Powering Progress Episode 10 – Building Resilience into AI Infrastructure

Aubrey Bishai [00:00:09]

Welcome back to Powering Progress, a podcast series exploring the infrastructure, energy, and market dynamics shaping the next generation of data centers and AI development. I’m your host, Aubrey Bishai, Chief Innovation Officer at Vinson & Elkins.

Much of the conversation around AI infrastructure has focused on growth, new data centers, unprecedented capital investment, rising power demand, and the race to build capacity fast enough to keep up. But every major infrastructure boom eventually enters a different phase.

Projects become operational, markets evolve, technologies advance, regulations change, and the contracts that looked straightforward during periods of rapid expansion are put to the test. So what happens when today’s assumptions no longer hold?

Joining me today is James Barratt, a partner in our international disputes and arbitration practice. James advises clients on complex cross-border disputes involving major infrastructure projects and long-term commercial arrangements. He has been at the forefront in identifying where disputes are likely to arise in the operation phase of the AI buildout.

Today, we’re discussing the next phase of the AI infrastructure story, where disputes may emerge, the lessons other industries can offer, and what companies should be doing now to build more resilient agreements for the future. James, welcome to the podcast.

James Barratt [00:02:00]

Thank you, Aubrey. Great to be here.

Aubrey Bishai [00:02:02]

History shows that large infrastructure sectors often go through a honeymoon period before disputes begin to surface. What tends to trigger that transition?

James Barratt [00:02:15]

Aubrey, I think what we’re really asking here is, what ends the honeymoon phase? Disputes typically surface not at the time when the deals are done. They surface instead in the operational phase, and typically only after the first real market shock. We’re in a boom market at the moment. Problems get absorbed, they get renegotiated. It’s in no one’s interest to litigate disputes at the moment.

However, there’s typically a trigger — be that a dip in demand, a jump in the cost of power, there’s an issue with a counterparty not performing, or the technology level moves on. That’s when people start to really focus on the terms of the contract and see where the leverage is. We’re now looking at deals that typically have a lifespan of 10, 15 years, signed in an enormous, exponential growth market.

These deals get tested in a downturn. Now look, I’m not here to predict doom and gloom by any stretch of the imagination. I think this is a fascinating build-out to watch, and it’s great that all of us have been a part of it. But it’s today’s contracts which will be the ones that we’ll be arguing over in the next downturn.

Aubrey Bishai [00:03:24]

James, you’ve authored this article, “The Coming Wave of Disputes in Data Center and AI Infrastructure,” where you draw parallels between the AI infrastructure market and other project-financed industries. Which sectors offer the most useful lessons, and why?

James Barratt [00:03:46]

Well, Aubrey, I think that the LNG sector, liquefied natural gas, is the closest thing we’ve got to a playbook in terms of looking at the disputes. Many have noticed that the deal structures, while typically based on co-location agreements, do draw a lot from the project finance world. We see that with take-or-pay provisions, offtake provisions, so on and so forth.

My point is that the disputes, when they arise, the participants to them are likely to draw on the LNG disputes playbook. And you can see that with the arc of what’s happened in the last 15 or 20 years with LNG disputes — waves of price review arbitrations when there’s a market dislocation, and when the economics are very different from where they were when the deal was signed.

My view is that we’re likely to see the same with force majeure, where one party attempts to seek relief, be it from an outage or some form of disruption. And often those claims may be spurious. Sometimes they may be well grounded. But what’s for sure is there’s going to be a fight on every level in terms of that, when the economics are no longer as closely aligned as they are now — for example, because of the change in the economics in terms of pricing of the technology or the compute power, or the ability to ensure that the grid is able to offer the energy that’s required.

The cautionary tale really, I think, from the LNG world is what we’ve seen in the Venture Global sagas, where we’ve had conflicting arbitration awards. Shell lost, BP won, but they’re both essentially on the same facts. And that’s, of course, dealing with the jurisprudence we have in international arbitration for LNG disputes.

There’s no jurisprudence we can sensibly rely on in terms of arbitral practice and precedent from the point of view of data center disputes. So therefore, I think what participants and what their lawyers are likely to do is they very much look to the types of arguments and the approach that’s been used by buyers and sellers in the LNG world.

So to my mind, that seems to me to be the closest thing we’ve got as a map to where this goes.

Aubrey Bishai [00:05:52]

We’ve talked a lot on this podcast about how power has become the defining challenge in this sector — both securing the supply itself and the long-term commitments like take-or-pay deals written on top of it. What types of disputes do you expect around power purchase and energy supply arrangements? And what happens to those take-or-pay structures if market conditions shift dramatically?

James Barratt [00:06:20]

Power is very much where the first issues are likely to land — first over getting and keeping supply, and then over the commitments which the parties have drafted on top. So if we look first at supply, three quick points on that. Where there’s a mismatch between when renewables can generate — because of course with solar, wind, hydro, there’s obviously environmental and geographical limits as to how they generate — when renewables generate, and when a data center actually needs power. So there’s a mismatch issue. Second, there’s curtailment.

Who pays for replacement power when the grid is down? And then third, unproven power tech — nuclear restarts, behind-the-meter gas. All of these as yet are fledgling technologies. What happens when something goes down and there’s a cut in the supply?

Similarly, if we look at take-or-pay, these provisions exist to make the projects bankable, financeable — lender certainty, secured capacity. However, if the price of compute falls, yet the contract price is locked in above the now-market price, one party may seek to make a reach for a force majeure argument if there are any grounds, or attempt to renegotiate or otherwise look at the contractual provisions and try and get out of the deal. By contrast, unlike LNG, there’s no set price review mechanism.

So as a result, to a large extent, a lot of potential for disputes has already been baked in to the financing and how these deals are structured. And one dispute over supply, or curtailment, or take-or-pay, may trigger breaches of financing agreements as well.

Aubrey Bishai [00:07:53]

So James, if you were advising developers, investors, operators, or customers entering into these agreements today, what practical steps would you recommend to reduce the likelihood of becoming tomorrow’s headline dispute?

James Barratt [00:08:10]

I think the key thing is to draft for the world that’s coming, not the one we’re in today. What do I mean by that? I mean, build in adjustment. Be precise, not vague, in terms of how we define things. Think about the big risks. We’ve talked about power, we’ve talked about curtailment. So draft carefully around what actually are the obligations there.

Similarly, think carefully about what are often standard pro forma clauses, but which actually need to be given real thought. Change-in-law clauses. What happens with tech obsolescence? Define what force majeure does or does not excuse. And most importantly, think long and hard about the right forum for dispute resolution. If it’s international arbitration, pick a well-regarded international seat, whether that’s New York, London, Singapore, elsewhere.

Focus, when there are disputes, on tech-literate arbitrators — those who actually understand the industry, actually understand that intersection that we’ve all been talking about between AI and power. And think also about early resolution. What are the options for the off-ramp to avoid becoming the next day’s headline news? So essentially, I’d say bring the dispute lawyers into the drafting room. We’re the ones who’ve seen what happens when these contracts break.

Let’s find a way to put the mechanics in place to ensure confidentiality, to ensure that these things don’t end up in newspapers, and also to ensure that there’s flexibility there to try to resolve the disputes amicably.

Aubrey Bishai [00:09:42]

Thank you, James, and thank you everyone for listening to Powering Progress. If you’d like to explore more insights on AI infrastructure and the legal frameworks behind it, visit velaw.com.