NVIDIA Pivots as Networking Stalls
Yes, $11B in Blackwell revenue is impressive. Yes, Nvidia's data-center revenue grew 93% year over year. Under the surface, however, there's trouble in networking. In the January quarter (Q4 FY25), networking revenue declined 9% year over year and 3% sequentially. In its earnings call, CFO Collette Kress said that Nvidia's networking attach rate was "robust" at more than 75%. Her very next sentence, however, hinted at what's happening underneath that supposed robustness.
"We are transitioning from small NVLink8 with InfiniBand to large NVLink72 with Spectrum-X," said Kress. About one year ago, Nvidia positioned InfiniBand for "AI factories" and Spectrum-X for multi-tenant clouds. That positioning collapsed when the company revealed xAI selected Spectrum-X for what is clearly an AI factory. InfiniBand appears to be retreating to its legacy HPC market while Ethernet comes to the fore.
![]() |
Nvidia Data-Center Revenue |
So how do we square 93% DC growth and >75% attach rate with a decline in networking revenue? We suspect the answer lies in the definition of attach rate, which likely counts only the NIC. In other words, >75% of Nvidia's GPUs were connected to ConnectX or BlueField NICs, but that doesn't mean they were also connected to Quantum or Spectrum switches. The primary alternative is ConnectX NICs using Ethernet as the interface to switch fabrics based on Broadcom's Jericho line.
With hindsight, the joint announcement with Cisco the day before Nvidia's earnings makes more sense. First, the partners will enable Cisco's Silicon One chips and systems to work with Nvidia "SuperNICs" under the Spectrum-X protocol/software umbrella. Note that Nvidia's SuperNIC nomenclature once equated to BlueField DPUs but now includes ConnectX-8 too. Thanks to Silicon One's programmability, Cisco plans to release Spectrum-X support by mid-2025. The longer-term plan includes Cisco switch systems built around Nvidia Spectrum silicon, enabling integration with NX-OS. Given both companies already have 51.2T switch chips and systems, the new Cisco platforms will likely target 102T or beyond.
The companies will probably fill in a few more details at GTC next month. With Ultra Ethernet coming soon, however, this expanded partnership provides Cisco and Nvidia the opportunity to strengthen their AI-networking positions relative to Arista and Broadcom.
Comments
Post a Comment