How AI is impacting global connectivity

The growing adoption of artificial intelligence (AI) is driving unprecedented demand for network capacity, but the connectivity infrastructure that powers the technology often goes unnoticed.

Speaking to Computer Weekly, Jürgen Hatheier, vice-president and international chief technology officer of Ciena, a networking systems, services and software provider, noted that AI for the telecoms world is still a mystery: “Nobody exactly knows what the impact on telecoms networks is going to be, except for the fact that there’s just going to be a lot more traffic.”

He pointed to the rapid pace of AI innovation, noting that even two years ago, the transformative potential of applications like ChatGPT was underestimated. The impending explosion of AI applications, from large vision models that process 4K video streams to personalised movie generation, will place unprecedented strain on existing networks.

Training the complex models that power these applications requires massive computational power, often located in datacentres strategically placed near renewable energy sources. Inference – the process of interacting with these models using real-world data – also generates significant traffic flow. Submarine cables, with their high capacity and global reach, are well-positioned to facilitate both training and inference workloads.

Hatheier acknowledged that while current capacity may suffice for today’s needs, future demands will require significant expansion. “Do we have enough capacity today? For today’s needs, yes, but for future needs, absolutely not,” he said, citing the 40-month lead time for new submarine cable builds as evidence of the burgeoning demand.

He also underscored the importance of strategic planning for service providers in the evolving AI landscape. “It’s not only about the technicalities, but also about the business strategy,” he said.

Ciena is advising telcos on how to position themselves within the AI ecosystem, moving beyond their traditional role as connectivity providers. He pointed to SKT’s success in developing its own AI chipset and platform as a prime example of a telco that has successfully navigated this transition.

In the Asia-Pacific region, Ciena is actively engaged with customers to support their growth ambitions. “We are very active on the subsea side,” Hatheier said, highlighting Ciena’s work on capacity upgrades for existing submarine cables. A single upgraded cable, he added, can now carry the equivalent capacity of the entire Starlink network.

Beyond subsea cables, Ciena is working with service providers to build terrestrial backhaul networks and facilitate the move towards network-as-a-service offerings. This involves modernising legacy systems, enabling automation and orchestration, paving the way for autonomous networks.

“We are helping telcos open up the networks and make them available as a consumable asset,” said Hatheier. This will enable cloud suppliers and application developers to leverage the network for innovative services, such as those provided by gaming companies that deliver unique experiences for users, he added.

Addressing the modernisation challenges telcos face in capitalising on the AI wave, Hatheier suggested a top-down approach, starting with the business vision and strategy before addressing the underlying technology stack. He also highlighted the importance of abstraction layers and the adoption of newer technologies like a 50-gigabit passive optical network (PON) to enhance agility and efficiency, as well as network slicing for critical applications such as emergency services.

“In every conversation around AI, including those with hyperscalers, the connectivity element isn’t getting much attention,” he said, adding that connectivity should be an integral part of planning, not an afterthought. “The industry has done a great job of connecting everybody, but we shouldn’t take it for granted, especially when technology innovation happens at a very fast pace.”