LEE MYALL AI INFRASTRUCTURE
The UK government has announced a new programme of investment in AI and the data centre market in conjunction with a number of hyperscalers, however the success of the initiative will depend for a large part on having adequate optical infrastructure. Lee Myall , CEO at Neos Networks, argues that bringing the UK’s infrastructure up to the required standard will not only involve building fibre to the edge, but also building new backbone networks, replacing ageing infrastructure and providing access to green energy in areas where it is most available. The UK’s AI strategy is ambitious, but realising it will take massive achievement and effort from both the private sector and government. BUT ACHIEVING IT WON’T BE EASY THE UK’S AI STRATEGY IS AMBITIOUS
HOW THE UK’S NETWORKS MUST EVOLVE TO MEET DIGITAL AMBITIONS AND UNLOCK ITS POTENTIAL AS AN AI SUPERPOWER In recent years, investment in the UK’s networks has focused on the data centre market, driving more demand across the national network. With over £14 billion in data centre investment pledged by the likes of Vantage, Nscale, and Kyndryl as part of the UK government’s new Artificial Intelligence Opportunities Plan, this demand won’t be slowing down any time soon. In fact, since then Google has also announced a significant investment in Hertfordshire to support AI innovation and digital services. The multiple large-scale data centre builds that are currently in the pipeline will have a significant impact on the demand faced by the UK’s digital infrastructure. While much of the existing core network has sufficient capacity today, the dispersal of data centres outside traditional hubs like London and Manchester will put pressure on the UK’s networks to evolve. THE UK’S GROWING DATA CENTRE DEMANDS This year, analyst house Gartner predicts that generative AI alone will account for 10% of all data produced, and that’s without accounting for the
data produced by non-generative AI. Looking forward, the global AI market is expected to grow at a CAGR of 36% in the years leading up to 2030, and the UK government is keen to secure its share. In the latest iteration of AI strategy, the government set a number of AI growth goals, including its aim to increase high performance computing capacity 20-fold over the next five years and establish AI growth zones across the country. With initiatives like the AI growth zones set to be established beyond traditional data centre locations, we’ll see deep-learning capable data centres being developed that can support inference closer to the edge. Computational infrastructure must be built out across the edge and between our core data centres to ensure high- speed, high-bandwidth, and low-latency connectivity. Computing power is increasingly being established outside of major cities like London and Manchester, largely dictated by the availability of power. Data centres are immensely power- hungry, and the availability of stable, cost-effective, and increasingly green energy is one of the biggest determining factors for site selection. This will be the driving force behind AI-ready data centre locations, rather than where historical infrastructure hubs exist. The
UK’s constrained energy grid means that areas with surplus generation capacity— such as near renewable energy sources like offshore wind farms or nuclear plants—are increasingly attractive locations for new AI-driven facilities. HIGH-PERFORMANCE NETWORK CONNECTIVITY NOW NEEDED ACROSS THE UK All the new investment in a more dispersed data centre landscape ultimately means more strain on the network. While the UK’s core backhaul capabilities are strong, the network edge is at risk of becoming a bottleneck, limiting AI’s true potential. With only the first AI growth zones announced, the exact location of the inference needs and subsequent fibre demands will be hard to determine. What is clear is the need for this fibre to extend right to the edge, to break out from the core, beyond the established backbone routes that line the UK. Having fresh fibre built out as close to the edge as possible with as much longevity as is feasible will also be an attractive prospect for hyperscalers. As they continue to invest in significant data centre builds like Nscale’s £2 billion data centre commitment, they will likely prefer to utilise fibre connections with as much lifespan left as possible, to avoid expensive fibre re-builds in the
16
www.opticalconnectionsnews.com
| ISSUE 41 | Q2 2025
Made with FlippingBook interactive PDF creator