With investment pouring into AI infrastructure, Damir Špoljarič, Managing Partner at Gi21 Capital, explains why capital is only part of the equation when it comes to building data centres that can deliver for decades to come.

The global AI explosion has transformed data centres into one of the world’s most sought-after infrastructure assets. The sector is flush with cash and optimism. Ambitious announcements arrive almost daily and investor presentations are chock-full of promises of unprecedented capacity.

If cooler heads do prevail, they’re likely to be those attached to the necks of experienced operators. These people know that a data centre doesn’t automatically succeed once a certain funding amount has been secured. A host of pieces must fall into place: location, power, regulation, a few thousand operational decisions, and future-proofing. They understand what it takes to get from concept to operation.

Money isn’t everything

Financing represents the greatest obstacle to market entry. Except for when it doesn’t.

Experienced operators know that money is, at most, half the battle. Power availability depends on reliable grid access, and a site without said access has limited value, no matter how much money you throw at it. An adequately powered and permitted site can be worth significantly more than undeveloped land because years of regulatory and infrastructure work have already been completed.

If there were as many data centres as there are investment proposals, every square inch of the planet would be covered. Instead, most end up in the digital dustbin, never making it out of the presentation slides. Construction risk, operational complexity, permitting timelines, and changing market conditions are lethal to most proposals.

The AI boom hasn’t been going on long enough to produce a surplus of experts. The pool of professionals with decades of experience delivering and operating large-scale facilities remains very small, and it gets even smaller when you consider that not every professional made the leap to AI-specific data centres.

The perfect location doesn’t exist

Location is everything, as they say. The right location is an amalgamation of priorities. Electricity prices have to be reasonable, ideally with some room to negotiate. Among the other major considerations are long-term grid capacity, political stability, environmental risks, telecommunications connectivity, latency requirements, and local planning regulations. Cooling strategies such as free cooling can significantly improve operating efficiency, although they depend entirely on local climate conditions. You’ll never get free cooling in Nigeria the way you can in Norway.

These variables differ from one country to another. Even comparing electricity costs across European markets requires detailed local expertise because pricing structures, regulatory frameworks, and subsidy mechanisms vary considerably. Make no broad assumptions. Every jurisdiction requires its own analysis.

Traditional cloud infrastructure had different requirements and concentrated around major connectivity hubs such as Frankfurt, London, Amsterdam, Paris, and Dublin. AI workloads are expanding into secondary markets where power is more readily available and latency requirements are more flexible.

The trouble is the lack of crystal balls. A facility designed today has to support hardware that has yet to be developed. Racks that once required around 10 kilowatts are now approaching 150 kilowatts within the same footprint. If you design only for current demand, you risk creating a facility that’s DOA – essentially obsolete before construction is even finished. It’s not difficult to understand why a facility built with extended operation in mind inevitably produces stronger outcomes. Plan for maintenance requirements, resilience, serviceability, and future upgrades from the get-go.

Perfect uptime and peak performance

Investors want confidence that facilities will perform as expected, particularly where new infrastructure has yet to demonstrate reliability under real workloads. A track record of delivering and operating data centres can therefore be an important consideration.

Maximum availability is often the goal, and organisations understandably want to demonstrate it. Much to everyone’s chagrin, truly uninterrupted service remains stubbornly difficult to achieve. Having built and operated a data centre that maintained 100% uptime for more than a decade, I know how much discipline it takes to deliver that level of reliability.

Industry standards around levels of redundancy differ, but meeting a certification alone does not eliminate operational risk. Every additional layer of resilience requires deliberate engineering choices and continuous operational oversight. The highest-performing operators often exceed standard redundancy requirements across cooling, power distribution, and security systems. Those decisions can be unpopular during the design phase because they increase complexity. Once built, however, they can dramatically reduce the likelihood of customer-affecting outages over the lifetime of the facility.

Beyond engineering, human error is still a major operational risk. Malfunctions happen at 3am, and they turn into disasters when a sleep-deprived engineer heads to the switching room and makes a mistake. Successful operators therefore place equal emphasis on process discipline, automation, and detailed operational procedures. Aviation is a helpful model here, thanks to its standardised protocol checklists and carefully designed processes that reduce the impact of fatigue and high-pressure decision-making. Resilience and uptime are ultimately the result of thousands of decisions made during design, construction, and daily operations.

Lessons for new builders

Contemporary demand for data centres is nothing short of extraordinary and is largely driven by AI and digital infrastructure investment. A boomtown always attracts newcomers. This is no exception.

It’s also no small undertaking. Success depends on a clear-eyed understanding of the full complexity of the business. Power strategy must be considered well before anyone breaks ground. The exact plot of ground that gets broken – the location – requires its own serious examination of energy markets, regulation, telecommunications, and environmental planning. Facilities must anticipate future technology rather than current hardware alone, and operational discipline must shape development, looking far into the future at decades of service.

The industry’s biggest competitive advantage is no longer simply access to capital. The money is here, and there’s plenty of it. Experience is what helps bring these variables together and turn investment into infrastructure that performs reliably long after the ribbon cutting.