How NVIDIA is leveraging Arm’s 5G infrastructure ecosystem to build the AI-on-5G platform

The convergence of 5G and AI at the edge has the potential to transform the world just as railroads and the internet did in earlier transformations. Intelligent devices and edge data centers featuring orders of magnitude leaps in bandwidth, latency, and the number of connections will pave the way to AI applications that deliver immersive experiences, industrial automation, and other services we cannot yet imagine. It will be a world swimming in bandwidth and AI compute.

And customers are responding. 5G will account for over 3.5 billion subscriptions by 2026 accounting for 40% of all mobile subscriptions, according to the latest Ericsson Mobility Report, while some large companies are experimenting with 5G to manage office space in a post-COVID world. 

AI workloads dominate the datacenter today. Multiple verticals like manufacturing, retail, healthcare are reaping significant benefits already.  

For example in manufacturing AI is being used to control machines, predict glitches in supply chains, detect defects and automate quality assessment. 5G enabled smart factory of the future will be a fully connected experience: connectivity will improve safety and quality while reducing the problem of unplanned downtime.

To realize the promise of 5G/AI convergence, it helps to look at some of the key technology requirements:

1. Better performance and power efficiency - the compute performance necessary to meet the network and AI workloads will go up exponentially, but the stringent cost and form factor requirements will mean power has to stay flat.  
2. Scalability – Rapid and straightforward deployment will be absolutely critical. Customers will need the ability to scale from small edge compute elements all the way to hyperscale cloud in a unified hardware and cloud native software environments.
3. Ecosystem - A broad software ecosystem that builds the various elements of the solution to realize applications that leverage 5G and AI. No one company has all the answers. 

Arm and NVIDIA are natural partners to realize the promise and convergence of 5G and AI. We have a shared vision, as well as complementary technology stacks, and a long-term roadmap that enables a hyper-converged infrastructure with 5G network building blocks and AI applications in highly integrated edge datacenter SOC.

• Arm CPU portfolio - Arm is the leader in providing the highest performance/power for a broad range of applications from the edge all the way to hyperscaler cloud datacenters. The increasing compute requirements at the networking edge demands scalability and specialized compute. Arm Neoverse CPUs enable Arm partners with high performance low power solutions to meet the fast evolving requirements.  
• NVIDIA AI-on-5G BlueField converged accelerator roadmap combines industry-leading NVIDIA DPUs and powerful Arm CPU computing with Nvidia GPU platforms, providing the full infrastructure-on-chip programmability and high-performance networking. Looking ahead, NVIDIA is focused on integrating functions, increasing performance, and lowering power consumption. This roadmap will enable AI + vRAN on a single card, even deliver AI services and vRAN on a single SoC. NVIDIA is committed to the aggressive silicon roadmaps to accelerate innovation for AI-on-5G. 
• Software & ecosystem support are key to realizing the AI-on-5G convergence. 

NVIDIA and Arm plan to onboard vRAN software providers to add enterprise and carrier-grade solutions for stack, edge, and core networks.  

NVIDIA Hyper-converged stack on BlueField and Arm Neoverse servers

Conclusion

For 5G to succeed, we need to get the underlying technology right. The highly dynamic market makes stringent demands on compute power efficiency, scalability, and software ease of use. Arm and NVIDIA are excited to collaborate in bringing AI on 5G convergence.

[CTAToken URL = "https://www.arm.com/solutions/5g/infrastructure" target="_blank" text="Visit Arm 5G Infrastructure Solutions Page" class ="green"]