Thank you for inviting me. I am the chief executive officer of Digital Infrastructure Ireland. By way of background, I have spent 31 years working in Ireland’s tech sector, so I have watched the evolution of the sector to the point that it is at now, where it contributes roughly 30 per cent per annum to the Government’s corporation tax take for the whole of the Republic of Ireland—it is a significant contributing factor to that.To talk to the origins and growth of the Irish data centre industry, one has to cast the lens back to the 1990s, because that was the beginning, when investments were made predominantly in the Dublin area to serve as the foundation for an evolving tech sector. That began with significant investment in a major fibre-optic network. Around Dublin, we have the M50 motorway, which is similar to the M25 around London—it kind of circles the city. On the periphery of the M50, a 16-duct fibre network infrastructure called the T50 network was built, and that network was woven in and out of every major industrial park on that M50 route. Those industrial park locations are where the majority of the data centre facilities that are based near or adjacent to Dublin have been deployed, and that has resulted in a reasonable concentration of data centre facilities in the greater Dublin area.Over the years, and particularly in the past 10 years, we have encountered a misalignment with Government investment policy on the national grid. We built the fibre infrastructure to attract the facilities, and they came as a result of that investment, and they continue to grow. However, 10-plus years ago, we put a pause on investment in the national grid, much of which was to do with the policies around doing less of everything to allow us to align better with our green targets that were coming from Europe. That included making less investment in the grid, so it was getting about €1.5 billion a year by way of investment, which is just about enough to keep the thing running but does not allow for any upgrade or expansion in capacity.The sector continued to grow and workloads continued to evolve. I was delighted to hear both the previous witnesses make the distinction between co-location facilities, which tend to house the compute resources of organisations that are proximate to the facilities, be those private enterprises or Government agencies, versus the hyperscale facilities, which quite often just house compute resources that belong to the owner of that facility. Along the route of the T50 network, there is a significant concentration of co-location facilities, many of which have already secured the status of critical national infrastructure. Because of the workloads that they house, they have been deemed to be critically important.The workloads have evolved over the years. Many of the data centre facilities in Ireland that emerged in the early 2000s began life serving up a means of software distribution and hosting some light compute workloads. The workloads have evolved significantly towards cloud-based workloads. Now, the vast majority of data centre facilities that are deployed across Ireland are responsible for hosting cloud-based workloads.The number of AI-focused data centres that Ireland has deployed today is actually zero. What is the difference between a cloud-focused data centre and an AI-focused one? It is predominantly about the power density that is deployed in the data centres. Typically, we measure that by the number of kilowatts per rack of capacity that is available to drive the compute resources deployed within each rack, and many of these data centres would house up to 1,000 racks of equipment, depending on their size. Typically, a cloud-focused data centre would have in the range of between 5kW and 15kW of capacity per rack, but what we are starting to see now in the AI world is the on-ramp for an AI-focused data centre being predominantly in the range of 70kW per rack, and that is rapidly increasing to 120kW for the newer NVIDIA Blackwell 200 series chipsets and onwards up to 150kW per rack for the 300 series.We are part of the European Union Data Centre Association, and I have sat in on its recent presentations. Recently, it invited NVIDIA to give a presentation, in which it talked about a roadmap towards 600kW per rack and a future density of 1,000kW—that is, a megawatt—per rack. The existing data centre facilities do not lend themselves to that sort of power density. The size of the grid connection that they have is not sufficient to support it. Equally, a data centre facility that has been designed for 1,000 racks could accommodate 10 of those future-state racks. You can see that Ireland faces some fundamental challenges if it wants to play a role in relation to the newer AI data centre facilities that are emerging.It was fascinating to hear the earlier conversations around sovereignty, and I note that we are seeing a push from the European Union towards sovereignty. The EU has announced that Europe needs to triple the number of data centres that it has in order to maintain its sovereignty status and be competitive in relation to its international peers in the US and China. It has also identified a need to build seven AI gigafactories, which are data centre facilities that house in excess of 100,000 graphics processing units. When you equate that down to megawatt scale, that is basically an entry level of about 250MW, which eclipses anything that is deployed in Ireland at this point in time.The way I look at things is that the world is at a time of transition with regard to digital workloads. The digital workloads that have serviced our societies for the past 20 or so years will continue to evolve, but there are newer, much more power-intensive workloads coming through that are predicated on the AI investment wave. Some of those workloads will be housed in dedicated AI data centres. However, equally, we will start to see smaller language models being used. Earlier, we heard reference to the concept of large language models, which are the very large compute deployments that have been used by the AI systems to date, but there are also small language models—or small large language models, as they are described. You will start to see enterprises applying their own AI workloads on their own compute resources, many of which will be in these co-location data centre facilities, and some of which will be in the hyperscale data centre facilities as well. That will put increasing upward pressure on the existing facilities with regard to power demands, too. Therefore, grid upgrades are going to be required just to house what is already there today and to continue to service the growth.I liken the situation to the transition that took place from the very beginning of the internet era, when we were on dial-up connections and we had limited time on the internet because other people wanted to use the phone and so on, to when we went to always-on broadband over copper networks and then on to fibre networks. Those changes fundamentally altered how people interacted with this new thing called the internet. If you go back to the start of that process, you will see that no one was able to predict with any degree of accuracy the ways in which we would all end up using internet and cloud-based services into the future.The exact same situation exists today with AI. Nobody can accurately predict what is going to happen. We all have ideas on where it is going to be able to contribute towards society, and there are lots of ideas about where it is going to detract from society, but nobody can make those predictions with any degree of accuracy. However, we are starting to see some novel innovations that will help society, such as new antibiotic breakthroughs that have been fundamentally driven by AI. A lot of work cases are coming through, but the reality is that we need the joined-up planning to be able to support the facilities if we want Ireland—or Scotland, in your case—to be part of that next wave of technological advancement.