DataBank is a data center infrastructure provider that develops and operates colocation, cloud, connectivity, and AI-ready data center facilities across multiple markets, supporting enterprises, hyperscalers, and AI workloads.
Raul, you have led DataBank through a period of significant growth. Can you tell us about that journey and what it has taught you about the infrastructure underpinning the AI boom?
I've been at DataBank for about 10 years. When we started in 2016, we were a small business with six data centers in three markets, serving primarily enterprise customers. Today, we operate 65 data centers across 25 markets, have 1.3 gigawatts of capacity installed or under construction, and serve more than 2,200 customers.
The biggest observation is that data centers are the foundation of technology adoption. Over the last decade, the world has continued to digitize, creating more demand for digital infrastructure. AI is simply the latest chapter in that trend.
Before ChatGPT, this sector was known as high-performance computing. We saw an early glimpse of its infrastructure requirements in 2018 through a project with Georgia Tech. What makes AI different is the sheer amount of compute, storage, networking, and power it requires. That demand is driving the current expansion of data center infrastructure, and I believe AI could ultimately be as transformative as the internet itself.
You have suggested that enterprise AI adoption may be slower and more gradual than many expect. Why do you believe that?
When the internet emerged, businesses struggled to understand how to restructure themselves around a completely new technology. We're seeing something similar today. Among our enterprise customers across industries such as healthcare, financial services, retail, and manufacturing, there is enormous interest in AI, but many are still figuring out how to integrate it into their existing operations and technology stacks.
The value of AI for software development is already clear, and productivity gains have been remarkable. Beyond that, we're beginning to see compelling use cases in areas like healthcare and scientific discovery. However, businesses are still operating complex organizations with existing responsibilities and systems. Integrating AI into those environments will take time. The technology's long-term potential is undeniable, but adoption across the broader economy will likely be more gradual than some forecasts suggest.
You have also spoken about tighter financing conditions and growing caution around AI investment. What concerns you?
There is an extraordinary amount of investment flowing into AI right now. We are seeing massive valuations, huge capital raises, and significant enthusiasm across the market. Having worked through previous cycles, including the dot-com era, the global financial crisis, and the software correction in 2022, I sometimes see valuations that make me pause and ask questions.
That said, I remain very optimistic about the long-term future of the technology. We are in an incredibly active phase of development, and investors are eager to support innovation. My point is simply that markets move in cycles, and history suggests that periods of intense enthusiasm are often followed by periods of reassessment.
Do you think some of the data centers being built today could eventually prove unnecessary?
It's an important question. Before ChatGPT, most data center development was concentrated in a relatively small number of major markets where hyperscalers had established their infrastructure. Since then, demand has increased dramatically. In the U.S., annual data center absorption has grown from roughly one gigawatt in 2022 to many times that level today, and total installed capacity has expanded significantly.
Much of that growth has occurred in regions that historically were not major data center markets. Developers have gone where power is available, including places like West Texas, Wyoming, Wisconsin, Kentucky, Louisiana, and North Dakota. The challenge is that nobody knows exactly how much infrastructure will ultimately be required. AI today relies heavily on power-hungry GPU architectures, but future semiconductor innovations could dramatically improve efficiency. Those developments will influence how much capacity is needed in the years ahead.
Can you tell us about DataBank's urban AI inference centers and the thinking behind them?
We are currently building around 700 megawatts across several major markets, including Dallas, Northern Virginia, Atlanta, San Diego, Kansas City, and New York. Our strategy is based on proximity to users. Latency matters, and consumers live in major metropolitan areas, so we believe infrastructure should be located close to where demand exists.
Inference, which is the actual use of AI models, benefits from being near end users. At the same time, traditional infrastructure demand remains strong. Our view is that over time the distinction between AI infrastructure and conventional IT infrastructure will fade. Future data halls will integrate GPUs, CPUs, storage, and networking into a single environment because AI capabilities will increasingly be embedded directly into the applications people already use every day.
What is the biggest bottleneck facing the data center industry today?
The biggest bottlenecks are power and supply chain constraints. Building new capacity is no longer something that can be done quickly. Developing a major data center project has become a five-year undertaking, and companies need the resources and patience to plan on that timescale.
The challenges range from securing equipment, materials, and skilled labor to obtaining power and working with utilities to deliver it. Those processes also take years. That's why we spend so much time searching for opportunities that can accelerate deployment. For example, our recent joint venture with Goodman Group enabled us to unlock 32 megawatts of power in Los Angeles with a service date by the end of 2027. We often describe that process as being a "truffle pig"—finding hidden opportunities that allow us to bring capacity online faster.
You are preparing to hand over the CEO role. What do you expect from your successor, and what will your own role be going forward?
I've worked closely with Kevin Ooley for the past 10 years, and I'm fully confident in both him and the broader leadership team. DataBank has achieved 46 consecutive quarters of positive net bookings, meaning we have consistently grown the business quarter after quarter for more than a decade. It has been a very disciplined and successful organization.
I expect that trajectory to continue. As for me, I look forward to remaining involved at the board level rather than in day-to-day operations. As the company has grown, long-term strategy and access to capital have become increasingly important. I am excited to work with Kevin, the management team, and the board to help shape the company's next decade of growth.
