Wall Street isn’t built for consortiums. Blood feuds flourish, and the pressure to outperform rivals leaves little reason to team up with them on investments. But the sheer size of the AI buildout has competitors playing nice.

“There’s enough to go around for all of us,” Brookfield Asset Management CEO Connor Teskey tells Semafor. Chalk it up to Canadians’ go-along-to-get-along instincts, but the race to shift the cost of the AI buildout — $7 trillion over the next decade, by Brookfield’s estimate — from the hyperscalers themselves to external money managers is making strange bedfellows.

Brookfield has a joint venture with Nvidia to buy up to $100 billion of AI infrastructure assets. It’s also in a consortium arranged by the chipmaker to corral another $500 billion alongside Apollo, Blackstone, and KKR — some of Brookfield’s fiercest rivals.

Data centers, as well as the hardware inside them and the power plants feeding them, are basically the same as bridges, ports, and other unsexy but economically fundamental stuff that firms like Brookfield exist to build, own, and operate — albeit “on a much larger scale and happening much faster,” he said. There will still be plenty of losers among data-center financiers, he said. For power projects, he’s wary of gas, which is slow to deploy, and keener on fuel cells, onshore renewables, batteries, and nuclear.

“If there’s a single narrative that is most starkly divorced from the fundamentals we’re seeing on the ground,” he said, “it’s the idea that the demand for AI infrastructure and the energy and the supply chain that supports it is in some way fragile.”Tim McDonnell and Liz Hoffman: Is power still the most important bottleneck for digital infrastructure?

Connor Teskey: By a very large margin. This imbalance and the capital requirements for energy are going to last through the end of this decade and beyond. But in an environment where all forms of generation are seeing increased demand, we see certain technologies growing the fastest — fuel cells because they’re the fastest to deploy; renewables because they’re the cheapest form of bulk electricity; batteries and energy storage because they provide stability to an increasingly volatile electric grid; and nuclear because it provides your large-scale baseload and energy security. Which offers the best returns?

Down-the-fairway, mature renewables — onshore wind and solar, battery storage — are absolutely seeing the most growth and have the most market participants. So maybe your return outcomes are more range-bound. Nuclear is a huge opportunity. There is incredible demand, huge capital requirements, and there are far fewer people who have the expertise in order to invest and help pull those projects off of the ground.

Therefore, the return requirements are going to be different. You’re part of the consortium that Nvidia put together. [Nvidia CEO] Jensen Huang has been propping up a huge chunk of the AI buildout. Are we now at a place where investors who do this for a living are stepping in and backfilling that support?

If I can play it back to you in a slightly different way, AI infrastructure is growing on a very traditional trajectory that we’ve seen [before]. Initially it’s done by independent capital providers who look to build and hold the assets. Once the asset class grows, you begin to see banks syndicate the financing. Other capital participants come in, whether it’s private credit or insurance. That is exactly what’s happened within AI infrastructure. The only thing that’s different about this one is it’s happening on a much larger scale and happening much faster.

I’ve never seen competitors at the same table this often. How do you think about places where you want a deal to yourself versus sharing it — and sharing the returns — with your competitors?

We’re seeing one of the biggest energy buildouts in history. We’re seeing global trade rewired with an increasing focus on resiliency and production of critical goods closer to home. These are mega-trends that require far more capital than what has traditionally been available from both governments and even the public markets. So why can you get five or six of the leading players at one table? Because there’s enough to go around for all of us. But if everyone is in this trade, what’s going to differentiate winners and losers?

While it’s easy to paint everything with a single wide brush, there is going to be a dispersion of who performs well and who has some disappointing results. To be absolutely clear, we’re not doing every opportunity we see. We are doing the tiniest fraction. And even with that very tight filter, we can deploy a tremendous amount of capital.

Returning to the energy supply question, where do you see gas fitting in here? You mentioned lower-carbon technologies, but all the tech companies are scrambling to get their hands on gas turbines.

Gas is going to be critically important. But try and order a new gas turbine today, it’s not coming in a decade that has a ‘two’ on the front of it. That’s why today we see the greatest amount of growth in things like solar and batteries.

When you look at some of the mounting public pushback on data centers, is that a risk to the pace of energy demand growth ahead? Is there a risk of overbuilding?

If there’s a single narrative that is most starkly divorced from the fundamentals we’re seeing on the ground, it’s the idea that the demand for AI infrastructure and the energy and the supply chain that supports it is in some way fragile. Or that there’s a small tweak in model efficiency that’s going to disrupt this supply-demand imbalance we have today. Even if demand , there would still be an imbalance that the current levels of supply are nowhere near servicing.

So will it be this way forever? No. But certainly today being a provider of incremental capacity into this supply chain is very attractive. It’s also important to recognize that what we’re building is under 20- or 25-year contracts with the greatest counterparties in the world. We’re going to get a return on our capital in that initial contract, regardless of what the demand environment looks like two or three decades from now.

Infrastructure projects historically have a public offtaker, like the Port Authority or municipal water departments. Private companies are on the other side of the AI projects. Is AI fundamentally infrastructure, or are we just calling it that?

In the past, economies ran on toll roads and railroads and pipelines that were physical. But as the world around us increasingly digitizes, data is going to need to be increasingly transported, processed, and stored just like every other commodity. So we view this as building the pipelines and the toll roads for a digital era.