Skip to content
    NVDA
    Earnings call· Apr 2025(Q1 FY26)

    NVIDIA Q1 FY26 earnings call NVDA

    May 28, 2025 Source

    Executive summary

    NVIDIA Q1 FY26 — Strong Blackwell Ramp and AI Inference Demand Offset China Headwinds

    NVIDIA delivered a strong quarter, driven by the rapid Blackwell ramp and surging AI inference demand, particularly for reasoning models. The company is expanding its AI factory deployments globally, with new growth engines in sovereign AI, enterprise AI, and industrial AI. However, new export controls on H20 significantly impacted China Data Center revenue, leading to a substantial charge and expected future revenue loss in the region.

    Highlights

    5

    • Total revenue of $44 billion, up 69% year-over-year, exceeding outlook.

    • Data Center revenue of $39 billion, grew 73% year-on-year, driven by Blackwell ramp.

    • Blackwell contributed nearly 70% of Data Center compute revenue in Q1 FY26.

    • Gaming revenue was a record $3.8 billion, increasing 48% sequentially and 42% year-on-year.

    • Returned a record $14.3 billion to shareholders in Q1 FY26 through share repurchases and cash dividends.

    Concerns

    4

    • Recognized a $4.5 billion charge for H20 inventory and purchase obligations due to new export controls.

    • Unable to ship $2.5 billion in H20 revenue in Q1 FY26 due to new export controls.

    • Expect a meaningful decrease in China Data Center revenue for Q2 FY26, reflecting a loss of approximately $8 billion in H20 revenue.

    • China Data Center revenue was down sequentially and below expectations due to H20 export licensing controls.

    Guidance & targets

    10
    CategoryTargetConfidence
    Total revenue
    $45 billion, plus or minus 2%
    high materiality
    High
    H20 revenue loss
    approximately $8 billion
    high materiality
    High
    GAAP gross margins
    71.8%
    medium materiality
    High
    Non-GAAP gross margins
    72%
    high materiality
    High
    Non-GAAP gross margins
    mid-70s range
    high materiality
    Medium
    GAAP operating expenses
    approximately $5.7 billion
    medium materiality
    High
    Non-GAAP operating expenses
    approximately $4 billion
    medium materiality
    High
    Full year FY26 operating expense growth
    mid-30% range
    medium materiality
    High
    GAAP and Non-GAAP other income and expenses
    income of approximately $450 million
    low materiality
    High
    GAAP and Non-GAAP tax rates
    16.5%, plus or minus 1%
    low materiality
    High

    Segment performance

    4
    SegmentRevenueYoYQoQMargin
    Data Center
    Strong growth driven by the Blackwell ramp and transition from Hopper. Significant impact from H20 export controls in China.
    Blackwell contribution to Data Center compute revenue: nearly 70%H20 revenue recognized in Q1: $4.6 billionH20 revenue unable to ship in Q1: $2.5 billionChina as percentage of Data Center revenue: slightly below expectations, down sequentially
    		$39 billion73%——
    Gaming
    Record revenue driven by strong adoption of Blackwell architecture by gamers, creators, and AI enthusiasts. Improved supply and availability.
    		$3.8 billion42%48%—
    Pro Visualization
    Tariff-related uncertainty temporarily impacted Q1. Strong demand for AI workstations expected to drive sequential growth in Q2.
    		$509 million19%flat—
    Automotive
    Year-on-year growth driven by ramp of self-driving solutions and robust end demand for NEVs.
    		$567 million72%-1%—

    Operational metrics

    34
    Non-GAAP gross margins
    61%
    Q1 FY26

    Excluding the $4.5 billion H20 charge, non-GAAP gross margins would have been 71.3%.

    GAAP operating expenses
    up 7%sequentially
    Q1 FY26

    Reflecting higher compensation and employee growth.

    Non-GAAP operating expenses
    up 6%sequentially
    Q1 FY26

    Reflecting higher compensation and employee growth.

    Capital returned to shareholders
    $14.3 billionrecord
    Q1 FY26

    A key element of the capital allocation strategy.

    AI factory deployments
    nearly 100twofold increase year-over-year
    Q1 FY26

    Pace and scale of AI factory deployments are accelerating across industries and geographies.

