Golden hockey-stick opportunities do not appear often. When they do, the right response is not to casually watch from the sidelines. It is to focus, go back to school, and do the homework. The AI data-centre buildout looks like one of those rare moments, because demand is moving faster than the physical infrastructure needed to support it.
The market has already understood the AI software and chip story. What I think is still being repriced is the second-order infrastructure layer underneath it: electricity, grid connection, cooling, backup generation, switchgear, transformers, substations, gas turbines, diesel and gas engines, batteries, power-management software, high-density rack systems, modular construction, and AI memory.
Why this matters is scale
Next-generation AI racks can consume dramatically more power than traditional data-centre racks. Some designs move from the old 5 to 15 kilowatt range toward more than 100 kilowatts per rack. A traditional data centre may need tens of megawatts. An AI-focused campus can require hundreds of megawatts. That single change rewrites the economics of the entire supply chain.
That is the golden hockey stick. It is not just more data centres. It is more power per building, more cooling per rack, more backup generation, more grid equipment, more modular construction and more memory bandwidth. If a data-centre operator cannot secure power, it cannot deploy compute. If it cannot cool the racks, it cannot run the machines. If it cannot source transformers, switchgear, generators, grid connections, memory or construction capacity, the project is delayed.
Where the demand actually lands
The opportunity is not a single stock or a single sector. It is a chain, and a delay anywhere in the chain holds up everything downstream. These are the areas Bill Moss AO and the BGW desk are tracking most closely.
The homework
A theme running this hot will attract capital that has not done the work. The discipline, in Bill's words, is to track backlog, orders, lead times, margins, capex, power availability, approvals, grid bottlenecks and quarterly commentary. Those are the signals that tell you whether supply is still constrained, which is where the strongest phase of the cycle tends to sit.
The mistake a golden hockey stick invites is complacency: assuming the obvious story is already priced and watching from the sidelines. The whole point of the setup is that the demand is visible while the physical response lags. That gap is the opportunity, and it is also where the work has to be done.
How we think about it at BGW
For our clients, the question is not whether the AI infrastructure theme is real. It is how to take exposure with discipline: across listed and private positions, sized to a portfolio, and bought at a price that respects how much of the story is already understood. That is the kind of work the BGW Investment Committee does on a theme like this, and it is the kind of conversation we are happy to have one on one.