Why advanced network infrastructure is essential for AI growth and innovation

The tools of tomorrow are currently powered by the networks of today. As we emerge into the next tech-enriched era, defined by artificial intelligence, we’re discovering that advanced network infrastructure is essential for AI growth and innovation. In this piece, we’ll look at the role it plays in AI, the impact of intra data centre networks, empowering edge computing and where the challenges still lie.

Table of Contents

The role of network infrastructure in AI

AI applications like large language models (LLM) and machine learning (ML) tools push traditional cloud networks to their limits. As this technology improves, it’s clear that our networks will need to evolve as well. Networks with high-speed, low-latency capabilities to handle the intense data flow within and between data centres will become the norm and more innovation is still needed. Ciena reports, “Just like customised AI-specific processors like Central Processing Units (CPUs) and Graphics Processing Units (GPUs) are being developed, network technology innovation is also required to fully optimise AI infrastructure. This includes advances in optical transceivers, Optical Circuit Switches (OCS), co-packaged modules, Network Processing Units (NPUs), standards-based UEC and UALink-based platforms and other networking technologies.”

Intra data centre networks

AI needs more than the data centres of today can offer. Intra data centre networks that operate at speeds of 400Gb/s and above for AI will enable the handling of data-heavy tasks, necessary for all this real-time processing and efficiency. A wish list of standards created by the Ultra Accelerator Link (UALink) Promoter Group and Ultra Ethernet Consortium (UEC) include network-centric infrastructure like higher bandwidths, lower latency, resource isolation, specialised NPU and faster dynamic resource scaling. Networks operating at 400Gb/s and 800Gb/s to 1.6Tb/s and higher will be required in the future. And then there’s the actual location and distance between these data centres to contend with – leading to interconnectivity challenges and campus structures.

Inter data centre connectivity and edge computing

Data centres must be interlinked to enable distributed AI processing across vast distances using optical transport solutions. The monthly AI data load by 2030 is estimated by Omdia at 148 exabytes (compared to 0.6 in 2023). Edge data centres that bring this AI closer to users, will improve speed and reduce latency, making applications like autonomous driving and real-time analytics increasingly viable. But to do this, data centres will need to be built in clusters and closer to the end user – and this will be at odds with many user’s Not In My Backyard mentality. It’s a challenge that leaders will need to address so this valuable data can move cost-effectively and swiftly from core data centres to edge data centres.

Sustainability and power efficiency concerns

Location and tech are the only challenges for AI. Its high energy demands can’t be understated. While there are ongoing innovations aimed at making network solutions more power-efficient and sustainable, they are unlikely to keep pace with the rapid deployment of AI. Barclays Research sheds a bit of light on the problem: “After decades of almost non-existent demand growth for electricity in the US, the AI revolution is expected to more than double data centre electricity needs by 2030 based on current grid capacity.” And since AI needs a constant supply, renewables like wind and solar are likely not up to the task. Sustainability is a challenge that industry leaders will need to tackle, and soon.

Overall, it’s clear that advanced network infrastructure is foundational for AI to scale and thrive. While there are some issues with sustainability and data centre interconnectivity to overcome, one thing is clear – as AI’s reach grows, so will the demand for smarter, faster and more efficient networks.

To find out more, visit Ciena’s website.