    Microsoft processed tokens
    over 100 trillionfivefold increase year-over-year
    Q1 FY26

    Representative of strong demand for Azure AI foundry and other AI services.

    NVIDIA Dynamo AI inference throughput
    30x
    Q1 FY26

    Turbocharges AI inference throughput for new reasoning models.

    Capital One agentic chatbot latency reduction
    5x
    Q1 FY26

    Reduced agentic chatbot latency with Dynamo.

    MLPerf inference throughput (GB200 NVL72)
    up to 30x highercompared to 8-GPU H200
    Q1 FY26

    Achieved through combination of performance per GPU and more GPUs connected via NVLink.

    Blackwell performance improvement (software optimizations)
    1.5x
    last month

    Software optimizations have already improved Blackwell's performance.

    Hopper inference performance increase
    4x
    over 2 years

    Benefit of NVIDIA's programmable CUDA architecture and rich ecosystem.

    Llama Nemotron accuracy boost
    20%
    Q1 FY26

    Post-training enhancements yielded accuracy boost for open reasoning models.

    Llama Nemotron inference speed increase
    5x
    Q1 FY26

    Post-training enhancements yielded inference speed increase for open reasoning models.

    Cisco model accuracy increase (NeMo)
    40%
    Q1 FY26

    Using NeMo, Cisco increased model accuracy.

    Cisco response time improvement (NeMo)
    10x
    Q1 FY26

    Using NeMo, Cisco improved response time.

    NASDAQ AI platform accuracy improvement (NeMo)
    30%
    Q1 FY26

    NASDAQ realized improvement in accuracy and response time in its AI platform's search capabilities.

    NASDAQ AI platform response time improvement (NeMo)
    30%
    Q1 FY26

    NASDAQ realized improvement in accuracy and response time in its AI platform's search capabilities.

    Shell custom LLM accuracy increase (NeMo)
    30%
    Q1 FY26

    Shell's custom LLM achieved accuracy increase when trained with NVIDIA NeMo.

    NeMo model training time acceleration
    20%compared to other frameworks
    Q1 FY26

    NeMo's parallelism techniques accelerated model training time.

    CrowdStrike detection triage speed
    2x faster
    Q1 FY26

    CrowdStrike realizing faster detection triage with 50% less compute cost using NVIDIA's AI security stack.

    CrowdStrike compute cost reduction
    50% less
    Q1 FY26

    CrowdStrike realizing faster detection triage with 50% less compute cost using NVIDIA's AI security stack.

    NVLink bandwidth
    14xbandwidth of PCIe Gen 5
    Q1 FY26

    NVLink offers significantly higher bandwidth.

    NVLink 72 bandwidth per rack
    130 terabytes per secondequivalent to world's peak Internet traffic
    Q1 FY26

    NVLink 72 carries massive bandwidth in a single rack.

    NVLink shipments
    exceeding $1 billion
    Q1 FY26

    NVLink is a new growth vector and off to a great start.

    Spectrum-X annualized revenue
    over $8 billion
    Q1 FY26

    Spectrum-X posted strong sequential and year-on-year growth.

    Spectrum-X and Quantum-X power efficiency
    3.5x
    Q1 FY26

    These platforms will enable next-level AI factory scaling through increased power efficiency.

    Spectrum-X and Quantum-X network resiliency
    10x
    Q1 FY26

    These platforms will enable next-level AI factory scaling through increased network resiliency.

    Spectrum-X and Quantum-X time to market acceleration
    1.3x
    Q1 FY26

    These platforms will accelerate customer time to market.

    Singapore billed revenue
    nearly 20%
    Q1 FY26

    Many large customers use Singapore for centralized invoicing.

    GeForce user installed base
    100 million
    Q1 FY26

    Represents the largest footprint for PC developers.

    Nintendo Switch consoles shipped
    over 150 million
    to date

    Making it one of the most successful gaming systems in history.

    TSMC fab design time savings (Omniverse)
    months
    Q1 FY26

    TSMC saves months in work by designing fabs virtually using Omniverse.

    Foxconn thermal simulations acceleration (Omniverse)
    150x
    Q1 FY26

    Foxconn accelerates thermal simulations using Omniverse.

    Pegatron assembly line defect rate reduction (Omniverse)
    67%
    Q1 FY26

    Pegatron reduced assembly line defect rates using Omniverse.

