NORBERT BIEDRZYCKI

The AI bubble is the wrong question

Microsoft has just pulled out of a deal to build a data centre in Loughton, near London. Not for lack of demand, but for lack of power: the grid connection there won’t be ready until the 2030s.

And it’s not the only case. Morgan Stanley estimates that by 2029 data centres in the US alone will need about 160 GW of power. Even counting what is already under construction and the power generated on site, there is a shortfall of about 76.5 GW. That’s roughly the installed capacity of every power plant in Poland combined.

For weeks the market has been asking whether the AI bubble is about to burst. I think that’s the wrong question.

The capital is there. So is the demand: data centre vacancy in North America is 1.4%, even though supply grew by more than a third in a year. What’s missing is far more mundane. Switchgear, which has gone up 60% in price since late 2021. Transformers. Megawatts.

This isn’t a capital problem. It’s a physics problem.

For boards, this has a very practical consequence. Most AI strategies I see assume compute will be available on demand and keep getting cheaper. That assumption rests on infrastructure nobody has built yet.

When did your board last ask where, physically, the compute in your AI strategy will come from?

Read next

  1. AI won’t run out of customers. It will run out of power and time to repay its debt

    Many people read the sell-off in data center stocks as the first crack in the AI bubble. I look at the same numbers and see something else. The demand is there, and it is substantial. The trouble lies elsewhere: in the construction itself, in access

  2. Data centers in orbit. The physics is starting to work. The spreadsheet isn’t, yet

    Space data centers are no longer a conference slide. We already have the first satellites, the first billion dollar valuations and the first delays. The delays tell us the most about this market. For the last two hundred years, industry has gone where the energy

  3. We may have fundamentally misunderstood the economics of AI, because it won't be freeLinkedIn

    For the past two years, we've heard almost exclusively one narrative: AI will make knowledge work cheaper, faster, and practically unlimited in cost. But early signals are starting to emerge that reality might look very different.