Nvidia (NVDA +0.99%) dominates the market for data center GPUs, and it's now following the well-tread playbook of expanding into adjacent areas. The company has been investing in networking for years, and with its push into rack-scale AI solutions, the networking business is exploding.
In the third quarter of fiscal 2026, Nvidia generated $8.2 billion in networking revenue, up 162% year over year. This category includes NVLink, which connects GPUs together, InfiniBand switches, and the Spectrum-X Ethernet networking platform. Nvidia noted that Meta, Microsoft, Oracle, and xAI are currently building massive AI data centers that will use its Spectrum-X Ethernet switches.
Image source: Nvidia.
Why AI networking is so critical
At CES 2026, Nvidia CFO Colette Kress disclosed that the networking attach rate, or the percentage of customers who buy full AI systems and also purchase networking products, is now close to 90%. Kress also noted that even customers who use their own customer AI chips for a portion of a deployment are still often using Nvidia's networking technology.
The networking needs of an AI data center are very different than the networking needs of a standard cloud data center. For AI training workloads, the data throughput between GPUs must be sufficiently high to prevent the GPUs from idling. The quick movement of data is also crucial for AI inference workloads.
Nvidia has grown to become a significant player in the Ethernet switching market, particularly when considering only the data center portion of the market. According to IDC, revenues for ultra-fast 800GbE switches nearly doubled sequentially in the third quarter of 2025. Nvidia now has an 11.6% share of the data center Ethernet switch market, behind only Arista Networks and Cisco Systems.
Networking growth could be supercharged by the Rubin platform
Nvidia unveiled its Rubin platform at CES, which brings together its GPUs, CPUs, and multiple networking technologies into 8-GPU and 72-GPU systems. The Vera Rubin NVL72 is a rack-scale solution that will be deployed by all the major AI cloud providers in 2026.
The Rubin platform features the new Spectrum-6 series of Ethernet switches, which include ports with 800 GB/s connectivity and can deliver up to 102.4 Tb/s of switching capacity. Nvidia's increasing focus on selling rack-scale AI systems, rather than data center GPUs alone, has the potential to drive networking revenue significantly higher in 2026 and beyond.
How big is the networking opportunity for Nvidia? MarketsandMarkets forecasts the AI networking market to grow from $14.9 billion in 2025 to $46.8 billion by 2029, representing a compound annual growth rate of 33.8%. Even if Nvidia loses some ground in the AI accelerator market to its competitors, the AI networking business is another growth engine for the company.
NASDAQ: NVDA
Key Data Points
Is Nvidia stock a buy?
Enormous AI data centers are being planned and built, with trillions of dollars likely to be invested in AI infrastructure. By 2030, McKinsey predicts that global capital spending on data centers will need to reach nearly $7 trillion to meet demand. A significant chunk of that spending will go straight to Nvidia.
While this outlook presents a strong bull case for Nvidia stock, investors should still exercise caution. It's impossible to predict what demand for AI will look like four years from now, and it's also impossible to know whether the industry is currently on track to overbuild AI data centers. If the industry does overbuild, like during other capital-intensive bubbles of the past, demand for Nvidia's AI systems isn't going to hold up.
In the long run, AI as a technology is absolutely here to stay, and the productivity benefits should not be underestimated. However, benefiting from AI as an investor may not be as simple as buying Nvidia stock. Nvidia is thriving today, but a lot can change in a few years as the AI industry continues to rapidly evolve.
