Nvidia Builds Multibillion-Dollar Networking Division That Rivals Its Chip Business

Nvidia's networking division has rapidly emerged as a powerhouse within the company, generating staggering $11 billion in quarterly revenue with a remarkable 267% year-over-year increase.

“Sentiment Remains High Beyond optics, GTC is drawing participation from across the enterprise AI world”

24/7 Wall St.

This explosive growth has positioned the networking business as Nvidia's second-largest revenue driver behind compute, creating a formidable second franchise that few industry observers fully recognize.

Benzinga

The division now represents a critical piece of Nvidia's infrastructure dominance that Wall Street has largely overlooked amid the usual frenzy over GPU shipments and gaming metrics.

What makes this achievement particularly notable is that it has achieved this scale in just a few years, transforming from a strategic acquisition into a multibillion-dollar behemoth that would rank as a Fortune 500 company on its own.

The origins of Nvidia's networking dominance trace back to its 2020 acquisition of Mellanox Technologies, an Israeli-founded networking company, for $7 billion in what appeared at the time to be an expensive but prescient move.

This deal brought critical technologies including InfiniBand, NVLink for GPU communication, Ethernet switching platforms like Spectrum-X, and co-packaged optics switches that form the backbone of modern AI data centers.

Bloomberg Línea

Together, these technologies provide all the essential components needed to build what Nvidia calls an 'AI factory'—a data center specifically designed for training AI models.

The timing of this acquisition proved perfect, as companies rushed to build AI infrastructure only to discover that networking bottlenecks were severely limiting performance, making high-speed networking essential for even the most powerful GPUs to operate effectively.

In the competitive AI networking landscape, Nvidia has established a significant advantage through its full-stack approach that provides architectural control across GPUs, interconnects, DPUs, switches, optics, and software.

“Lea más en Bloombergc”

Bloomberg Línea

This comprehensive integration allows the company to deliver complete, supported systems that address the specific needs of hyperscalers building GPU-rich clusters and enterprises deploying AI applications requiring predictable latency.

While competitors like Broadcom, Marvell, and Cisco advance 51.2T and 102.4T Ethernet switch silicon, and AMD's Pensando and Intel's IPU initiatives target offload and data plane acceleration, Nvidia's strength lies in its ability to offer a unified solution.

Kevin Cook, a senior equity strategist at Zacks Investment Research, highlighted this competitive edge by noting that Nvidia's networking business now does 'in one quarter what Cisco's business does in a year,' generating greater revenue than even Cisco's entire networking operation.

The strategic importance of Nvidia's networking division extends far beyond mere revenue generation—it has become essential to amplifying the value of the company's core GPU business.

Without high-speed networking infrastructure, even the most powerful GPUs sit idle, waiting for data transfers that severely limit performance.

CNBC

Nvidia's networking technologies serve as the 'circulatory system' for AI clusters, enabling the seamless data flow that makes modern AI processing possible.

This positioning allows the networking division to complement and enhance rather than compete with the chip business, creating a powerful synergy within Nvidia's overall AI infrastructure strategy.

The division's focus on building 'AI factories'—data centers specifically designed for AI training—further solidifies its role as an indispensable component of modern AI infrastructure.

Looking ahead, Nvidia's networking division appears well-positioned to capture a growing share of the rapidly expanding AI infrastructure spending market.

“At Nvidia's annual developer conference on Monday, CEO Jensen Huang took the stage to a packed house and said he expects purchase orders between Blackwell and Vera Rubin to reach $1 trillion through 2027”

CNBC

Analysts expect continued growth as companies increasingly recognize that networking represents a critical bottleneck in AI deployment that must be addressed alongside compute capabilities.

CoinDesk

The division's focus on next-generation technologies like co-packaged optics, which replace copper interconnects with light-based data transmission, indicates a forward-looking approach to addressing the scaling challenges of AI infrastructure.

With the 'inference inflection point' now upon us—where AI systems are transitioning from training to production deployment—the demand for robust networking solutions is expected to accelerate further.

Nvidia's ability to tightly integrate its networking solutions with its GPU roadmap and aggressive development cycles suggests the company is positioned to maintain its leadership in this critical segment of the AI ecosystem.