Q&A with Steve Kundich: Next Generation Digital Infrastructure to Support AI Workloads

Steve Kundich has been leading data center design and innovation across the industry for more than 20 years, creating and deploying cutting-edge solutions for some of the world's most transformative companies. In this Q&A, we will learn more about his current role at Ada Infrastructure and his insights on developing sustainable data centers that can adapt to meet today's emerging technologies.

What are your responsibilities at Ada Infrastructure?

My job is to build a design program that meets the business objectives of the company and our customers from both a process and product standpoint. This entails setting the foundational design direction, cultivating partner relationships, and working with the leadership team to plan out the roadmap.

What drew you to the company?

Ada Infrastructure currently has 1 GW of secured capacity across seven inflight campuses in Japan, Brazil, and the UK. This global portfolio is a blank canvas, at a time when the rapid adoption of AI is redefining the digital infrastructure needed to support it sustainably.

As an architect and data center designer, this is an intriguing opportunity: I get to work with a team of data center industry experts along with a global partner ecosystem aligned with our vision, and some of the world’s largest and most innovative companies to design and build the infrastructure that will power our AI future for decades. 

What are your priorities in developing the design strategy?

Future-readiness is built into our reference design. When working at the scale we are, that involves developing a design strategy that is repeatable, flexible, and scalable, while also taking into account factors such as local climate zones, regulations, the surrounding community, and customer-specific requirements.

Sustainability is part of this and core to everything we do. Since we’re not bogged down with legacy data centers, we have a real opportunity to work with hyperscalers and partners to make significant strides in sustainable design – from construction to operations. We also have a great partner in GLP, with strategic assets, established supply chains, and renewable energy expertise and resources.

Safety in design is the other priority – we have put processes in place to ensure our data centers are designed from the beginning for safe construction and operations. This is an area our president, Jennifer Weitzel, is deeply invested in and focused on.

How does AI play into your designs for future-readiness?

Future-readiness is about having flexible space, power, and cooling.

When I first got into the data center business in its early days, we used chilled water units – or CRAH units – to cool data halls before direct/indirect air cooling became the norm. With AI-specific chips driving up densities – along with increased power and cooling requirements – the industry is now coming full circle back to liquid.

Our system is set up to enable both air and liquid cooling: air for lower density workloads and liquid for higher density, mainly AI, workloads. We’re targeting air cooling for racks averaging 10 kW to 12 kW, and liquid for anything beyond that, but each customer will have their own requirements. The advantage we have is that our design is ready day one for both air and liquid cooling.

It’s almost impossible for companies to anticipate and plan the amount and types of capacity they’ll need and how quickly AI hardware will ramp up in the data hall. It’s conceivable we could reach densities of 40 kW to 70 kW a rack for AI and ML workloads.

The flexibility to scale up liquid cooling to AI racks will be critical, but you have to allow space for equipment, piping, and valving to support it. Having flexible space, power and liquid cooling built into our design gives us and our customers an advantage and the ability to adapt to AI workloads and chip advances as needed.

What are some of the roadblocks you see in meeting growing demand for capacity for AI workloads?

We luckily don’t have the same constraints other providers have. For example, we don’t have legacy data centers that will be difficult to retrofit to meet rapidly emerging and future needs. Other providers also don’t have ready access to strategically located real estate and other resources that we do through our relationship with GLP.

Another challenge we see, that is industrywide, is the global supply chain constraints. Our team at Ada Infrastructure leverages long standing relationships with major suppliers in conjunction with our design strategy (standard, repeatable, flexible) to meet the capacity needs of our customers.

Lean construction – eliminating as much design, construction, and infrastructure as possible – is a particular focus of mine. Off-site fabrication and near-site fabrication will play a big part in this – limiting the impact of supply chain issues while also feeding into our safety and sustainability commitments.  

Our team has deep experience deploying global data center strategies for companies like Meta, Microsoft, AWS, and others.

Learn more about our approach to health & safety here, and how our CISO fosters a security culture here.