Data Centers Aren’t the Bubble With Eugene McGrane

Podcast Briefing: an Edited Q&A

The following Q&A has been adapted from the HC Commodities Podcast and edited for clarity and length.

Are data centres a bubble?

Paul Chapman: We’re seeing a lot of commentary comparing AI to the dot-com era. Are data centres at risk of becoming overbuilt assets, or are they something more structural?

Eugene McGrane: It is reminiscent of those cycles, but I think we have to separate what’s happening in public markets from what’s happening on the physical infrastructure side. Where I don’t think there’s a bubble is in the industrial backbone supporting this level of compute.

The numbers are meaningful. In 2025, there were roughly 12.5 gigawatts under construction globally, up significantly from just a few years earlier, and now we’re talking about more than 30 gigawatts. That kind of scale reflects real demand. Companies are building because they believe they need this infrastructure to remain competitive, even if returns aren’t fully visible today.

The economics of owning compute

Paul Chapman: Even if valuations feel stretched, it seems like owning data centre infrastructure today is highly attractive. Is that a fair way to think about it?

Eugene McGrane: I think that’s broadly right. If you own the infrastructure that provides compute today, you are in a strong position because it’s scarce and difficult to replicate. But I’d frame it slightly differently.

This is less like building a refinery and more like owning key infrastructure within a network, whether that’s ports or rail junctions. The value comes from the location, the connectivity, and especially access to power. The question of who ultimately uses that compute remains open, but the underlying need for it is not.

Power as the central constraint

Paul Chapman: You’ve mentioned power a few times. Is that really the defining bottleneck today?

Eugene McGrane: Power is the moat. It’s not just how much power is available, but how you secure it, how you price it, and how quickly you can access it.

What’s changed over the last few years is the scale. We’ve moved from looking for 50 megawatt sites to hundreds of megawatts or even gigawatt-level developments. That completely changes the equation. At that scale, you’re not just finding a site, you’re effectively building an energy solution around your data centre.

Community pushback and permitting challenges

Paul Chapman: There’s been a growing backlash in parts of the US around data centres. What’s driving that?

Eugene McGrane: A lot of it comes down to how quickly the scale has changed. Communities were caught off guard by the size and impact of some of these projects.

There are real concerns around power usage, water consumption and local impact. And in the US in particular, infrastructure development requires navigating multiple layers of approval. If you don’t have community buy-in, projects can stall. That’s led to a situation where more experienced developers are now taking the lead, because they understand how to manage those dynamics.

Engineering and technological evolution

Paul Chapman: Are those concerns being addressed through better design and engineering?

Eugene McGrane: Yes, and quite quickly. Cooling technologies, power management, and overall efficiency are improving. There’s also more emphasis on integrating with power systems in a way that reduces impact.

You’re starting to see more advanced solutions, including different power mixes and even early-stage developments in new generation technologies. There are examples of fusion projects being contracted, which would have been unthinkable a few years ago. It shows how much innovation is being pulled through by demand.

Who is building the infrastructure?

Paul Chapman: Are these assets being built speculatively, or is demand already locked in?

Eugene McGrane: The competition for capacity is intense, so most projects are not being built without a clear path to utilisation. Hyperscalers are building directly, but there’s also a broader ecosystem of developers and capital providers involved.

In some cases, developers will secure land and power access before bringing in partners, but given the scale of investment required, you typically see funding aligned early. The market is still moving quickly enough that there’s very little idle capacity being created.

The risk of overbuilding

Paul Chapman: One of the usual outcomes in capital cycles is overcapacity. Do you think that’s likely here?

Eugene McGrane: It’s possible in pockets, but the barriers to entry are high. The permitting challenges, power constraints and capital requirements all act as filters.

What we’re seeing is that a lot of the early speculative projects have already fallen away. The projects that remain tend to be larger, better funded and more likely to be delivered. Vacancy rates are still very low, and pre-leasing is strong, so it’s difficult to argue that we’re already overbuilding in a significant way.

Technology risk and obsolescence

Paul Chapman: One potential risk is technological change. If compute becomes more efficient, could today’s infrastructure become obsolete?

Eugene McGrane: There’s always that risk, but I think the broader trend is still an increase in data generation and usage. Even with efficiency gains, demand tends to expand to fill capacity.

We’re only just beginning to understand what’s possible with the amount of data we’re collecting and processing. There are both commercial and societal drivers pushing for more compute, not less. So while technology will evolve, the need for infrastructure is unlikely to disappear.

Capital intensity and the “treadmill” effect

Paul Chapman: Is there a risk that companies are simply spending to keep up, rather than to generate incremental returns?

Eugene McGrane: That’s a real consideration. There’s a historical parallel with industries that required constant reinvestment just to maintain competitiveness.

In that sense, some companies may find themselves on a capital investment treadmill, where returns don’t increase in proportion to spending. But they still have to invest, because not doing so would mean falling behind.

The long-term outlook

Paul Chapman: If we step back, how should we think about the longer-term trajectory for this sector?

Eugene McGrane: The infrastructure being built today is supported by long-term commitments. Power generation assets, for example, typically require 15-year agreements.

So the investments are being made with a long horizon in mind. There may be volatility along the way, including valuation corrections or pauses in development. But the underlying demand for compute and the infrastructure that supports it looks structural rather than cyclical.