financing.
We may also enter into debt financing transactions or access
the debt capital markets.
• We are actively progressing potential debt transactions,
including asset-backed financing and corporate-level debt.
• In July we entered into our first secured debt financing,
raising $775 million priced at SOFR + 2.50%. The vehicle
is backed by deployed GPU infrastructure and contracted
cash flows from an agreement with an investment-grade
customer and demonstrates our ability to fund growth
on attractive terms. We have more than $40 billion of cus-tomer commitments.
• We plan to continue tapping into these financing options.
Our financing options include our at-the-market (ATM) program,
which we began using this quarter.
11
• Through June 30, 2026, we sold 12.7 million Class A shares
at a weighted-average price of $223.6 per share, generating
gross proceeds of approximately $2.8 billion.
• As of June 30, 12.3 million shares remained available under
the program.
And we may also evaluate other financing options, and will ulti-mately pursue the vehicles that best serve the long-term interests
of the business.
Guidance
The company will share a detailed view of guidance on its earn-ings call and webcast.
Earnings webcast
Nebius Group will host a conference call and earnings webcast
at 5:00 a.m. Pacific time/8:00 a.m. Eastern time/2:00 p.m. Central
European time on August 12, 2026, to discuss these financial
results. To register to participate in the conference call, or to listen
to the live audio webcast, please visit Nebius’s Investor Relations
website at group.nebius.com/investor-hub.
A replay will be available on the same website following the call.
Forward-looking statements
This document contains forward-looking statements that involve
risks and uncertainties. All statements contained or implied other
than statements of historical facts, including, without limitation,
statements regarding our business plans, market opportuni-ties, capacity buildout plans, capital expenditure requirements,
financing requirements and projected financial performance, are
forward-looking statements. In some cases, these forward-look-ing statements can be identified by words or phrases such
as “may,” “will,” “expect,” “anticipate,” “aim,” “estimate,” “intend,”
“plan,” “believe,” “potential,” “continue,” “is/are likely to” or other
similar expressions. In addition, these forward-looking statements
reflect our current views with respect to future events and are not
a guarantee of future performance. Actual results may differ mate-rially from the results predicted or implied by such statements,
and our reported results should not be considered as an indica-tion of future performance. The potential risks and uncertainties
that could cause actual results to differ from the results pre-dicted or implied by such statements include our ability to:
obtain sufficient financing and manage our liquidity and capi-tal resources to support our operations and growth; successfully
identify, develop and bring online additional data center capacity
on a timely and cost-effective basis, including securing suita-ble sites and access to power; implement and maintain effective
internal control over financial reporting; manage supply chain risks
and secure required equipment, hardware, materials and services
on acceptable terms; compete effectively in a dynamic and com-petitive market while generating sustained customer demand;
and manage dependence on key vendors and adapt to techno-logical change.
Many of these risks and uncertainties depend on the actions
of third parties and are largely outside of our control. Our actu-al results of operations may also differ materially from those
stated in or implied by such forward-looking statements
as a result of a variety of factors, including those described
under the captions “Risk Factors” and “Operating and Financial
Review and Prospects” in our Annual Report on Form 20-F for
the year ended December 31, 2025 filed with the U.S. Securities
and Exchange Commission (“SEC”) on April 30, 2026, which
is available on our investor relations website at https://group.nebi-us.com and on the SEC website at www.sec.gov. All information
in this document is as of the date hereof, and the Company under-takes no duty to update this information unless required by law.
In addition, statements that “we believe” and similar state-ments reflect our beliefs and opinions on the relevant subject.
These statements are based upon information available to us
as of the date of this document, and while we believe such infor-mation forms a reasonable basis for such statements, such
information may be limited or incomplete, and our statements
should not be read to indicate that we have conducted an exhaus-tive inquiry into, or review of, all potentially available relevant
information. These statements are inherently uncertain, and inves-tors are cautioned not to unduly rely upon these statements.
We operate in an evolving environment. New risks emerge from
time to time, and it is not possible for our management to predict
all risks, nor can we assess the effect of all factors on our busi-ness or the extent to which any factor, or combination of factors,
may cause actual results to differ materially from those con-tained in any forward looking statements. You should not rely
upon forward looking statements as predictions of future events.
We undertake no obligation to update or revise any forward-look-ing statements, whether as a result of new information, future
events or otherwise.
Disclaimer
Links to third-party websites are provided for informational pur-poses only; Nebius is not responsible for the content contained
on or accessible through the linked sites.
