Skip to content
    NBIS
    Earnings call· Dec 2024(Q4 FY24)

    Nebius Group N.V. NBIS

    Feb 20, 2025 Source

    Executive summary

    Nebius Group Q4 FY24 — Aggressive AI Infrastructure Build-Out and Strong ARR Momentum

    Nebius Group, now a publicly traded AI specialized cloud, is aggressively expanding its global AI infrastructure and sales capabilities. The company is focused on its full-stack AI technology, significant capital investment, and strengthening its corporate structure to capitalize on the vast AI market opportunity, despite Q4 timing issues and initial sales ramp-up challenges.

    Highlights

    5
    • Resumed public trading and raised $700 million in an oversubscribed offering, attracting investors like NVIDIA.

    • Expanded data center footprint and GPU deployments, including new H200 clusters in Iceland and Kansas City, bringing total capacity to 30,000 GPUs.

    • Achieved March annualized run rate revenue of at least $220 million, with a projected December 2025 ARR target of $750 million to $1 billion.

    • Subsidiary Toloka grew full year revenue by 140%, and TripleTen's revenue was up 250% with 14,000 students.

    • Avride secured partnerships with Uber and Grubhub, deploying 100 robots for 1,000 daily deliveries at Ohio State University.

    Concerns

    4
    • Q4 revenue dynamics impacted by longer deal lead times and increased customer selectivity, leading to anticipated churn not being replaced.

    • Sales hires require 6 months or more to become fully productive, impacting immediate revenue flow.

    • Experienced price pressure on H100 GPUs in Q4, though mitigated by cost structure and shift to Blackwells.

    • EBITDA is expected to remain negative for the full year 2025, with breakeven anticipated at some point during the year.

    Guidance & targets

    12
    CategoryTargetConfidence
    Annualized Run Rate Revenue
    at least $220 million
    medium materiality
    High
    Annualized Run Rate Revenue
    $750 million to $1 billion
    high materiality
    High
    Revenue
    $500 million to $700 million
    high materiality
    High
    EBITDA
    negative for the full year, but pass breakeven at some point during the year
    high materiality
    Medium
    Data Center Capacity
    100 megawatt
    high materiality
    High
    Finland Data Center Expansion
    first part of the delivery to be completed
    medium materiality
    High
    Finland Data Center Expansion
    remaining capacity will come online
    medium materiality
    High
    Kansas City Cluster Availability
    up and running
    medium materiality
    High
    Iceland H200 Cluster Availability
    available for customers
    medium materiality
    High
    Blackwell B200 GPU Availability
    available in Q2
    high materiality
    High
    Blackwell GB200 GPU Availability
    available in the platform in Q3
    high materiality
    High
    Blackwell GB300 GPU Deployment
    plan to deploy
    high materiality
    High

    Segment performance

    3
    SegmentRevenueYoYQoQMargin
    Toloka (Data Training Platform)
    Grew full year revenue by 140% and diversified its customer base by adding several leading AI labs. Completed transition to a new platform tuned for complex GenAI tasks.
    140%
    TripleTen (EdTech Business)
    Doubled new student additions year-over-year in 2024 and maintained its position as a leading IT boot camp. Revenue was up 250%.
    New student additions: doubled year-over-yearTotal students: 14,000
    250%
    Avride (Autonomous Technology Platform)
    Signed contracts with Uber (as one of three autonomous technology partners) and Grubhub. Uber Eats deployed robots in Austin, Dallas, and Jersey City. Grubhub deployed 100 robots at Ohio State University, achieving 1,000 daily deliveries. Received certification in Japan.
    Robots deployed (Ohio State University): 100Daily deliveries (Ohio State University): 1,000

    Operational metrics

    16
    Capital raised in Q4
    $700 million
    Q4 FY24

    Raised in an oversubscribed offering.

    Cash and investments balance
    $2.4 billion
    end of last year

    Including the $700 million raised in Q4.

    March annualized run rate revenue
    $220 millionat least
    March

    Based on contracts already in place.

    Total current GPU capacity
    30,000
    end of March

    With additions of Iceland and Kansas City clusters.

    Deployed H200 GPUs
    20,000
    end of March

    Portion of total current GPU capacity.

