IPI Partners: Size matters for hyperscale data centers

This article is sponsored by IPI Partners

Artificial intelligence is no longer a distant and vague concept, it is here and its consequences are tangible, believes Matt A’Hearn, partner at IPI Partners, a global investment platform focused exclusively on data centers and other technology and connectivity-related real assets.

Indeed, AI is now exploding. With an expected annual growth rate of 37.3 percent between 2023 and 2030, according to consulting firm Grand View Research, AI is expected to increase demand for data center capacity, led by hyperscale companies. This need will make partnerships with specialized developers of data centers stronger than ever, A’Hearn argues.

How will AI’s rise impact hyperscalers and their need for data centers?

AI is a technology that is interfacing with pretty much all products that large-scale technology companies – hyperscale companies – deal with. If you listen to their earnings releases, AI is often intertwined in every part of their business. So, we think it is going to have an impact on all applications, whether it is the cloud, search, or other products or services.

AI has really driven the need for hyperscalers to scale their data center capacity in a meaningful way, on a global basis. If we think about most of these companies, they have cloud businesses that continue to grow at a very rapid pace, so they need data center capacity to support that growth. AI is a step function on top of that; it is really increasing the timing sensitivity and the scaling sensitivity for hyperscale companies to support this growth.

Can the current stock support AI’s growth?

The short answer is ‘no’. In terms of the amount of available capacity today, we were already at a point where the existing stock cannot support the current expected growth in the market. Now that we layer on AI, it is even more of a challenge.

Data centers are continuing to grow at scale, but there are challenges that exist from a supply standpoint that are driven in many respects by power availability. While data center campuses have continued to get larger on the demand side, the supply side has been more challenged than ever to be able to find large-scale sites with access to power.

Where can these large data centers be built?

When we think about the use cases for AI, many of them still rely on proximity of population, meaning large data center markets will continue to be a major focus for the supply side. That means we will continue to see Northern Virginia, Northern California, or markets like Frankfurt and Tokyo as the starting point. But then we also often see a focus on edging out beyond just these central major markets.

It is going to be a bit of an expansion, we believe, of some of the existing major data center markets to provide supply to support AI.

Even in those situations, it is going to be challenging and newer markets are starting to be discussed and evaluated where there is large-scale power, which comes with its own complications from a network configuration standpoint.

But there is an interest in finding some of these other markets where there is capacity. Large-scale utility infrastructure, together with the renewable component, is a focus for the industry.

How will these constraints hyperscalers face affect future development?

The constraints will drive them to use other parties to help solve problems they may have. Particularly, the scarcity of the power that is needed will cause them to look to partner with groups who have the teams available to find those unique ways to solve problems. So, we are going to see more creativity around the market.

Is there greater pressure now to build green data centers?

Yes. Hyperscale companies are leading the world in putting out the most ambitious sustainability targets and seek experts to work with them to achieve these commitments.

It is important to have an active dialogue with them around how to find sites with power availability, which is ideally renewable power. But not just power, because there are also other sustainability metrics like water use.

There is a data center site in Oslo, for instance, that has a unique sustainability story that I think we will see more and more. The heat that is coming out of that data center site is being utilized by some 5,000 homes in a central heating district within the city.

I expect we will see more examples like that. The puzzle in this industry is how you utilize other aspects of the data center to give a sustainability benefit and I think there will be some very inventive solutions.

How have tenants’ appetites changed? What is the average length of a tenancy contract for data centers?

Hyperscale data center contracts tend to be 10-15 years, so they are long-term leases. From the hyperscalers’ perspective, it helps to provide security in terms of knowing that capacity will be available on a go-forward basis.

That is important given the magnitude of the investment they will put into the data center. They might build the shell, and the mechanical and electrical infrastructure, but they are investing multiples of that in their own servers and storage, network infrastructure, cabling and so on.

They are then starting to put their customers, ultimately, on those servers and storage. As they think about that business going forward, because it is so difficult to move, they want some level of security around what the cost of the capacity is. So, they are often looking for long-term leases with us, as well.

From a leasing standpoint, we see a lot of build-to-suit development projects and hyperscalers are often looking for a longer runway to be able to ramp up into campuses.

A decade ago, a hyperscale customer might take a data center hall in an existing multi-tenant building, which started to shift a few years ago to wanting their own building. Now, we are seeing that they want their own campuses where they know they have room to grow.

With multiple buildings on the entire campus, often it might be necessary to phase the construction: one building gets built, followed by a second, and additional buildings over time. So, having that path that allows expansion for future growth on a much larger scale has been a positive evolution over the last couple of years.

How can private equity companies partner with hyperscalers?

It starts, to some extent, with helping them solve problems and doing it on a scaled basis. That is very hard to do. You have to build trust, which means you really have to build those relationships over years of work – developing and delivering for customers, operating for them – because the infrastructure they put into a data center is mission critical for their businesses.

It is important to have the scale needed on a global basis to compete and ultimately have the required relationships. Just to give a sense, we talk in terms of megawatts when we talk about how large a site is – that is the amount of power that can serve a site.

If you are thinking about a 100MW campus, which is a big campus, that is over $1 billion of capital that is needed for that one campus.

These are capital intensive projects, so while there is a lot of interest associated with the industry, there are very few people who can play at the scale that is necessary.

Do hyperscalers tend to build their own data centers or are they increasingly leasing?

They have always built some of their own data centers. That happened historically, it happens today, and it will happen going forward. But we continue to see, because of the demand dynamics and the challenges with supply, that hyperscalers are increasingly turning to outside groups to help find solutions to those problems.

These hyperscale companies are more reliant than ever on trusted partners for scale, for timing sensitivity and for developing superior land banks, especially as we think about international markets where they don’t have as much of a team as they may have in the US. Enabling partners to achieve faster time to market than they could on their own is a meaningful value proposition in today’s environment where speed at scale is so important.

If you think about what is happening in a data center, the data being processed is extremely valuable – especially to these businesses. Data center operators need to get this right, so hyperscalers prefer to work with a smaller group of experienced investors of scale instead of having a very broad universe. It is critical to have partners that can scale with them on a global basis.