Nvidia is quietly building a multibillion-dollar behemoth to rival its chips business
Nvidia's networking division hit $11B in quarterly revenue, nearly matching Cisco's annual business.
Nvidia has quietly built a networking empire that generated $31 billion in revenue last year, making it the company's second-largest business unit behind its core compute division. This growth stems from the strategic $7 billion acquisition of Mellanox in 2020, a move CEO Jensen Huang made to complete the 'AI factory' stack. Last quarter alone, the networking segment reported $11 billion in revenue—a staggering 267% year-over-year increase. According to Zacks Investment Research analyst Kevin Cook, that single-quarter figure surpasses Cisco's entire networking business and nearly matches what Cisco does in a full year.
The division's technology is critical for modern AI infrastructure. It includes NVLink for high-speed GPU-to-GPU communication, the InfiniBand and Spectrum-X platforms for AI-optimized networking, and co-packaged optics switches. Senior VP Kevin Deierling notes that Nvidia sells this as a fully integrated, full-stack solution through partners, not as individual components. This approach, combined with the explosive demand for AI processing, has transformed networking from a connectivity afterthought into what Huang calls 'the foundation' of the data center as the new unit of computing. The business provides all the necessary tech to construct dedicated 'AI factories' for training large models.
- Nvidia's networking business generated $11B in Q1 revenue, a 267% YoY increase, totaling $31B for the fiscal year.
- The division, built from the $7B Mellanox acquisition, now surpasses Cisco's entire networking business in quarterly revenue.
- It provides the full-stack tech (NVLink, Spectrum-X) needed to build 'AI factories'—specialized data centers for model training.
Why It Matters
Nvidia now controls both the AI brains (GPUs) and the nervous system (networking), locking in dominance over the entire AI infrastructure stack.