Every project has a real constraint and a stated one, and they are often not the same thing. The stated constraint is what the plan is built around. The real constraint is what actually decides whether you finish. Find the gap between them and you’ve found where most projects quietly fail.
Watch what’s happening in artificial intelligence right now and you see this in giant, glowing letters. The whole industry talks about models, talent, data, algorithms. But through 2026 the conversation has shifted to something far less glamorous: power. Electricity. The data centres that run these models consume staggering amounts of it, and the binding constraint on AI’s growth is increasingly not cleverness but kilowatts. So you get the strange spectacle of technology giants – Microsoft, Google, Amazon – striking deals with nuclear and energy providers, and a wave of mergers where the prize isn’t an AI company at all, but the power to keep one running. One reported tie-up paired a media-and-tech group with a fusion company. The hottest commodity in the smartest industry on earth turns out to be the most basic thing imaginable: a reliable supply of electrons.
Here’s why that should interest anyone who runs projects. The AI industry spent years optimising the visible constraint – better models, more data, faster chips – and is now being governed by an invisible one it under-planned for. The thing that decides whether the project can scale was never the thing the project was mostly about. It was sitting underneath, in the infrastructure, in the boring layer nobody puts on the keynote slide.
Your projects have this too. You build the plan around the obvious constraint – usually time, or budget, or scope – and you manage that one diligently. Meanwhile the real binding constraint sits somewhere quieter and unexamined. It might be a single overloaded specialist everyone depends on. A piece of ageing infrastructure the whole programme rests on. A regulatory approval. A supplier with no backup. The energy, the bandwidth, the one scarce resource that everything else silently assumes will be there. You optimise the constraint you can see, and you get blindsided by the one you didn’t.
The discipline is to go looking for the real constraint before it finds you. Ask the uncomfortable, unglamorous question early: when this project scales up, what actually runs out first? Not what we’re focused on – what we’re not focused on that everything depends on. The answer is rarely the thing in the title of the project. It’s usually somewhere in the infrastructure, the plumbing, the assumed-reliable layer that nobody owns because everyone takes it for granted.
And once you’ve found it, plan around it as the priority it actually is, not the afterthought it looks like. The AI giants are now spending billions to secure power because they belatedly understood it was the real limit. The cheaper move – the one available to the rest of us – is to identify our equivalent of “power” at the start, while there’s still time to secure it calmly, rather than discovering mid-flight that the whole project is throttled by something we never put in the plan.
So look hard at your next project and ask what it really runs on. Not the headline. The hidden dependency underneath. Find your version of electricity before you’re the one bidding desperately for it, and you’ll have solved the problem the smartest industry in the world is only now waking up to.
Leave a Reply