    Potential scalable data center capacity
    300more than
    future

    All new facilities are potentially scalable to this total.

    Toloka full year revenue growth
    140%YoY
    FY24

    Growth versus 2023.

    TripleTen new student additions growth
    doubledYoY
    FY24

    New student additions year-over-year.

    TripleTen revenue growth
    250%
    FY24

    Revenue growth for the year.

    TripleTen total students
    14,000
    end of FY24

    Record high for the business.

    H200 GPU payback period
    2.5 to 3-year range
    current

    Based on company calculations for Hopper generation GPUs.

    Dutch tax payment
    $180 million
    February

    Paid to authorities to settle remaining potential liability.

    Dutch tax liability reduction
    $220 million
    current

    Amount freed up from the previously guided $400 million for expansion plans.

    Avride daily deliveries
    1,000
    daily

    Achieved in the first week of deployment for Grubhub partnership.

    Avride robots deployed
    100
    current

    Deployed for Grubhub food delivery service.

    Sales team productivity ramp-up
    6 months or more
    ongoing

    Typical time for new sales hires to become fully productive.

    Industry KPIs

    9
    MetricValueDetails
    Capacity CAPEX100megawatts
    Revenue growth$500 million to $700 millionUSD
    Arr net new arr$750 million to $1 billionUSD
    Bookings billingspositive demand
    Pricing model mixdiverse
    Customer account counthundredscustomers
    Large deal new logo metricsmultiple thousandsGPUs
    Operating FCF margin rule of 40negative
    Ai product adoption monetizationlaunched

    Orderbook & backlog

    1
    Blackwell presalespositive demandend of 2024

    Working through initial customer demand; discussing reserve contracts from 1 year and higher; pre-commitments in multiple thousands of GPUs.

    Product announcements

    6
    ProductTypeDetails
    AI cloud platformlaunch
    Inference as a Service platform (AI Studio)launch
    H200 cluster in Icelandlaunch
    Blackwell B200 GPUslaunch
    Blackwell GB200 GPUslaunch
    Blackwell GB300 GPUslaunch

    Deals & partnerships

    3
    NVIDIA, Accel, OrbisEntered capital structure via oversubscribed $700 million offering$700 million

    A deal that saw NVIDIA, Accel, Orbis, and others enter Nebius Group's capital structure.

    UberAvride autonomous technology partner

    Avride signed a contract with Uber as one of just three autonomous technology partners in the U.S. Uber Eats has already deployed Avride robots for food delivery in Austin, Dallas, and Jersey City.

    GrubhubAvride robots for food deliveries on college campuses

    Avride signed a contract with Grubhub to use Avride robots for food deliveries for college campuses across the United States, starting with Ohio State University.

    Capital programs

    2
    Finland data center expansionunderway

    Benefit: tripling the capacity

    Construction is going according to schedule, with the first part of delivery in Q3 and remaining capacity online around year-end.

    New U.S. build-to-suit facilityannounced

    Benefit: equipped for Blackwell deployments

    Signed a significant new build-to-suit facility in the U.S. specifically for Blackwell deployments; details to be updated soon.

    Risks & headwinds

    5
    Longer deal lead timesQ4 FY24

    mostly timing issues

    Mitigation: Customers became much more selective and wanted to do more in-depth proof-of-concept testing. Sales hires take 6 months or more to become fully productive.

    Customer churnQ4 FY24

    couple of large customers completed their engagement with us in Q4

    Mitigation: Anticipated, but not replaced due to longer lead times for new deals. Expected in dynamic AI market.

    H100 price pressureQ4 FY24

    some price pressure in Q4

    Mitigation: Mitigated by strong cost structure and shift to higher-priced Blackwell generations.

    Negative EBITDAFY25

    EBITDA will remain negative for the full year

    Mitigation: Expected to pass breakeven at some point during the year as Blackwells come online and sales ramp up.

    Dutch tax liabilityQ1 FY25

    $180 million paid

    Mitigation: Settled with authorities, freeing up $220 million from the previously guided $400 million for expansion plans.

    What to watch in Q1 FY25

    5

    Kansas City cluster availability

    end of this quarter
    Currentin deployment phase
    Targetup and running

    Why it matters

    Marks first U.S. capacity, crucial for U.S. customer base expansion.

    our first U.S. cluster in Kansas City, which is now in deployment phase and should be up and running by the end of this quarter.

