Skip to content
    NVDA
    Earnings call· Jan 2025(Q4 FY25)

    NVIDIA CORP NVDA

    Feb 26, 2025 Source

    Executive summary

    NVIDIA Q4 FY25 — Record Data Center Revenue Driven by Unprecedented Blackwell Ramp

    NVIDIA achieved record Q4 FY25 results, primarily fueled by unprecedented demand for its Data Center products, particularly the rapid ramp of the Blackwell architecture. The company is navigating a complex supply chain to meet surging AI infrastructure needs, while also managing a temporary dip in gaming and networking revenues. Management remains highly confident in sustained long-term growth driven by the pervasive integration of AI across all industries and the continuous evolution of its product roadmap.

    Highlights

    5

    • Record Q4 revenue of $39.3 billion, up 78% year-on-year and 12% sequentially, exceeding outlook of $37.5 billion.

    • Data Center revenue reached a record $35.6 billion in Q4, up 93% year-on-year and 16% sequentially, driven by the Blackwell ramp.

    • Blackwell sales exceeded expectations, delivering $11 billion in Q4, marking the fastest product ramp in company history.

    • Automotive revenue was a record $570 million, up 103% year-on-year and 27% sequentially.

    • Returned $8.1 billion to shareholders in Q4 through share repurchases and cash dividends.

    Concerns

    3

    • Gaming revenue decreased 22% sequentially and 11% year-on-year to $2.5 billion in Q4, impacted by supply constraints.

    • Networking revenue declined 3% sequentially in Q4.

    • Gross margins expected to be in the low 70s during the Blackwell ramp, with Q1 FY26 guidance at 71%, down from 73.5% in Q4 FY25.

    Guidance & targets

    14
    CategoryTargetConfidence
    Total revenue
    $43 billion, plus or minus 2%
    high materiality
    High
    Data Center revenue growth
    sequential growth
    high materiality
    High
    Gaming revenue growth
    strong sequential growth
    medium materiality
    High
    Networking revenue growth
    return to growth
    medium materiality
    High
    GAAP gross margin
    70.6%, plus or minus 50 basis points
    high materiality
    High
    Non-GAAP gross margin
    71%, plus or minus 50 basis points
    high materiality
    High
    Gross margins (Blackwell ramp)
    low 70s
    high materiality
    High
    Gross margins (fully ramped Blackwell)
    mid-70s
    high materiality
    Medium
    GAAP operating expenses
    approximately $5.2 billion
    medium materiality
    High
    Non-GAAP operating expenses
    approximately $3.6 billion
    medium materiality
    High
    Full year FY26 operating expenses growth
    grow to be in the mid-30s
    medium materiality
    High
    GAAP and non-GAAP other income and expenses
    approximately $400 million
    low materiality
    High
    GAAP and non-GAAP tax rates
    17%, plus or minus 1%
    low materiality
    High
    Automotive vertical revenue
    approximately $5 billion
    medium materiality
    High

    Segment performance

    4
    SegmentRevenueYoYQoQMargin
    Data Center
    Record revenue driven by the Blackwell ramp and Hopper 200 continued sequential growth. Large CSPs represented about half of revenue. Regional cloud hosting increased as a percentage of data center revenue. Data center sales in China remained well below levels seen at the onset of export controls. Networking revenue declined 3% sequentially but is expected to return to growth in Q1 FY26.
    Blackwell sales: $11 billionLarge CSPs revenue growth: nearly 2x YoYConsumer Internet revenue growth: 3x YoYEnterprise revenue growth: nearly 2x YoYNetworking attach to GPU compute systems: >75%
    		$35.6 billion93%16%—
    Gaming
    Q4 shipments were impacted by supply constraints, despite strong demand throughout the holiday. Full year revenue was $11.4 billion, up 9% year-on-year. Expect strong sequential growth in Q1 FY26 as supply increases.
    		$2.5 billion-11%-22%—
    Professional Visualization
    Key industry verticals driving demand include automotive and health care. NVIDIA technologies and generative AI are reshaping design, engineering, and simulation workloads, fueling demand for NVIDIA RTX workstations. Full year revenue was $1.9 billion, up 21% year-on-year.
    		$511 million10%5%—
    Automotive
    Record revenue driven by the continued ramp in autonomous vehicles, including cars and robotaxis. Toyota is adopting NVIDIA Orin, and Aurora and Continental will deploy driverless trucks powered by NVIDIA DRIVE Thor. NVIDIA DRIVE Hyperion has passed industry safety assessments.
    		$570 million103%27%—

    Operational metrics

    18
    Total Revenue Growth
    78%YoY
    Q4 FY25

    Total revenue for Q4 FY25 was $39.3 billion.

    Total Revenue Growth
    12%QoQ
    Q4 FY25

    Total revenue for Q4 FY25 was $39.3 billion.

