Tony Jefferson is Director of New Markets and Enterprise Development at Enspec Power. With more than 30 years of experience across the global energy sector, he works closely with developers, network operators and major infrastructure projects to address the challenges shaping the future of power systems. Here he shares his reflections from Data Centre Live London.
What happens when compute capability evolves faster than infrastructure can be delivered?
Recent advances in compute capability, the processing power delivered by GPUs, AI accelerators and the infrastructure that supports them, have been extraordinary. There is every indication that this pace of development will continue.
Yet the infrastructure required to support that growth operates on very different timescales.
Grid connections, substations, transmission reinforcement, major electrical equipment and planning processes are often measured in years rather than months.
That disconnect feels increasingly important.
Throughout the event, discussions repeatedly returned to power availability and what many are now referring to as “time to power”. Historically, site selection has often been driven by connectivity, access to customers and established technology hubs.
Today, power is becoming a much bigger part of the conversation.
Another theme I found particularly interesting was the growing relationship between energy and compute. For decades, infrastructure planning has largely focused on moving power to demand. Increasingly, there is discussion around whether some forms of energy intensive infrastructure may begin to locate closer to available generation capacity.
If access to power becomes one of the primary constraints on future growth, that shift could have significant implications for both the energy sector and digital infrastructure.
My biggest takeaway from the event was simple.
The AI revolution is not just a technology story.
It is also an infrastructure story.
And perhaps most importantly, it is becoming a timing story.
The organisations that successfully deliver the next generation of AI infrastructure are likely to be those that can align technological ambition with the realities of infrastructure delivery, power availability and long term system planning.
