EPS · actual vs est
Revenue · actual vs est
Summary
Generated 2026-11-17
Management highlights
Leadership and continuity
- The board said Jensen Huang voluntarily resigned as president, CEO, and director following disclosure of personal behavior inconsistent with company standards. The board accepted the resignation, does not intend to reinstate him, appointed CFO Colette Kress interim CEO, and began a permanent CEO search led by Heidrick & Struggles.
- Management characterized operations as business as usual, citing employee commitment, leadership continuity, and support from customers and suppliers. The board also created a national security and export-control oversight committee.
Demand, products, and ecosystem
- Management described demand for AI compute as sustainable and broad-based across hyperscalers, frontier labs, AI-native companies, enterprises, and governments. Cloud-industry backlog reached $2.5 trillion, up from $2 trillion three months earlier.
- Frontier labs are expected to contribute about 15% of Data Center revenue this year. Management highlighted rapid growth in open and agentic AI, with agentic workloads at about 80% of OpenRouter tokens and weekly token volume up 35x year over year.
- NVIDIA announced an agreement to acquire Hugging Face and emphasized open models, enterprise adoption, and an expanding partner ecosystem. Token prices have fallen nearly 60% since May while usage and compute demand have increased; GPU rental rates have also risen.
- Physical AI revenue nearly tripled year over year and is on track to exceed $25 billion for the full year. Sovereign AI revenue more than tripled year over year.
Supply, execution, and infrastructure
- Demand continues to exceed supply. NVIDIA is expanding capacity with suppliers across wafers, substrates, and advanced packaging, and said the Grace Blackwell-to-Rubin transition has been smooth. Hyperscalers brought Rubin into production in about one month, half the time required for Grace Blackwell.
- Q3 purchase commitments increased to $390 billion. Data-center lease and cloud-service commitments exceeded $100 billion; neocloud commitments were $32 billion. Total guarantees were $118 billion, including $105 billion of credit support for the Portsmouth campus.
- Management said it will be thoughtful about capital participation and disciplined on commitments, while considering support for supplier capacity and selected customer infrastructure. NVIDIA’s networking revenue exceeded $21 billion; its co-packaged-optics roadmap remains on track.
- The new DSX Pod reference architecture is designed to shorten AI-factory deployment from months to weeks. Initial customer shipments are planned for early next year.
Segment performance
NVIDIA reported Q3 revenue of $116 billion, more than doubling year over year and rising 20% sequentially. Segment and product revenue disclosed:
- Data Center: $108 billion, up 21% sequentially. This represented approximately 93.1% of total revenue. Hyperscale revenue was $60 billion, up 23% sequentially (approximately 51.7% of company revenue); AI Clouds, Industrial and Enterprise (ACI&E) revenue was $47 billion, up 18% sequentially and nearly 3x year over year (approximately 40.5% of company revenue). The two subcategory figures total $107 billion, reflecting rounding or classification differences versus reported Data Center revenue.
- Vera Rubin platform: nearly $20 billion in its first production-shipment quarter; this is a product/platform figure within Data Center, not a separate reportable segment.
- Edge computing: $8.2 billion, up 14% sequentially and 42% year over year (approximately 7.1% of company revenue).
- Networking: revenue exceeded $21 billion; no growth rate or separate revenue contribution was stated. It is included within the company’s broader reporting, not an additional segment to add to total revenue.
- Licensed Hopper Data Center products sold to China: $2.4 billion in Q3. NVIDIA said it is no longer shipping these products to China and excludes such sales from guidance. Non-GAAP gross margin was 74.9% and GAAP gross margin was 74.8%, both above outlook due to product mix. Operating income and EPS reached records, but their absolute values were not provided in the remarks.
Guidance
- Q4 FY2027 revenue is expected to be $130 billion, plus or minus 3%, with sequential growth driven by AI-infrastructure demand and the Rubin ramp. Q4 includes an additional week.
