Mark Bieberich, Vice President of Portfolio Marketing at Ciena, details how hyperscalers’ regional build-outs and booming subsea routes are magnifying demand for ultra-scalable, automated DCI and cloud on-ramps.

In an era defined by digital transformation and, increasingly, AI, the demand for high-capacity, low-latency connectivity has never been greater. This is where wavelength services (also referred to as ‘wave services’) come in. 

Wave services are dedicated, high-capacity connections over a shared optical fibre infrastructure. They form the foundation of high-capacity wireline communications, enabling connectivity between data centres and cloud regions, and across continents through submarine and terrestrial networks.

AI-driven data centre expansion, cloud services growth, and data sovereignty concerns are all shaping the wave services market landscape. The stage is now set for a much denser and modernised global connectivity fabric in years to come.

AI is fuelling the demand for data centre connectivity

According to a recent study, data centre experts predicted that 43% of new data centre facilities will be dedicated to AI workloads by 2030. Given the need for more power to feed the power-intensive AI compute capacity, the location of these new data centres is transforming from where they would traditionally be sited, to areas with more power availability. 

For AI training, which is often the more power-intensive task, data centres are increasingly being constructed in remote areas, closer to power sources. This proximity enables tighter integration with the power grid, eases the burden on urban infrastructure, and reduces the need to transmit electricity over long distances. 

In contrast, inference compute is being built closer to the end user, whether in regional or even edge data centres. The location of inference ultimately depends on the specific requirements of each use case – and there are plenty of new ones emerging every week, pushing for more distributed compute sites requiring low-latency, high-capacity connectivity.

This decentralisation, and expansion of an interconnected, global AI fabric, demands increasingly high-performance connectivity. A new generation of wave services is driving this shift by facilitating data centre interconnect (DCI), providing access to AI clouds through ultra-scalable cloud on-ramp connections. 

Moreover, beyond wave services alone, this shift requires collaboration across the entire ecosystem – from hyperscalers to data centre and network providers – to ensure AI demand is being met.

Hyperscalers pivot their cloud strategies 

Another key driver shifting the demand for wave services is data sovereignty. As governments tighten regulations around where data can be stored, processed, and transmitted, hyperscalers deploy cloud infrastructure in more regions, creating new sites that need to be interconnected to their networks while acting as new access termination hubs for regional connectivity. 

Wave services play a key role in facilitating DCI required to integrate these new data centres with existing ones, and to regionally connect customers through cloud on-ramp connections.

It is not only about connecting new sites but doing so with much higher capacity to meet the requirements of emerging AI applications. AI workloads are becoming more multimodal, involving more bit-heavy image and video data on top of the text data of early genAI apps, like AI chatbots. 

Use cases are quickly multiplying, with AI being increasingly rolled out for every aspect of our work and digital lives. Take the example of Google, which saw a 50x increase in the volume of tokens processed by their AI products and APIs in just the last year. 

To enable the growth of AI applications and the utility of billions invested in AI infrastructure, the network must be designed to support wave services at increasingly higher capacity levels.

Terrestrial and submarine infrastructure expansion magnifies interconnect need

As AI workloads multiply and data centre proliferation soars, the growth of undersea cables is booming, with over 160,000km of new submarine cables planned for this year. 

New landing stations terminating these cables represent additional terrestrial network opportunities, as they need to be connected to regional data centres to backhaul submarine traffic. Wave services serve both submarine and terrestrial networks, providing reliable, scalable, and cost-effective capacity over distances spanning several thousands of kilometers. With heightened demand for securing submarine networks, they can also be encrypted, forming an additional layer of data protection for critical applications. 

Earlier this year, Lumen successfully established a 1.2 terabit wavelength service trial across 3,050 kilometers on Lumen’s Ultra-Low-Loss (ULL) fiber network, making it the world’s longest 1.2 terabit non-regenerated signal. Using 800G interfaces, Lumen successfully tested and qualified the services to support wavelength, Ethernet, and IP services over the 1.2 Tbps single carrier channel. This real-world trial illustrates what’s becoming commercially viable now, and what is likely to be mainstream within two to three years.

Building smarter networks for tomorrow’s demands

Looking ahead, as data centre network architectures decentralise and cloud-native AI applications expand, the demand for adaptable, high-performance connectivity will only increase. Service providers must rise to the challenge, leveraging AI-driven automation to activate wave services faster, assure service levels more reliably, and adapt to increasingly unpredictable AI traffic demand. 

In fact, leveraging AI to modernise and expand wave service portfolios is essential to meeting AI-fueled customer demands, whether from enterprises, governments, or other service providers. This trend is indicative of a new generation of wave services: smarter, more flexible, more scalable offerings that underpin the rapid growth of a globally interconnected AI network fabric.