While market attention remains fixated on NVIDIA's AI processors, CEO Jensen Huang has quietly built a multibillion-dollar secondary pillar for the company. The data center networking business, strategically established through acquisitions in 2020, has become one of NVIDIA's most profitable and fastest-growing divisions.
Within just a few years, the networking segment designed to interconnect data centers has evolved into NVIDIA's second-largest revenue driver, trailing only its computing business. According to NVIDIA's latest financial report, the division generated $11 billion in revenue last quarter, representing a 267% year-over-year surge, with annual revenue exceeding $31 billion.
This remarkable growth rate and scale have fundamentally reshaped the competitive landscape of the networking equipment market. Kevin Cook, senior equity strategist at Zacks Investment Research, noted that NVIDIA's quarterly networking revenue now exceeds the full-year networking revenue projections of established giant Cisco Systems.
Foundation for AI Factories The explosive growth of NVIDIA's networking business directly benefits from soaring AI processing demands. The division's technology portfolio includes NVLink for connecting GPU racks, InfiniBand Switches for in-network computing, Spectrum-X for AI networking over Ethernet, and co-packaged optical switches.
These technologies collectively provide all necessary infrastructure for building "AI factories"—specialized data centers dedicated to training AI models.
Kevin Deierling, NVIDIA's senior vice president of networking, stated: "People typically think of networking as 'I have a printer and need to connect it.' But from day one of our acquisition, Jensen Huang declared that the data center is the new computing unit. Networking isn't just moving bits between compute nodes—it's actually the foundation."
Vision Behind the $7 Billion Acquisition This massive business originated from NVIDIA's $7 billion acquisition of Israeli networking company Mellanox in 2020. Deierling joined NVIDIA through this acquisition.
Initially, Deierling didn't fully understand Huang's rationale for acquiring Mellanox at that time, but he now comprehends the strategy. Owning networking capabilities enables NVIDIA to bundle GPUs with optimally matched networking technology.
"When Jensen Huang acquired Mellanox in 2020, he recognized this as the missing piece to make GPUs a complete solution," analyst Cook commented.
Beyond technological leadership, the success of NVIDIA's networking business also stems from its unique business model. Deierling emphasized that NVIDIA sells these technologies exclusively as full-stack solutions rather than individual components, distributing them through partners.
"I can't think of another company with our full-stack capabilities," Deierling said. "We've built the complete computing stack, a fully integrated stack, then brought it to market through all our partners."
At the Nvidia GTC technology conference on March 16, NVIDIA further strengthened this advantage by introducing a series of networking system updates, including the Rubin platform (featuring six new chips), Inference Context Memory Storage platform, and more efficient Spectrum-X Ethernet Photonics switches.
"Networking is no longer an external device for connecting printers or other slow I/O devices," Deierling emphasized. "It's the foundation of the computer. In the past, computers had backplanes inside. Today, the network serves as the backplane for AI factories—it's absolutely critical."
Comments