Nvidia Steals the Show at the Edge

Smart City Paris

By: R. Scott Raynovich


I did not make it down to Los Angeles this week for the Mobile World Congress event. But from what I saw, Nvidia's huge move into the edge and an its announced alliance with Ericsson, Microsoft, and Red Hat is the most important news of the show.

Nvidia unveiled the EGX Edge Supercomputing Platform at MWC this week. Originally announced in May, EGX includes Nvidia's high-performance gear for low-latency apps such as Artificial Intelligence (AI), combined with with high-performance network interface cards (NICs) from Mellanox, which Nvidia is in the process of buying (but it's still awaiting government approvals).

Nvidia is also partnering with Ericsson on software-defined radio-access networks (RAN), Red Hat OpenShift for Kubernetes container orchestration as well as Red Hat's telco platform, and Microsoft on the Azure Cloud.

This is huge. This is a strong alliance, with a compelling vision for the edge. It revolves around building an architecture with high performance infrastructure for distributed apps.

It's About "Smart Everything"

Nvidia founder and CEO Jensen Huang declared this as part of the "smart everything revolution" in his keynote on Monday.

"5G is set to turbocharge the intelligent edge revolution," Huang told the crowd at MWC, according to ZDnet. "Fusing 5G, supercomputing, and AI has enabled us to create a revolutionary communications platform supporting, someday, trillions of always-on, AI-enabled smart devices."

This comes on a week in which dismal earnings news from Nokia shows that the communications infrastructure market is not going to get any easier. Everybody is looking at 5G and the "edge" as the next opportunity, but many people have trouble espousing a vision of how it works and why they're the leader.

Nvidia has a story. It comes down to technical details around hardware-based performance and integration with cloud-native software.

The edge is hyped by the marketers for sure, but that doesn't mean they're wrong. There is huge opportunity here as bandwidth and apps gets fueled by 5G and more computing power closer to mobile users, machinery, and vehicles. But the problem with hype is that it is broad-based, and figuring out who wins comes down to execution details.

Nvidia has a real product with real customers. Reference customers already include Walmart and BMW. Walmart can use EGX to compute more than 1.6 terabytes of data generated each second. Some applications include using AI to alert staff to needs for shelf re-stocking, or monitoring store services such as checkout lines. BMW is using EGX to power AI-driven video analytics to automate inspections, using EGX.

This partnership has depth and breadth in its understanding of the edge. Red Hat, along with VMware, is a leader in virtualized technologies for communications environments such as Kubernetes and container-based applications, which will be critical to the edge because of their capabilities to distribute computing power. Nvdia's new software platform, the Nvidia Edge Stack, is optimized for container-based applications.

And on the virtualized 5G front, Nvidia also made a savvy move by partnering with Ericsson. Huang said Nvidia has been working with Ericsson on technologies for virtualized 5G radio access networks (RAN), which was also announced this week. Nvidia and Ericsson are touting this as the world's first software-defined 5G RAN. This is important because past generations of RAN have been largely proprietary and hardware-driven, so this will be the first generation of cloud-native RAN. A software developer kit (SDK) called Nvidia Aerial will allow providers to build their own software-defined 5G wireless RAN.

High-performance NICs

The other key part of this edge ecosystem is Nvidia's forthcoming acquisition of Mellanox, assuming it can get done. Some of the hang-up on the deal has been attributed to waiting for Chinese government approval, so the improving tone of trade relations should help push this over the line.

Mellanox takes a unique approach to optimizing NICs to eliminate the computing overhead of virtualization and deliver blazing performance that has been a differentiator for low-latency AI apps in the data center. As these apps move to the edge, the NIC will be just as important, if not more so. Nvidia was smart to see the potential, and pounced. Now the vision is expanding. Nvidia sees the power of taking its powerful AI chips to the edge, integrating with Mellanox NICs, and then integrating with the virtualization and software platforms from Red Hat and Azure.

Nvidia's EGX announcement was well-timed, thorough, credible, and connected to the right partners. I haven't seen another edge announcement with this type of breadth. The rest of the industry is now going to be scrambling to come up with a more powerful edge story. And that will be difficult.