NVIDIA Corporation (NVDA) Earnings

Management highlights

- **New Reporting Framework Transition** - Updated segmentation to better reflect current and future growth drivers, splitting business into two core market platforms: Data Center (with Hyperscale and ACIE sub-segments) and Edge Computing - Hyperscale includes revenue from public cloud and largest consumer internet companies; ACIE addresses growth in AI purpose-built data centers and AI factories across industries and countries; Edge Computing covers agentic and physical AI devices including PCs, gaming consoles, workstations, robotics, and automotive - Historical revenue breakdown for 9 prior quarters has been posted to NVIDIA's investor website - **Product and Deployment Milestones** - Blackwell architecture ramp is the fastest product launch in NVIDIA's history, with GB300 and NVL72 seeing particularly strong demand from frontier model builders and hyperscalers; hundreds of thousands of Blackwell GPUs have been cumulatively deployed - GPT 5.5 was co-designed, trained, and is served on Blackwell; Microsoft's Farweave AI data center (powered by hundreds of thousands of Blackwell GPUs) is live ahead of schedule; AWS will add more than 1 million Blackwell and Rubin GPUs starting 2026; Google will offer Blackwell to cloud customers with confidential computing capabilities - NVIDIA's share of frontier AI compute is growing, with Anthropic added as a new strategic partner alongside existing major model builders including OpenAI, Meta, and xAI - Spectrum-X (NVIDIA's AI-optimized Ethernet platform) is now larger than all other Ethernet network peers combined; InfiniBand grew more than 4x YoY driven by next-generation XDR technology deployments - Vera, the first CPU purpose-built for agentic AI, is on track to start production shipments in Q3 2026 as part of the VeraRubin platform; Vera delivers up to 1.5x faster per-core performance, 2x performance per watt, and 4x density per rack compared to x86 alternatives, opening a new $200 billion total addressable market (TAM) for NVIDIA - VeraRubin will deliver up to 35x higher inference throughput and up to 10x greater AI factory revenue compared to Blackwell; Google's XGS bare metal instances will support up to 960,000 Rubin GPUs across multiple sites - Blackwell Ultra swept all MLPerf inference benchmarks, delivering a 2.7x throughput increase and 60% reduction in cost per token for GV300 compared to 6 months prior - Physical AI collaboration is expanding: NVIDIA will power Uber's robotaxi fleet across nearly 30 cities and 4 continents by 2028, with leading industrial, surgical, and humanoid robotics companies building on NVIDIA technology - **Market and Ecosystem Dynamics** - AI infrastructure build-out is accelerating, driven by 1) ongoing transition of hyperscale workloads from CPUs to GPUs, and 2) accelerating adoption of native AI products and services across all industries; annual global AI infrastructure spending is on track to reach $3 trillion to $4 trillion by the end of the 2020s - NVIDIA holds a dominant, largely unique market position in fast-growing underpenetrated segments including ACIE, sovereign AI, and physical AI; ACIE growth is driven by demand from hundreds of thousands of enterprises and industrial companies across the world that require fully integrated AI factory solutions - NVIDIA has increased total supply (including inventory purchase commitments and prepaids) to $145 billion to meet ongoing demand; long-term supplier partnerships support the company's ability to scale - **Capital Allocation Update** - Prioritizes R&D and strategic investments to expand the AI market and strengthen NVIDIA's competitive position - Increased quarterly dividend from $0.01 to $0.25 per share (corrected from an initial $0.20 announcement during the call) - Announced a new $80 billion share repurchase authorization, in addition to the $39 billion remaining under the existing plan; expects to return ~50% of annual free cash flow to shareholders in 2027 - Returned a record $20 billion to shareholders in Q1

Guidance

- Q2 2027 total revenue is expected to be $91 billion, plus or minus 2%, with sequential growth driven primarily by the Data Center segment - GAAP and non-GAAP gross margins are expected to be 74.9% and 75% respectively, plus or minus 50 basis points; full-year 2027 gross margins are still expected to be in the mid-70% - Q2 GAAP and non-GAAP operating expenses are expected to be approximately $8.5 billion and $8.3 billion respectively; full-year 2027 operating expenses are now expected to grow in the upper 40% range YoY, driven by higher R&D and increased AI tool usage to boost productivity - Full-year 2027 GAAP and non-GAAP tax rates are expected to be between 16-18% (excluding discrete items from material tax changes), a downward revision from the prior expectation of 17-19%, driven by changes in geographic mix - Management reaffirmed confidence in $1 trillion in total Blackwell and Rubin revenue forecast from 2025 through 2027; standalone Vera CPU revenue is incremental to this $1 trillion base - No China data center compute revenue is included in the outlook, as no revenue has been generated from H200 shipments to date and regulatory import approval remains uncertain