- FY2028 revenue growth guidance was raised to approximately 75% year over year, from the prior 70%, despite comparison against a higher FY2027 base. Management cited stronger available supply alongside continued demand growth; it did not confirm the analyst’s suggested 100% demand-growth framing.
- Q4 GAAP and non-GAAP gross margins are expected at 71%–72%, which management expects to be the low point. FY2028 margins are expected to improve to 72%–73%, unchanged from the prior outlook.
- Q4 GAAP and non-GAAP operating expenses are expected to be approximately $11 billion and $10.7 billion, respectively, including effects from the extra week and transition-related costs. Investment will continue in areas supporting market growth and NVIDIA’s position.
- FY2027 GAAP and non-GAAP tax rates are expected to be 16%–18%, excluding discrete items and material tax-environment changes.
- NVIDIA expects total CPU revenue to more than double in FY2028. It plans to complete its share-repurchase authorization by the end of FY2028 and return excess free cash flow, net of strategic uses, to shareholders.
Risks
- The CEO transition creates leadership and employee-retention uncertainty. The board said it will conduct a thorough search and declined to discuss the circumstances or ongoing investigation.
- U.S. export controls create uncertainty for China-related sales. The Q4 outlook contains no licensed Hopper Data Center sales to China, and the company has stopped shipping those products there.
- Supply remains constrained relative to demand, with capacity expansion dependent on suppliers and critical components, including wafers, substrates, and advanced packaging.
- NVIDIA has substantial purchase commitments, leases, cloud commitments, and guarantees. Some customer or neocloud infrastructure plans depend on securing financing and completing construction; guarantees generally take effect only once projects are ready, and may not apply if those prerequisites fail.
- Management acknowledged that it must assess customer capital access, data-center readiness, and committed offtake when allocating constrained supply. It also said future financing and guarantees will be evaluated selectively and with capital discipline.
- Gross margins are expected to bottom in Q4, reflecting supply costs and customer purchase orders already honored; management expects recovery in FY2028 but noted pricing and cost considerations.
Q&A highlights
Q: What supports the increase in FY2028 growth guidance to 75%, and is demand still above that level? / A: Management cited both continued demand growth and additional supply it expects to provide. It said export-control changes are still being assessed and did not reaffirm a 100% demand-growth estimate.
Q: How will NVIDIA preserve employee and customer confidence during the CEO transition? / A: Kress emphasized employee commitment, an experienced leadership team, and continued customer and supplier engagement. The board is considering internal and external CEO candidates and said it will prioritize finding the right leader over rushing the appointment.
Q: How does NVIDIA allocate scarce compute among hyperscalers, labs, and neoclouds? / A: Allocation depends on whether customers have ready land, power, and data-center shells, sufficient capital, and credible offtake so they can use shipments efficiently. Management said allocation is not based on different customer pricing and will expand as supply grows.
Q: Are neocloud guarantees new, and when do they apply? / A: Kress said the disclosure clarifies existing practice, not a change. Guarantees generally begin after the customer raises capital, builds and prepares the facility, and can procure and install the compute; delays before readiness can mean the guarantee does not take effect.
Q: How will NVIDIA balance buybacks against supplier or customer financing? / A: Shareholder returns remain a top priority, with plans to use the expanded buyback authorization by FY2028. Capital may also support supplier capacity and selected infrastructure or guarantees, subject to close review of counterparties and project strength.
Key numbers
Reported versus consensus
Earnings calendar feed
| Metric | Reported | Consensus | Delta | Prior year |
|---|---|---|---|---|
| EPS | $2.68 | $2.48 | +8.0% | $1.30 |
| Revenue | $115.84B | $108.74B | +6.5% | $57.01B |
Transcript
November 17, 2026Full transcript unavailable for redistribution
The structured summary above covers the available call sections. Full transcript text is not included on this page.
Continue exploring
Prior quarters
This page presents the stored structured earnings-call summary and deterministic earnings calendar values. How this is generated. For informational purposes only; not investment advice.