    Industry KPIs

    8
    MetricValueDetails
    Lead timesSupply chain quite busy
    Ai data center revenue$39 billionUSD
    Fab capacity utilizationProcess qualification underway
    Bookings net order intakeMore orders today
    Design wins socket pipelineTens of thousands of Blackwell GPUs deployed; ramping to hundreds of thousands of GB200sGPUs
    Inventory channel inventory$4.5 billionUSD
    Node platform ramp scheduleBlackwell ramp fastest in company's history
    End market segment revenue mixData Center: $39B; Gaming: $3.8B; Pro Visualization: $509M; Automotive: $567MUSD

    Product announcements

    14
    ProductTypeDetails
    GB200 NVLlaunch
    Blackwell Ultraroadmap
    GB300 systemslaunch
    Llama Nemotron family of open reasoning modelslaunch
    NVIDIA NeMo microserviceslaunch
    NVLink Fusionlaunch
    Spectrum-X and Quantum-X silicon photonics switcheslaunch
    GeForce RTX 5060 and 5060 Tilaunch
    Nintendo Switch 2launch
    DGX Sparklaunch
    DGX Stationlaunch
    Isaac GR00T N1launch
    NVIDIA Cosmos World Foundation modelslaunch
    NVIDIA Isaac platform for health care simulationlaunch

    Deals & partnerships

    26
    MicrosoftDeployment of Blackwell GPUs for AI services

    Microsoft has already deployed tens of thousands of Blackwell GPUs and is expected to ramp to hundreds of thousands of GB200s with OpenAI as one of its key customers.

    OpenAIKey customer for Microsoft's GB200 deployments

    OpenAI is one of Microsoft's key customers for the ramp to hundreds of thousands of GB200s.

    AT&T, BYD, Capital One, Foxconn, MediaTek, TelenorLeveraging NVIDIA's full-stack architecture for AI factory deployments

    Industry leaders like AT&T, BYD, Capital One, Foxconn, MediaTek, and Telenor are underpinning AI factory deployments with NVIDIA's full-stack architecture.

    Saudi ArabiaStrategic sovereign cloud project

    Announced a 500-megawatt AI infrastructure project in Saudi Arabia.

    U.A.E.Strategic sovereign cloud project

    Announced a 5-gigawatt AI campus in the U.A.E.

    PerplexityLLM provider adopting NVIDIA technology

    Developer engagements increased, with adoption ranging from LLM providers such as Perplexity.

    Accenture, Cadence, Deloitte, MicrosoftTransforming work with NVIDIA's reasoning models

    Leading platform companies are transforming work with NVIDIA's reasoning models.

    CiscoIncreased model accuracy and improved response time in code assistant using NeMo

    With NeMo, Cisco increased model accuracy by 40% and improved response time by 10x in its code assistant.

    NASDAQRealized improvement in accuracy and response time in AI platform's search capabilities using NeMo

    NASDAQ realized a 30% improvement in accuracy and response time in its AI platform's search capabilities using NeMo.

    ShellAchieved accuracy increase for custom LLM when trained with NeMo

    Shell's custom LLM achieved a 30% increase in accuracy when trained with NVIDIA NeMo.

    Yum! BrandsBringing NVIDIA AI to restaurants to streamline operations

    Partnership with Yum! Brands to bring NVIDIA AI to 500 of its restaurants this year and expanding to 61,000 restaurants over time to streamline order-taking, optimize operations and enhance service.

    Check Point, CrowdStrike, Palo Alto NetworksUsing NVIDIA's AI security and software stack for cybersecurity workflows

    Leading companies are using NVIDIA's AI security and software stack to build, optimize and secure agentic workflows.

    CoreWeave, Microsoft Azure, Oracle Cloud, xAIAdoption of Spectrum-X

    Adoption of Spectrum-X is widespread across major CSPs and consumer Internet companies.

    Google Cloud, MetaNew customers for Spectrum-X

    Google Cloud and Meta were added to the growing list of Spectrum-X customers this quarter.

    MediaTek, Marvell, Alchip Technologies, Astera LabsASIC providers leveraging NVLink Fusion

    Enabling key partners, including ASIC providers, to leverage NVLink Fusion to connect respective ecosystems.

    Fujitsu, QualcommCPU suppliers leveraging NVLink Fusion

    Enabling key partners, including CPU suppliers, to leverage NVLink Fusion to connect respective ecosystems.

