Building data centers to suit AI workloads is very different than building to suit traditional cloud or other workloads. Just as the various requirements for AI model training and inference are driving significant shifts in the location of data centers, they’re also shifting the timing requirements. In our “Building to Suit in an AI World” series, we first explored modeling for training or inference. The unique requirements of AI workloads are driving significant developments in new markets, as well as advancements in edge computing. Learn more here: Building to Suit – Location.
Now, let’s explore the importance of having capacity ready when demand arises, particularly with the right partner with the right resources. AI development and adoption are in the very early days, making demand forecasting incredibly difficult. Consumers have turned to AI for a wide range of use cases. AI is driving value in finance, healthcare, robotics, and beyond. And there are so many more opportunities yet to be realized — opportunities that could wildly change how much data center capacity AI will demand. Likewise, innovations in AI development could dramatically change the data center requirements for AI model training and inference.
Even as future data center capacity needs for AI are unclear, speed to market is essential. AI companies need capacity ready when the demand arises or risk losing market share. As our Chief Marketing and Product Officer Phillip Marangella explained in a Data Center Frontier article about adapting data center infrastructure for the age of AI, “The scale for both hyperscale and the edge has increased by at least 3-5x, and the speed of delivery has been reduced in some cases to half the traditional timeframes.”
Regardless of the capacity needs, it’s unlikely that they can be met with existing capacity, as vacancy rates have fallen to record lows in all of the primary markets globally. Without off-the-shelf capacity, development needs to be fast. However, this is increasingly difficult as years of high demand have led to land and power constraints in many primary markets. Securing entitlements and power interconnection is now often the longest part of the development process.
Building to Suit Timing Requirements
Developing quickly, even in constrained markets, requires land and power banking. In markets where land and power constraints have added years onto development timelines, having banked land and power enables us to develop new capacity relatively quickly. In Japan, we are establishing a 200MW data center platform, with more growth ahead. In Malaysia, we have 130MW across key areas, including Kuala Lumpur, Cyberjaya, Bukit Jalil and Johor. And those are just a few examples of markets with opportunities for expansion that are ready to develop.
Ensuring capacity is ready when it’s needed requires significant capital and commitment to invest it. Meeting AI demand alone will require a $5.2 trillion investment in data center capacity by 2030. Much of that investment will need to be made in advance—for things like prefabrication, global purchase agreements for long-lead equipment, power reservations with utilities, and the acquisition of larger land banks to cut time to market. Financially backed by EQT and Sixth Street, we have access to capital for projects worldwide and can fund scalable capacity to meet customers’ future requirements as needed.
A partner with creative approaches can dramatically accelerate speed to market. We leverage joint ventures, partnerships, and strategic acquisitions to augment organic capacity capabilities. For example, our strategic investment in Chayora, a leading data center provider in China, enabled our customers to access data center capacity in Beijing and Shanghai. Our joint venture with Adani Enterprises, AdaniConneX, has fast-tracked the development of a network of hyperscale data centers in India, including one in Chennai.
Bottom Line
Future data center capacity demand is wildly unpredictable. Yet, having capacity ready when the demand arises is essential. However, speed to market is challenging given that existing capacity is fully utilized in many primary markets. Success requires a developer with banked land and power, access to significant capital, and a creative approach to new markets. A developer like EdgeConneX.
Previously in the Building to Suit in an AI World Series: Build to Suit – Location.
Learn more about EdgeConneX next-gen, high density solutions for AI and HPC HERE.
Up Next: Build to Suit – Density. Stay tuned!