How AI is changing the way we work with GCNs

By Robert McCabe, Director of GCN Sales at Colt

The IT and telecommunications sector is said to have the highest AI adoption rate at 29.5% and this is driving a massive increase in both bandwidth requirements and the need for low-latency connections close to end users.

As a partner to the world’s biggest global content networks (GCNs) and hyperscalers, we’re seeing a shift in their needs driven by the impact of AI and its massive compute requirements. As this drives demand for scalable optical networking services, GCNs are looking for greater base infrastructure such as dark fibre, cable and duct, all of which require significant construction.

AI workloads can be broadly split into two systems, training engines and inference engines, each of which have different infrastructure requirements that we’re enabling:

  • Training engines
    Training engines manage the learning stage such as analysing datasets and they require massive computational power. GPUs are needed to process this data which is why we’re seeing customers looking for more data centres.
    The training stage is less sensitive on location, so real estate can be cheaper and further from metro areas. These are large sites which require significant fibre connectivity and it’s where we’re working with customers on the construction of new routes, often over considerable distances, in a highly competitive market.
  • Inference engines
    Inference engines are more location sensitive as there is a need to reduce latency for the end user. To infer information for presentation to a user, based on a set of rules and data, then represent that back as part of a user experience or service is incredibly latency bound. AI use cases of this sort, as part of an end user service, will live or die by its ‘live’ feel.
    This is driving demand for data centres or real estate closer to metro areas to meet tight latency SLAs (service level agreements). Our customers now need denser metro networks, often requiring construction as physical routes require full diversity from each other.

How we’re meeting demand

As a specialist digital infrastructure company with over 30 years’ experience, we’re used to enabling the latest trends for some of the world’s most demanding businesses. While generative AI might be new, the need for network breadth and depth is familiar.

In this case, we’re helping customers with both aspects, partnering with them on new route builds and proactively investing in our network to meet future demand. For our GCN customers, this can mean buying fibre at extreme scale in the long-haul or metro areas or enabling less tenured infrastructure consumers with managed optical networks, spectrum services or wavelengths. These partnerships are key to our mutual, long-term success and it enables Colt’s network growth.

As demand from our GCN customers increases, so does the complexity and commercial scrutiny required. Our approach is one of partnership, aligning closely with our customers at multiple levels in the organisation to ensure that all key parties are involved at the right stages and customers have direct contact with the right experts. The Colt GCN sales team utilises the expertise of our Engineering, Sales, Network Growth and Commercial teams, all working together to make it simple for the customer. Our intention is to serve our customers with a flat consultative organisation with a pioneering peer-to-peer approach.

As AI adoption accelerates, demand for digital infrastructure will only continue to grow. Our partnerships with the world’s biggest GCNs and hyperscalers are delivering the long-distance and metro connectivity required, all through close partnerships.

To find out more information regarding our infrastructure and backbone services, head to our website.

 

What's your goal today?

1. Are you on the Colt IQ network?

Our network connects over 31,000 buildings worldwide powering companies such as Hitachi, Atos, Forbes, Arthur D Little, Brussels Airlines and thousands of others. Find out if you're Colt connected now.

2. Learn about digital infrastructure

We've written thousands of guides and white papers, regularly publish content on our blog and host regular events on everything from enterprise network connectivity, to cloud, digital transformation and the hybrid workforce.

3. Join our team

To learn more about joining our team of over 5000 people around the world, and to browse our current open roles visit https://careers.colt.net/.