    NintendoLeveraging NVIDIA's neural rendering and AI technologies for Switch 2

    The recently unveiled Nintendo Switch 2 leverages NVIDIA's neural rendering and AI technologies, including next-generation custom RTX GPUs with DLSS technology.

    Databricks, SAP, Schneider ElectricDeepened Omniverse integration and adoption

    Deepened Omniverse's integration and adoption into some of the world's leading software platforms.

    KION Group, Pegatron, AccentureLeveraging Omniverse Blueprint for robotic fleet management

    New Omniverse Blueprint such as Mega for at-scale robotic fleet management are being leveraged by these companies to enhance industrial operations.

    TSMC, Quanta, Foxconn, PegatronUsing Omniverse for manufacturing processes

    Showcased Omniverse's traction with technology manufacturing leaders, demonstrating significant efficiency gains.

    GMBuilding next-gen vehicles, factories, and robots using NVIDIA AI

    Partnering with GM to build the next-gen vehicles, factories and robots using NVIDIA AI, simulation and accelerated computing.

    Mercedes-BenzFull-stack solution for self-driving

    Now in production with full-stack solution for Mercedes-Benz starting with the new CLA, hitting roads in the next few months.

    1X, Agility Robotics, Figure AI, Uber, WaabiIntegrating Cosmos into operations for synthetic data generation

    Leading companies have begun integrating Cosmos into their operations for synthetic data generation.

    Agility Robotics, Boston Dynamics, XPENG RoboticsHarnessing Isaac's simulation to advance humanoid efforts

    These companies are harnessing Isaac's simulation to advance their humanoid efforts.

    GE HealthcareUsing NVIDIA Isaac platform for health care simulation

    GE Healthcare is using the new NVIDIA Isaac platform for health care simulation built on NVIDIA Omniverse and using NVIDIA Cosmos for platform speed, development of robotic imaging and surgery systems.

    Taiwan governmentPartnership for Taiwan's first AI factory

    Announced Taiwan's first AI factory in partnership with Foxconn and the Taiwan government.

    Capital programs

    4
    TSMC fabs and advanced packaging plants in Arizonaunderway

    Benefit: 6 fabs and 2 advanced packaging plants

    TSMC is building 6 fabs and 2 advanced packaging plants in Arizona to make chips for NVIDIA. Process qualification is underway with volume production expected by year-end. NVIDIA has made substantial long-term purchase commitments to support these investments.

    SPIL and Amkor packaging, assembly and test facilities in Arizonaunderway

    SPIL and Amkor are investing in Arizona, constructing packaging, assembly and test facilities. NVIDIA has made substantial long-term purchase commitments to support these investments.

    Foxconn AI supercomputer factory in Houstonunderway

    Benefit: 1 million square foot factory to build AI supercomputers

    NVIDIA is partnering with Foxconn to construct a 1 million square foot factory in Houston to build AI supercomputers. NVIDIA has made substantial long-term purchase commitments to support these investments.

    Wistron AI supercomputer plant in Fort Worth, Texasunderway

    Wistron is building a similar plant in Fort Worth, Texas. NVIDIA has made substantial long-term purchase commitments to support these investments.

    Risks & headwinds

    3
    China export controls on H20Q1 FY26 and ongoing

    $4.5 billion charge for inventory and purchase obligations; $2.5 billion in H20 revenue unable to ship in Q1 FY26; expected $8 billion loss in H20 revenue for Q2 FY26; $50 billion China AI accelerator market effectively closed to U.S. industry.

    Mitigation: Evaluating limited options to supply compliant Data Center compute products; considering new products for the Chinese market within stringent limits.

    Challenging operating environmentQ1 FY26

    General market conditions impacting Q1 FY26, though revenue exceeded outlook.

    Mitigation: Strong product demand and execution, particularly Blackwell ramp.

    Tariff-related uncertaintyQ1 FY26

    Temporarily impacted Q1 Pro Visualization systems demand.

    Mitigation: Strong demand for AI workstations expected to drive sequential revenue growth in Q2.

    What to watch in Q2 FY26

    5

    Blackwell Ultra production shipments

    Q2 FY26
    CurrentSampling began earlier this month
    TargetProduction shipments commence later this quarter

    Why it matters

    Indicates the next phase of NVIDIA's product roadmap and continued innovation in AI compute.

