Where do you see the greatest opportunities emerging across the data centre value chain today?
Our view is that value creation has moved upstream in the data centre value chain. While operations remain important, the highest-value component today is the ability to energise land. We still see operations as an important and valuable component in the chain, but the provision of these services is maturing, and what is most valuable today is powered land.
Capturing that value requires a broader capability set than many investors have traditionally needed. It starts with sourcing and structuring land transactions, navigating planning processes and understanding the design requirements of different end users. It also requires the ability to identify and secure power solutions.
Often, securing power involves much more than applying and waiting for a response. Utilities are becoming more cautious about allocating capacity to speculative data centre projects without clear visibility on end-user demand.
As a result, even for a strategy focused just on powering land, the ability to engage with hyperscalers and colocation operators is becoming an important factor in obtaining power access.
Once land has been powered, there are multiple paths available, from selling the site through to developing a powered shell, partnering with an operator or helping seed a larger operating platform. The key differentiator is the ability to secure and energise the land in the first place.
How concerned should investors be about access to power as demand continues to accelerate?
In most markets, the issue is not that grids lack power. The challenge is how long it takes to connect to that power. Many existing grid systems were not designed to accommodate modern data centre requirements. Moving from a distribution-level connection to transmission-level access often requires significant time and investment. As a result, the primary risk is less about whether power can eventually be secured and more about how long it takes to obtain it.
Speed matters because delays can affect development timelines, increase holding costs and make it harder to satisfy current demand. To improve outcomes, investors need the capability to access power through multiple pathways.
These can include gas generation, renewable energy solutions and front-of-the-meter battery energy storage systems.
There are also opportunities to work more creatively with grid operators through optimisation strategies and capacity-sharing arrangements supported by behind-the-meter storage.
The challenge for market participants is that every site may require a different solution. There is no single answer to securing power, which is why expertise across energy infrastructure, storage and grid optimisation is becoming more valuable.
How is AI changing data centre development requirements, and where are the investment implications most significant?
AI users generally have two distinct requirements: inference and training.
Inference workloads have their own requirements but are more likely to compete with hyperscalers and colocation operators for traditional locations. Training workloads, however, are different. They prioritise locations with access to very large amounts of power at the lowest possible cost and can often tolerate greater power latency.
That dynamic opens opportunities in more regional locations outside traditional availability zones. It also supports development models that emphasise speed, efficiency and lower construction costs, including single-storey modular facilities.
We are beginning to see AI training demand emerge across APAC. These companies are typically focused on securing real estate and infrastructure solutions while retaining responsibility for the chips themselves.
Their priorities are speed, access to power and the ability to develop large-scale campuses that can support requirements of 1GW or more. There is also a strong focus on reducing the cost of power wherever possible, including through battery storage and other efficiency measures.
From an investment perspective, AI training facilities introduce several considerations. Many projects have limited alternative-use potential, creating greater dependence on the underlying customer. In some cases, investors may face single-tenant exposure where the customer’s credit profile does not resemble that of a traditional institutional-grade occupier.
The capital requirements are also substantial. Large amounts of preferred equity are appearing in capital structures, while transaction arrangements are evolving to include parent-company guarantees and residual-value protections designed to attract the scale of capital required.
As AI adoption continues to accelerate, these projects are likely to become a more prominent part of the data centre landscape.
For investors, understanding the differences between traditional hyperscale demand and emerging AI requirements will be critical to identifying where value can be created in the next stage of the sector’s evolution.
