When 5G and AI become indistinguishable: NVIDIA's approach

When 5G and AI become indistinguishable: NVIDIA's approach

May 28, 2024

How is NVIDIA working to design and build solutions for the telecom sector? And how have NVIDIA's products evolved to meet the industry’s demand?

Artificial Intelligence is the most powerful technological force of our time. Enabled by 5G’s high-performance, ultra-reliable and highly secure connectivity, AI applied to the network edge will deliver connected intelligence across every industry. It will transform the business landscape and accelerate economic growth. This opens up new use cases in smart cities, smart manufacturing, smart retail, automated warehouses, and more. The NVIDIA AI-on-5G platform is spearheading the push to combine AI and 5G at the network edge to accelerate this digital transformation that is expected to create over $10 trillion in economic value.

Combining AI and 5G under NVIDIA’s AI-on-5G platform unlocks new commercial opportunities for enterprises, telecom companies, and other stakeholders, whether on premises, in the field or in the cloud. For enterprises, running AI applications at the edge on a high-performance 5G RAN is a key factor in realizing the Industry 4.0 vision of an interconnected and integrated system of cyber-physical systems. For telcos, deploying AI applications at the 5G edge creates new revenue sources, positioning AI to be the critical application for 5G and turning the discussion on 5G rollout from a capital expenditure discussion into a revenue and monetization discussion.

The AI-on-5G platform opens a new technical playbook. Today, 5G telecommunications infrastructure and the AI computing infrastructure are evaluated, deployed, and managed independently. This is an inefficient strategy, because both AI and 5G require computational oomph that can be provided by the same high-performance computing platform. Because of this, using the same computing platform to enable AI workloads to run seamlessly over a 5G network delivers both technical and cost efficiencies. This brings lower total cost of ownership (TCO) for equipment, power, and space, enabling the auto-scaling and pooling of compute resources. It also delivers higher security for AI.

NVIDIA AI-on-5G addresses this opportunity. Based on the NVIDIA EGX™ class of servers, it’s a single, hyperconverged, GPU-accelerated, edge “data center in a box” that can deliver high performance for both AI workloads and 5G RAN. With the AI-on-5G platform, the 5G RAN is implemented as an additional software stack on the GPU-accelerated computing platform. This makes it possible for an enterprise to add and manage 5G connectivity as part of their IT infrastructure and a telco to transform all 5G gNBs into edge data centers. The five building blocks of the NVIDIA AI-on-5G platform are:

  1. The EGX server
  2. The NVIDIA Aerial™ 5G vRAN software development kit (SDK)
  3. The Aerial A100 card,
  4. NVIDIA’s portfolio of edge AI applications, including NVIDIA Metropolis™ and NVIDIA Isaac™
  5. NVIDIA’s partner ecosystem of equipment makers, independent software vendors (ISVs), developers, and more.

Artificial Intelligence is the most powerful technological force of our time.

NVIDIA has proven to be a power player in several sectors: GPUs, data centers, CPUs and AI, to name a few. Now, it is aiming for 5G. Why? What does 5G mean to NVIDIA?

NVIDIA targets the most complex and demanding workloads to accelerate. NVIDIA has proved its value in Graphics, High performance computing and AI. NVIDIA is often the top performer for accelerating parallel computation workloads such as AI. This is published in the MLPERF benchmarks tests.

5G is a demanding workload, especially when it comes to mid-band and high-band frequencies. The NVIDIA stack of chips and systems and its AI operating software CUDA is highly suited to solving this computing challenge. The future market is worth trillions of dollars if we consider that every 5G mast and tower has the potential to be an AI + 5G mini data center. Furthermore, 5G is essential for modern AI workloads such as Computer Vision for Industrial Inspection Smart Cities to be adopted and deployed at scale. Graphical workloads such as photorealistic AR and VR can only be rendered on edge compute nodes with GPUs, the transport has to be high bandwidth and low latency. Only 5G makes this possible at scale. AI on 5G is essential to our customers and our ambition.

Running AI applications on 5G will make the vision of AI applications anytime, anywhere come true. It will bring extended coverage, mobility support, configurable quality of service, improved reliability and enhanced security. This will enable, for example, more powerful deep learning training and inference — making AI better able to guide traffic flows, route autonomous vehicles, make factory robots more efficient at picking and packing goods, and much, much more.

Could you share NVIDIA’s roadmap for 5G? What kind of solutions and products are you offering or planning to offer in the future?

I recommend watching Ronnie’s address on July 1 about AI on 5G or Jensen’s keynote chapter about AI on 5G. Our Aerial SDK development roadmap is available under NDA.

Is 5G just the very first step towards a future where edge computing solutions by NVIDIA are everywhere?

As mentioned above, 5G is a huge enabler for AI adoption and Edge Computing. Our platform and SDK are open to all. The infrastructure is supported in all the major cloud providers and OEM. Our GPU scales from the device edge with Jetson all the way to the Data Centre on a single architecture. For a real glimpse into the future, I recommend taking a look at NVIDIA Omniverse.

I hope this has offered good insight into why we are so focused on 5G and how we are enabling the ecosystem.