    Sampling of GB300 systems began earlier this month at the major CSPs, and we expect production shipments to commence later this quarter.

    Q&A highlights

    6

    How much of the reasoning AI demand can NVIDIA serve, how big is the inference business, and is NVL72 rack scale necessary for reasoning inference?

    NVIDIA aims to serve most of the demand. Grace Blackwell NVLink 72 is ideal for reasoning AI due to the 100-1000x more tokens required for step-by-step problem-solving. It offers 40x higher speed and throughput compared to Hopper, significantly reducing cost and improving quality of service. The company redesigned supercomputers for this purpose and is now in full production.

    “Compared to Hopper, Grace Blackwell is some 40x higher speed and throughput, compared. And so this is going to be a huge, huge benefit in driving down the cost while improving the quality of response with excellent quality of service at the same time.”

    asked by Joseph Moore · answered by Jen-Hsun Huang

    2 min read6 chapters

    Detailed Narrative

    01

    Blackwell Ramp and AI Inference Surge

    NVIDIA's Blackwell architecture is experiencing its fastest ramp in company history, contributing nearly 70% of Data Center compute revenue in Q1 FY26. The introduction of GB200 NVL is enabling data center-scale workloads and achieving the lowest cost per inference token. Hyperscalers are deploying nearly 1,000 NVL72 racks or 72,000 Blackwell GPUs per week, with Microsoft alone ramping to hundreds of thousands of GB200s. The company notes a sharp jump in inference demand, with reasoning AI models driving a step-function surge in compute intensity.

    02

    China Export Controls Impact

    New U.S. export controls on H20, a data center GPU designed for the China market, significantly impacted Q1 FY26 results. NVIDIA recognized $4.6 billion in H20 revenue prior to April 9 but was unable to ship an additional $2.5 billion. This resulted in a $4.5 billion charge for inventory and purchase obligations. The company expects a meaningful decrease in China Data Center revenue in Q2 FY26, reflecting an estimated $8 billion loss in H20 revenue, and believes the $50 billion China AI accelerator market is effectively closed to U.S. industry.

    03

    Networking Innovation and Growth

    Networking revenue grew 64% quarter-over-quarter to $5 billion, driven by continued adoption of NVIDIA's platforms for AI factory workloads. NVLink, the scale-up platform, is a new growth vector with Q1 shipments exceeding $1 billion. Spectrum-X, the enhanced Ethernet solution, posted strong growth and is now annualizing over $8 billion in revenue, with Google Cloud and Meta added as new customers. NVIDIA also introduced Spectrum-X and Quantum-X silicon photonics switches for next-level AI factory scaling.

    04

    New Growth Engines: Sovereign, Enterprise, and Industrial AI

    NVIDIA is identifying new growth engines beyond traditional cloud AI. Sovereign AI sees nations investing in national AI platforms, with nearly 100 NVIDIA-powered AI factories in flight globally. Enterprise AI is taking off with new products like RTX Pro enterprise server, DGX Spark, and DGX Station designed for on-prem deployment. Industrial AI, driven by global reshoring of manufacturing, leverages Omniverse and AI for digital twin simulation and robotics, with companies like TSMC and Foxconn seeing significant benefits.

    05

    Gaming and AI PC Momentum

    Gaming revenue reached a record $3.8 billion, up 42% year-on-year, fueled by strong adoption of Blackwell architecture. The company improved supply and availability in Q1 and expects continued efforts in Q2. NVIDIA expanded its AI PC laptop offerings, including models capable of running Microsoft's CoPilot+. The launch of GeForce RTX 5060 and 5060 Ti brought Blackwell architecture to mainstream gaming, and the upcoming Nintendo Switch 2 will leverage NVIDIA's neural rendering and AI technologies.

    06

    Capital Allocation and Manufacturing Investments

    NVIDIA returned a record $14.3 billion to shareholders in Q1 FY26 through buybacks and dividends, emphasizing its capital return program. The company is also making substantial long-term purchase commitments to support onshore manufacturing in the U.S., including TSMC's fabs and advanced packaging plants in Arizona, and Foxconn and Wistron's AI supercomputer factories in Texas, aiming for chip-to-supercomputer production in America within a year.

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