    Total Revenue Growth
    114%YoY
    FY25

    Total revenue for fiscal 2025 was $130.5 billion.

    Data Center Compute Revenue Growth
    18%QoQ
    Q4 FY25

    Data Center compute revenue jumped sequentially.

    Data Center Compute Revenue Growth
    over 2xYoY
    Q4 FY25

    Data Center compute revenue jumped year-on-year.

    Inference Cost Reduction
    200x
    last 2 years

    Achieved through performance and pace of innovation.

    Perplexity Monthly Queries
    435 million
    monthly

    Perplexity reduced its inference costs 3x with NVIDIA Triton Inference Server and TensorRT-LLM.

    Perplexity Inference Cost Reduction
    3x
    Q4 FY25

    Achieved with NVIDIA Triton Inference Server and TensorRT-LLM.

    Microsoft Bing Visual Search Speed Up
    5x
    Q4 FY25

    Achieved for Visual Search across billions of images with NVIDIA TensorRT and acceleration libraries.

    Developer Ecosystem
    5.9 million
    Q4 FY25

    Number of developers continuously improving customer economics through full stack optimizations.

    Networking Attach Rate
    >75%
    Q4 FY25

    Robust attach rate for networking to GPU compute systems.

    GeForce RTX 50 Series AI TOPS
    up to 3,400
    Q4 FY25

    Powered by Blackwell architecture, 5th generation Tensor Cores and 4th generation RT Cores.

    DLSS 4 Frame Rate Boost
    up to 8x
    Q4 FY25

    AI-driven frame generation turning 1 rendered frame into 3.

    Max-Q Battery Life Extension
    up to 40%
    Q4 FY25

    New NVIDIA Max-Q technology for laptops.

    Non-GAAP gross margin
    73.5%down sequentially
    Q4 FY25

    Down sequentially as expected with first deliveries of Blackwell architecture.

    GAAP operating expenses growth
    9%QoQ
    Q4 FY25

    Reflecting higher engineering development costs and higher compute and infrastructure costs for new product introductions.

    Non-GAAP operating expenses growth
    11%QoQ
    Q4 FY25

    Reflecting higher engineering development costs and higher compute and infrastructure costs for new product introductions.

    Share buyback / capital return
    $8.1 billion
    Q4 FY25

    Returned to shareholders in the form of share repurchases and cash dividends.

    Industry KPIs

    4
    MetricValueDetails
    Ai data center revenue$35.6 billionUSD
    Design wins socket pipelineToyota adopting NVIDIA Orin; Aurora and Continental deploying driverless trucks with NVIDIA DRIVE Thor
    Node platform ramp scheduleBlackwell architecture; Blackwell Ultra; Vera Rubin
    End market segment revenue mixData Center: $35.6 billion; Gaming: $2.5 billion; Professional Visualization: $511 million; Automotive: $570 millionUSD

    Product announcements

    5
    ProductTypeDetails
    GeForce RTX 50 Series desktop and laptop GPUslaunch
    NVIDIA Llama Nemotron model family NIMslaunch
    NVIDIA Cosmos World Foundation Model Platformlaunch
    Blackwell Ultraroadmap
    Vera Rubinroadmap

    Deals & partnerships

    10
    Microsoft Azure, GCP, AWS, OCIDeployment of GB200 systems

    These large CSPs were some of the first to stand up Blackwell.

    CoreWeaveBuilding large AI factories with Spectrum-X

    CoreWeave has been public about the successful bring up of their systems.

    xAIAdopting GB200 to train and inference next generation Grok AI models

    Part of consumer internet revenue growth driven by generative AI use cases.

    MetaAndromeda advertising engine runs on NVIDIA's Grace Hopper Superchip

    Serving vast quantities of ads across Instagram and Facebook applications.

    SAP and ServiceNowAmong the first to use NVIDIA Llama Nemotron model family NIMs

    Helping developers create and deploy AI agents across various applications.

    IQVIA, Illumina, Mayo Clinic, ARC InstituteUsing NVIDIA AI for healthcare

    Health care leaders leveraging generative and agentic AI.

    Hyundai Motor GroupAdopting NVIDIA technologies to accelerate AV and robotics development and smart factory initiatives

    Announced at CES.

    ToyotaWill build its next-generation vehicles on NVIDIA Orin running the safety certified NVIDIA DriveOS

    Toyota is the world's largest automaker.

    Aurora and ContinentalWill deploy driverless trucks at scale powered by NVIDIA DRIVE Thor

    Announced at CES.

    CiscoIntegrating Spectrum-X into their networking portfolio

    Cisco has a large enterprise footprint and global reach.

    Risks & headwinds

    4
    Gaming revenue impacted by supply constraintsQ4 FY25

    Gaming revenue decreased 22% sequentially and 11% year-on-year to $2.5 billion in Q4 FY25.

