NTT DATA’s Michael Wheeler talks widely on the trends and challenges at the heart of today’s industry
As AI workloads become increasingly distributed across data centres, clouds and regions, what role will high-capacity IP networks play in the ecosystem?
AI is an increasingly compelling part of the market ecosystem today. Data centres are probably feeling the biggest effect of that outside the physical infrastructure tied to compute, such as GPUs, and the growth in scale and investment in the sector is huge. While IP networks around the world may be seen as the ‘unsung heroes’ of that environment, they are critical to making sure the infrastructure being built out for AI can be used effectively, providing the ‘spinal cord’ that connects it all. As a result, I don’t think one part of this ecosystem is more important than another per se; the networks, data centres, software and GPUs all work together to form an environment in which both businesses and consumers are able to do more than ever before.
Where do you see the greatest pressure points in global internet infrastructure today?
With the way in which AI functions nowadays, you have two ‘buckets’ with quite different requirements – ongoing training, and the application of that training through inference. Both of these elements are critical, but I think the second is where we’ll see most evolution and therefore the larger pressure, because it’s less predictable and is happening extremely fast at this point. This will put the pressure points more to the edge at latency- and time-sensitive parts of the network, in metro- or regional-based rather than core subsea and terrestrial infrastructure. That will also happen as the need grows to be as rapidly responsive to the user as possible.
Operators of large-scale networks often highlight the number of points of presence they operate, but is it more important to have the largest possible footprint or highest-quality connectivity?
There are always going to be some trade-offs, but I think that perspective matters mostly based on the organisation’s particular business strategy. A tier-one carrier catering to both consumers and businesses on a more metro, regional or national level may want to be in more facilities to give itself a bigger footprint in its specific area. However, when it comes to a business like ours at NTT DATA, which provides connectivity to internet-centric businesses globally, we don’t intend to be in every single location. Instead, we’re focused on sites where the most internet traffic is being exchanged and provided to ISPs, telcos and content companies. I think what ultimately matters is are you generating returns for your shareholders and getting the results you need for your business strategy? After all, it doesn’t matter how many sites you’re in or where you are if you’re losing money.
With NTT DATA operating a major tier-one global backbone, where is the company seeing the strongest growth in demand for services today both geographically and by customer segment?
We’re seeing a lot of growth from the Asia-Pacific region, with a big uptick in demand for our 400G ports driving a large increase in traffic there. But North America and Europe remain very strong regions for us too. On the customer side, we continue to see demand and traffic from businesses across many different segments that put internet at the heart of what they do. By their nature, content delivery networks are large traffic generators, while major cloud companies running AI initiatives and the gaming sector have also continued to drive significant growth.
How do you see customer expectations evolving when it comes to network resilience, redundancy and security?
Resilience and redundancy are more critical than ever, with more and more reliance on the network to enable people to function in their jobs and tasks, particularly in the days of hybrid working since Covid. People need consistent connection to applications, services and software-as-a-service platforms. It is similarly crucial to keep ahead when it comes to security, with the threatscape evolving and bad actors harnessing AI to generate and manage attacks. At the same time, tools have become more effective to deal with attacks, meaning we have the means to continue evolving to make networks as secure as possible now and moving forward.