    Q&A highlights

    10

    Can you discuss the broader plans for capacity expansion over 2025?

    Andrey Korolenko detailed the expansion of the Finland data center, the new facility in France, and the Kansas City cluster. For 2025, a new H200 cluster in Iceland will be available in March, bringing total capacity to 30,000 GPUs. A new build-to-suit facility in the U.S. for Blackwell deployments was signed, and the company is on track for 100 MW in operations by year-end, with potential scalability to over 300 MW.

    Our guidance calls for 100 megawatt in operations by end of this year, and we are on track to deliver that. Also, I would like to highlight that all these new facilities are potentially scalable to a total of more than 300 megawatts as and when we are ready to do so.

    asked by Tom Blackwell · answered by Andrey Korolenko

    3 min read7 chapters

    Detailed Narrative

    01

    Strategic Focus on AI Infrastructure and Full-Stack Advantage

    Nebius Group is positioning itself as a leading AI infrastructure company, emphasizing its proven track record in building and operating efficient data centers with stable power and GPU-based AI compute. The company highlights its unique full-stack approach, which includes proprietary hardware development, an AI-specific cloud platform, and value-added services. This strategy aims to provide a significant cost advantage, flexibility, and reliability, enabling software-like margins and extending the economic life of its GPUs.

    02

    Q4 Achievements and Capital Infusion

    In Q4 FY24, Nebius resumed public trading as the first AI specialized cloud, a significant milestone achieved faster than anticipated. The company successfully raised $700 million in an oversubscribed offering in December, attracting high-quality investors such as NVIDIA, Accel, and Orbis. This capital raise, combined with its public listing, provides Nebius with enhanced access to efficient financing options to fuel its aggressive growth plans.

    03

    Aggressive Capacity Expansion and GPU Deployment

    Nebius is rapidly expanding its global data center footprint. In Q4, it expanded its data center in Finland, added a facility in France, and announced its first U.S. cluster in Kansas City. For 2025, new H200 clusters in Iceland and Kansas City will bring total current capacity to approximately 30,000 GPUs by March. The company is on track to have 100 megawatts in operations by year-end 2025, with potential scalability to over 300 megawatts, and is preparing for the deployment of next-generation Blackwell GPUs (B200, GB200, GB300) throughout the year.

    04

    Sales Function Ramp-up and Customer Dynamics

    The company has been actively building out its global sales and marketing teams, particularly in the U.S. While Q4 experienced longer deal lead times and some customer churn due to increased selectivity and platform migration, Nebius reports strong momentum entering 2025, with March annualized run rate revenue of at least $220 million. The customer base is diversified, serving GenAI startups and large enterprises with a mix of short-term on-demand and anticipated longer-term contracts for Blackwell deployments.

    05

    Performance of Subsidiary Businesses

    Nebius's business units demonstrated strong performance. Toloka, its data training platform, grew full-year revenue by 140% and expanded its client portfolio to include leading AI labs. TripleTen, the EdTech business, doubled its new student additions year-over-year in 2024 and saw revenue increase by 250%, ending the year with 14,000 students. Avride, the autonomous technology platform, secured partnerships with Uber and Grubhub, deploying robots for food delivery in multiple U.S. cities and receiving certification in Japan.

    06

    Capital Allocation and Future Funding

    Nebius ended last year with $2.4 billion in cash. While significant capital is required for 2025 expansion plans, management is confident in its ability to fund growth through efficient financing options and strong investor interest. The payback period for Hopper generation GPUs is estimated at 2.5 to 3 years, with expectations for faster returns on Blackwells due to cost efficiencies from proprietary hardware and high-margin software services.

    07

    Regulatory and Geographic Considerations

    Nebius, headquartered in the Netherlands, is actively expanding its physical and operational presence in the U.S. The company anticipates qualifying for 'universal verified end user' status under U.S. AI regulations and does not foresee material negative impacts from potential tariffs, given its global footprint and U.S. expansion focus. Management believes its tech knowledge and know-how position it well to contribute to large-scale AI infrastructure projects like the 'Stargate Project'.

    AI-generated summary of the company’s earnings call. Not investment advice.