    Mitigation: Expect strong sequential growth in Q1 FY26 as supply increases.

    Data Center sales in China impacted by export controlsQ4 FY25 and ongoing

    Data center sales in China remained well below levels seen at the onset of export controls, approximately at the current percentage.

    Mitigation: Will continue to comply with export controls while serving customers; believe China shipments will remain roughly at current percentage absent regulatory changes.

    Gross margin compression during Blackwell rampQ1 FY26 and during Blackwell ramp

    GAAP gross margin was 73% and non-GAAP gross margin was 73.5% in Q4 FY25, down sequentially. Q1 FY26 non-GAAP gross margin expected to be 71%, plus or minus 50 basis points. Gross margins expected to be in the low 70s during the ramp.

    Mitigation: Focusing on expediting manufacturing to meet demand. Once fully ramped, opportunities to improve cost and return to mid-70s gross margins late this fiscal year.

    Uncertainty regarding future tariff impactFuture

    A little bit of an unknown

    Mitigation: Awaiting further understanding of the U.S. government's plan (timing, where, how much); will always follow export controls and/or tariffs.

    What to watch in Q1 FY26

    5

    Gaming revenue recovery

    Q1 FY26
    Current$2.5 billion (Q4 FY25), down 22% QoQ
    TargetStrong sequential growth

    Why it matters

    To confirm supply constraints are easing and demand is translating into revenue for the Gaming segment.

    However, Q4 shipments were impacted by supply constraints. We expect strong sequential growth in Q1 as supply increases.

    Q&A highlights

    6

    With the blurring lines between training and inference, what does this mean for inference-dedicated clusters and NVIDIA's impact?

    Jensen explained three scaling laws: pre-training, post-training, and test-time/reasoning AI. He noted that post-training and reasoning AI can demand significantly more compute (100x or more for reasoning) than pre-training. Blackwell is designed for this, offering 25x higher throughput for reasoning AI. The architecture is fungible, allowing unified data centers for various AI workloads.

    “The amount of computation you use for post training is actually higher than pretraining. And it's kind of sensible in the sense that you could, while you're using reinforcement learning, generate an enormous amount of synthetic data or synthetically generated tokens.”

    asked by Christopher Muse · answered by Jen-Hsun Huang

    2 min read5 chapters

    Detailed Narrative

    01

    Blackwell Architecture Ramp and Demand

    The Blackwell architecture saw an unprecedented🌐 ramp, delivering $11 billion in Q4 FY25, making it the fastest product ramp in company history. Production is in full gear across multiple configurations, and supply is rapidly expanding to meet strong customer demand. Initial deployments are already seeing clusters of 100,000 GPUs or more, with shipments having commenced for several such infrastructures. This rapid deployment underscores the urgency of customers to scale AI capabilities.

    02

    AI Scaling Laws and Inference Acceleration

    Management highlighted three scaling laws driving AI compute demand: pre-training, post-training, and test-time scaling/reasoning AI. Post-training and model customization, using techniques like fine-tuning and reinforcement learning, can collectively demand orders of magnitude more compute than pre-training. Reasoning AI, exemplified by models like OpenAI o3 and Grok 3, can require 100x more compute per task. Blackwell is specifically architected for reasoning AI inference, offering up to 25x higher token throughput and 20x lower cost versus Hopper 100.

    03

    Geographic and Customer Dynamics

    Large CSPs represented about half of Data Center revenue in Q4, nearly doubling year-on-year, and were early adopters of GB200 systems. Regional cloud hosting also increased its percentage of Data Center revenue, reflecting global AI factory build-outs. Data Center sales in China remained competitive and well below pre-export control levels, expected to stay at current percentages without regulatory changes. The U.S. saw the strongest sequential growth in Data Center revenue due to the initial Blackwell ramp.

    04

    Product Roadmap and Annual Cadence

    NVIDIA is maintaining an annual product cadence, with Blackwell Ultra launching in the second half of FY26, followed by Vera Rubin. The transition from Hopper to Blackwell was complex due to significant system architecture changes (NVLink 8 to NVLink 72). However, subsequent transitions like Blackwell Ultra will slot in more easily as the system architecture remains consistent. The company is working closely with partners on these transitions, ensuring a continuous pipeline of advanced products.

    05

    AI's Pervasive Impact and Future Growth Vectors

    AI is rapidly integrating into every application and service across consumer, enterprise, and physical domains, making it a mainstream technology. Management believes AI will address a larger part of the world's GDP than any prior software tool. New growth vectors include agentic AI for enterprise (e.g., for employee productivity, design, operations) and physical AI for robotics (e.g., autonomous vehicles, humanoid robots), which require new types of AI to understand the physical world. These represent brand new computing opportunities.

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