Use of Non-GAAP financial
measures
To supplement the financial information prepared and presented
in accordance with U.S. GAAP, we present the following non-GAAP financial measures: Adjusted EBITDA/(loss) and Adjusted
net income/(loss). The presentation of these financial measures
is not intended to be considered in isolation or as a substitute for,
or superior to, the financial information prepared and present-ed in accordance with U.S. GAAP. For more information on these
non-GAAP financial measures, please see the tables captioned
“Reconciliations of non-GAAP financial measures to the most
directly comparable U.S. GAAP measures”, included following
the accompanying financial tables. We define the various non-GAAP financial measures we use as follows:
12
• Adjusted EBITDA / (loss) means U.S. GAAP net income/
(loss) from continuing operations before (1) depreciation
and amortization, (2) share-based compensation expense,
(3) acquisition and other corporate transaction-related costs,
(4) interest income, (5) interest expense, (6) income/(loss)
from equity method investments, (7) gain from revaluation
of investments in equity securities, (8) other income/(loss),
net, (9) income tax expense/(benefit).
• Adjusted net income / (loss) means U.S. GAAP net income/
(loss) from continuing operations before (1) share-based
compensation expense, (2) acquisition and other corporate
transaction-related costs, (3) amortization of debt discount
and issuance costs, net of interest expense capitalized, (4)
foreign exchange gains/(losses) and (5) gain from revalua-tion of investments in equity securities. Tax effects related
to the listed adjustments are excluded from adjusted net
income.
These non-GAAP financial measures are used by management
for evaluating financial performance as well as decision-making.
Management believes that these metrics reflect the organic, core
operating performance of the company, and therefore are useful
to analysts and investors in providing supplemental information
that helps them understand, model and forecast the evolution
of our operating business.
Although our management uses these non-GAAP financial meas-ures for operational decision-making and considers these financial
measures to be useful for analysts and investors, we recognize
that there are a number of limitations related to such measures.
In particular, it should be noted that several of these measures
exclude some recurring costs, particularly share-based compen-sation. In addition, the components of the costs that we exclude
in our calculation of the measures described above may differ
from the components that our peer companies exclude when they
report their results of operations.
Below we describe why we make particular adjustments to cer-tain U.S. GAAP financial measures:
Net income/(loss) from discontinued operations
We present Adjusted EBITDA/(loss) and Adjusted net income/(loss)
excluding any effects of our discontinued operations.
Information on our discontinued operations is disclosed in our
Annual Report on Form 20-F for the year ended December 31,
2025 filed with the U.S. Securities and Exchange Commission
(“SEC”) on April 30, 2026.
Share-based compensation expense
Share-based compensation is a significant expense item
and an important part of our compensation and incentive pro-grams. As it is highly dependent on our share price at the time
of equity award grants, we believe that it is useful for inves-tors and analysts to see certain financial measures excluding
the impact of these charges in order to obtain a clearer picture
of our operating performance.
Foreign exchange gains/(losses)
The functional currency of Nebius Group N.V. is the United States
Dollar, which is also the Group’s reporting currency. Foreign
exchange gain/(loss) dynamics reflect changes in the U.S. dol-lar value of monetary assets and liabilities that are denominated
in other currencies, as well as changes in the functional curren-cies of foreign subsidiaries’ monetary assets and liabilities that
are denominated in currencies different from their respective local
currencies. Because foreign exchange fluctuations are outside
of our operational control, we believe that it is useful to present
Adjusted EBITDA/(loss), adjusted net income/(loss) and relat-ed margin measures excluding these effects, in order to provide
greater clarity regarding our operating performance.
Acquisition and other corporate transaction-related costs
We believe that it is useful to present Adjusted net income/(loss),
Adjusted EBITDA/(loss) and related margin measures exclud-ing impacts not related to our operating activities. Adjusted
net income/(loss) and Adjusted EBITDA/(loss) exclude certain
expenses related to M&A activities and other expenses related
to corporate transactions.
Amortization of debt discount and issuance costs, net of inter-est expense capitalized
We also adjust net income/(loss) for interest expense repre-senting amortization of the debt discount and issuance costs
related to our convertible senior notes, net of interest expense
capitalized into cost of our property and equipment. Debt dis-count represents the accretion of the nominal amount of notes
payable at maturity, unless the relevant notes have been earli-er repurchased, redeemed or converted in accordance with their
terms. We adjust net income/(loss) for the interest expense rec-ognized from amortization of the debt discount and issuance
costs due to the significantly different timing of payment in rela-tion to the operating results.
Contact investor relations:
[email protected]