Nvidia (NVDA +3.03%) continues to be a dominant force in the artificial intelligence (AI) boom, thanks largely to its industry-leading graphics processing units (GPUs). But another type of technology is becoming increasingly important to the company's growth.
Nvidia generated a record $14.8 billion in data center networking revenue in its fiscal 2027 first quarter (which ended April 26), up 199% year over year. Networking revenue rose from $8.6 billion in its fiscal 2024 to $13 billion in its fiscal 2025 and to $31.4 billion in its fiscal 2026. Its opportunity in the space could become even larger as AI clusters scale further and require increasingly powerful networking infrastructure to connect many thousands of accelerator chips together.
Image source: Getty Images.
Networking is becoming a major growth engine
Modern AI systems increasingly depend on high-performance networking as well as raw computing power. Training and running increasingly sophisticated models requires massive amounts of data to be moved rapidly among large numbers of accelerators and data storage devices. If the network cannot keep up, communication can become a bottleneck that reduces GPU utilization and slows AI workloads.
NASDAQ: NVDA
Key Data Points
Nvidia's NVLink technology connects GPUs inside powerful AI rack systems, while Spectrum-X Ethernet connects servers and racks across larger data centers. As AI systems scale, the company has more opportunities to sell networking and interconnect technology alongside its computing hardware.
According to research firm IDC, Nvidia captured 21.5% of data center Ethernet switching revenue in the first quarter of 2026, making it the market leader.
Custom AI chips are strengthening the networking opportunity
Custom AI chips are a growing competitive threat to Nvidia's GPUs. But the NVLink Fusion rack-scale platform could cushion the company against that threat by allowing some custom processors to work alongside its networking and infrastructure technologies. That gives it another way to benefit from the rising AI infrastructure spending, even in cases when it doesn't supply the AI accelerators.
Nvidia has also expanded its partnership with custom chip designer Marvell Technology (MRVL +2.25%). Under the partnership, Marvell will provide custom accelerators and the networking hardware needed to connect those processors at high speeds, while Nvidia will supply technologies including NVLink high-speed interconnect technology, Spectrum-X switches, ConnectX network adapters, and BlueField data processing units.
Nvidia's networking business will not benefit from every custom AI chip. Customers can still choose competing networking technologies, and its networking gains may not fully compensate for the loss of lucrative GPU sales to rival chipmakers.
Competition is also significant, with market research firm IDC estimating Arista Networks' share of the data center Ethernet switching market at 20.7%, marginally below Nvidia's 21.5% share.
Still, Nvidia is trying to sell more of the technology that goes into each AI data center, rather than relying so heavily on its GPUs.
Networking could help support a $7.5 trillion market capitalization for Nvidia
Nvidia is trading now at around 24.1 times Wall Street's fiscal 2027 earnings estimate of about $9 per share (as of Aug. 12). Analysts currently expect its earnings to increase to approximately $12.90 per share in fiscal 2028.
If Nvidia delivers on those expectations and continues to trade at roughly the same forward valuation, its market capitalization could approach $7.5 trillion, compared to roughly $5.3 trillion today. However, valuation compression would limit its upside even if earnings rise.
Still, networking is no longer a peripheral business for Nvidia. As AI factories become larger and more communication-intensive, Nvidia's ability to sell more of the infrastructure surrounding its chips could become an increasingly important part of its next phase of growth.