Segment performance

NVIDIA reported total Q1 2027 revenue of $82 billion, up 85% year over year (YoY) and 20% sequentially. Under the new reporting framework split into two core market platforms: 1. **Data Center**: Generated $75 billion in revenue, up 92% YoY and 21% sequentially, accounting for 91.46% of total company revenue. Within Data Center: - Hyperscale: $38 billion, ~50% of Data Center revenue, up 12% quarter over quarter (QoQ) - ACIE (AI clouds, industrial, enterprise): $37 billion, ~50% of Data Center revenue, up 31% QoQ, with AI cloud revenue more than tripling YoY - Data center computing: $60 billion, up 77% YoY - Data center networking: $15 billion, nearly tripled YoY 2. **Edge Computing**: Generated $6.4 billion in revenue, up 10% QoQ and 29% YoY, accounting for 7.8% of total company revenue. Physical AI revenue exceeded $9 billion over the last 12 months.

Risks & headwinds

- Regulatory uncertainty remains for data center business in China: while the U.S. government has approved licenses for H200 shipments to Chinese customers, no revenue has been generated to date, and it remains unclear whether imports will be allowed into the country - NVIDIA remains not immune to potential supply chain challenges, despite increased total supply commitments and strong long-term supplier relationships - VeraRubin production and ramp depends on successful integration of multiple complex custom components, with timing of market availability dependent on completing system assembly from production - LPX (SRAM-based low-latency accelerator) has a limited addressable market due to its constrained throughput and context processing capacity, remaining a niche product for the foreseeable future

Analyst Q&A

• Q: What is the rationale for the new segmentation, and how does the new CPU business fit across segments? /

  A: Segmentation was updated to reflect NVIDIA's business growth and diversity. AI has diverse applications across industries, runs in different environments from hyperscale clouds to on-prem industrial facilities to the edge, and has different governance requirements. The new segmentation simplifies understanding of three distinct core areas: hyperscale clouds, ACIE (covering AI natives, enterprise/industrial on-prem, and sovereign AI), and edge/robotic AI. Each segment has different technology stacks, go-to-market models, and customer bases, with ACIE serving hundreds of thousands of companies worldwide compared to just a handful of hyperscalers. /

• Q: Do you expect to grow faster than hyperscale CapEx, and will hyperscale CapEx continue growing rapidly after this year? /

  A: NVIDIA should grow faster than hyperscale CapEx because our data center business has two large high-growth segments: the established hyperscale segment (which is growing to $1 trillion in annual CapEx and will continue growing rapidly) and the fast-growing ACIE segment. ACIE includes AI native clouds, enterprise/industrial on-prem AI factories, and sovereign AI, which is extremely fragmented, poorly understood by investors, and growing incredibly quickly. Most of this market cannot be served by semi-custom chip designs, and NVIDIA is uniquely positioned to serve this segment with a fully integrated, open platform solution, giving us large market share and strong upside. /

• Q: How will VeraRubin impact NVIDIA's inference market share going forward? /

  A: NVIDIA's inference market share is growing very quickly. We have added major new frontier model partners like Anthropic this year, and we are rapidly expanding capacity to support their growth, gaining significant new share. Nearly all frontier model companies will adopt VeraRubin immediately from launch, making it even more successful than Grace Blackwell. Additionally, NVIDIA already holds nearly 100% share of inference in the fast-growing ACIE segment and is the only major provider of physical AI inference, further driving share gains. /

• Q: Is the $20 billion 2027 standalone Vera CPU revenue incremental to existing product revenue, and is the CPU business incremental to GPU workloads rather than cannibalistic? /

  A: The $20 billion figure refers entirely to standalone Vera CPU revenue, which is incremental to CPU revenue included in VeraRubin systems. Agentic AI requires CPUs to run orchestration, tool use, and I/O for agents, while heavy thinking and inference workloads run on GPUs, so CPUs are incremental, not cannibalistic to GPU demand. Vera is purpose-built for agentic AI workloads, which will require billions of agents over time creating massive new demand for custom-designed CPUs that optimize for speed rather than traditional cloud core rental economics.