Nvidia’s Mellanox bet keeps paying off as quarterly networking revenue hits $11 billion
The Israeli unit has become critical to Nvidia’s dominance in AI infrastructure, powering the surge in data-center revenue.
When Nvidia paid $6.9 billion in 2020 to acquire Mellanox Technologies, an Israeli specialist in high-speed communications chips, the deal was framed as a strategic complement to its dominant graphics processors. Six years later, that acquisition has become one of the most consequential bets in the company’s history.
In its latest earnings report, Nvidia disclosed that its networking division generated $10.98 billion in quarterly revenue, a 263 percent increase from a year earlier. For the full year, networking revenue exceeded $31 billion, more than ten times its level in fiscal 2021, the year the Mellanox acquisition closed.
The scale of the growth reflects a shift in the economics of artificial intelligence infrastructure, where the ability to move data between processors has become as important as the processors themselves.
“Networking, a cornerstone of our data center scale infrastructure offering, was a standout this quarter, generating $11 billion in revenue, up more than 3.5x year-over-year,” Nvidia’s chief financial officer, Colette Kress, said on the company’s earnings call. Demand, she added, was driven by record adoption of Nvidia’s networking technologies, including NVLink, Spectrum-X Ethernet and InfiniBand.
Six months ago, the division was already expanding rapidly, generating $7.25 billion in quarterly revenue and nearly doubling year over year. At the time, it was clear that Mellanox’s technology had become deeply embedded in the infrastructure powering AI data centers.
The latest results show that acceleration has intensified dramatically. The division’s $10.98 billion in quarterly revenue now represents a scale comparable to that of major standalone semiconductor companies.
Much of that activity is concentrated in Israel, particularly at Nvidia’s research and development center in Yokneam, established through the Mellanox acquisition. The site serves as the company’s primary hub for communications chips, which connect thousands of processors into unified computing systems.
These components perform a critical function: allowing vast numbers of AI processors to operate as a single machine. Without high-speed networking, the computational power of individual chips cannot be fully utilized.
The surge in networking revenue reflects the rapid build-out of AI data centers by large technology companies. Hyperscalers are expected to spend at least $630 billion on infrastructure in 2026, much of it devoted to processors and the systems that connect them.
As AI models grow larger and more complex, they must run across thousands of interconnected chips. Networking technologies such as NVLink and InfiniBand enable these processors to communicate with minimal delay, making large-scale AI computation possible.
Kress said growth was driven primarily by NVLink scale-up switches used in systems built around Nvidia’s newest generation of AI infrastructure. These technologies allow customers to integrate massive numbers of processors into unified computing environments.
“NVLink scale-up fabric has revolutionized computing and demonstrates the power of extreme co-design across all of the chips of the supercomputer and the full stack,” she said.
Nvidia also announced it would enable Amazon Web Services to integrate NVLink into its custom silicon.
The company reported January-quarter revenue of $68.13 billion, up 94 percent from a year earlier, as demand for AI infrastructure continued to expand.
CEO Jensen Huang said the shift toward AI-driven computing was structural and unlikely to reverse. “This new way of doing computing is not going to go back,” he said.
