In the third quarter of fiscal 2026, NVIDIA (NYSE: NVDA) posted another record-setting performance, reporting $57 billion in revenue, up 62% year-over-year, and a sequential increase of $10 billion, or 22% with datacenter revenue hitting an all-time high of $51 billion, growing 66% from the prior year.
Compute sales rose 56% year-over-year, propelled by the GB300 product ramp, while networking revenue more than doubled, fueled by demand for NVLink scale-up capabilities and strong performance across Spectrum-X Ethernet and Quantum-X InfiniBand offerings.
NVIDIA devoted extensive attention to its technology direction by emphasizing a full-stack design that integrates GPU, CPU, networking, and software to support next-generation AI and computing workloads.
CEO Jensen Huang described the current environment as “the first time since the dawn of Moore’s Law” that the world is undergoing such wide-ranging changes in computing paradigms, with NVIDIA at the core of enabling these transitions across industries and use cases.
Why It Matters: NVIDIA’s announcements are highly relevant for companies planning infrastructure upgrades and AI integrations. With ecosystem alignment across major cloud service providers, enterprise software vendors, and global manufacturers, NVIDIA is making a case that its developments are easier for organizations to run their most critical AI workloads on one consistent architecture. As CFO, Colette Kress emphasized, “Our architecture will drive the most revenues per watt,” reinforcing the growing emphasis on total cost of ownership and long-term system value in compute procurement.
- NVIDIA Expects $500 Billion in AI Infrastructure Demand by 2026: Kress reiterated that NVIDIA remains on track to generate $500 billion in cumulative revenue from its Blackwell and Rubin platforms through the end of calendar 2026. She noted that the company has already delivered $150 billion toward that figure and highlighted new commitments, including the agreement with Saudi Arabia’s KSA to deploy an additional 400,000–600,000 GPUs over a three-year period.
- A Unified Architecture Serving All AI Workloads: NVIDIA reinforced its philosophy of “one architecture” that supports AI phases across every major deployment scenario. Huang emphasized that the company’s CUDA-X libraries and full-stack engineering make it possible to run scientific simulations, structured data pipelines, and machine learning inference all on the same infrastructure. This approach reduces difficulty for enterprises managing multi-cloud and hybrid environments by ensuring workloads can be seamlessly transitioned between training and production across NVIDIA’s GPU-accelerated platforms.
- Networking Capabilities Are Becoming a Core Differentiator: NVIDIA’s networking business generated $8.2 billion in revenue during the quarter, a 162% year-over-year increase. The company also introduced Spectrum-XGS, a new scale-across solution designed for gigascale AI factories. Customers like Microsoft are using NVIDIA’s networking switches in their next-generation AI facilities. This traction shows that high-speed, AI-specific networking is now a prerequisite for scaling large models. With NVLink Fusion being adopted by Fujitsu and Intel, NVIDIA is integrating its networking stack into broader CPU ecosystems as well, aiming to connect disparate components into a cohesive data center fabric.
- Enterprise Adoption of Agentic and Generative AI Continues to Expand: A wide range of companies are deploying NVIDIA’s full-stack solutions to improve operational efficiency and user experience. Salesforce has reported a 30% increase in new code development productivity thanks to its use of Cursor, a coding assistant running on NVIDIA hardware. RBC has slashed analyst report generation times using Agentic AI applications. Unilever is using NVIDIA’s digital twin technology to double the speed of content creation while cutting costs in half. Furthermore, Palantir’s Ontology platform is now being optimized using NVIDIA CUDA-X libraries. These use cases reveal how organizations are embedding AI into core workflows, resulting in tangible performance improvements that directly impact productivity and cost structure.
- Preparing for the Next Generation with Rubin and Grace Blackwell: Looking ahead, NVIDIA is preparing to ramp production of the Vera Rubin platform in the second half of 2026. This third-generation rack-scale system is designed to deliver a substantial performance-per-dollar improvement over current-generation Blackwell systems, while maintaining compatibility. NVIDIA has already received silicon from supply chain partners. The company aims to improve energy efficiency and throughput with each new generation in the face of global power and data center constraints.
Go Deeper -> NVIDIA Earnings Report – MarketBeat
Our Latest CIO Field Notes
Disney’s Tech Strategy Is Reshaping Consumer Engagement and Monetization
Content consolidating.
Trusted insights for technology leaders
Our readers are CIOs, CTOs, and senior IT executives who rely on The National CIO Review for smart, curated takes on the trends shaping the enterprise, from GenAI to cybersecurity and beyond.
Subscribe to our 4x a week newsletter to keep up with the insights that matter.
