When we talk about the data centre boom, an important piece of nuance is lost. There is simply no historical precedent for the kind of growth the industry is seeing. Goldman Sachs projects a 165% rise in global data centre power consumption by 2030, driven almost entirely by AI workloads.   

Consequently, the pressure to build fast, build big, and build efficiently all at the same time has never been more intense, nor has the margin for error ever been smaller. Having all three seems nearly impossible.  

However, it’s not only possible to build data centres with significantly more efficient operations and sufficient compute power for AI workloads. Schneider Electric is seeing firsthand how more data centre builders are realizing that the architectural and electrical decisions they make at the very beginning of the design stage are the ones that determine whether or not their data centre can meet a hyperscaler’s technical and sustainability standards—and therefore, determine whether or not their data centre will be successful.  

Measuring efficiency: Understanding PUE  

A key metric for measuring how efficiently a data centre uses energy is PUE (Power Usage Effectiveness), calculated by dividing total facility power consumption by the power delivered to IT equipment. A perfect score is 1.0. Higher scores represent more energy being consumed by cooling, power conversion, lighting, and other uses that don’t produce more compute.  

In Europe, data indicates the average PUE is around 1.6, although northern and central European data centres typically have 10-15% lower PUEs. Google lists the industry standard as 1.56 and reports an average PUE of 1.09 in its fleet. 

At Start Campus’ SINES DC in Portugal, the target PUE is 1.1; ECO DC, in Sweden, has a certified PUE of 1.22. Schneider Electric is a long-term partner for both projects. These two companies don’t have the same levels of global resources—how are these world-class results possible without equally world-class resources? 

The answer: five key decisions, made early on, with the express goal of lowering PUE.  

5 design decisions that set the foundation for hyperscalers 

1. Locking in a renewable energy strategy 

For hyperscalers, 100% renewable energy is no longer a differentiator when choosing a facility partner. Instead, it’s become a prerequisite.  

That means negotiating the Power Purchase Agreement (PPA) in the design phase is now even more critical. The details of a data centre’s PPA shape the facility’s entire carbon profile, its relationship to the grid, its long-term energy costs, and the confidence to credibly report 100% renewable energy consumption to customers trying to meet their own climate commitments.  

In practice, securing a PPA before construction begins requires a builder to know how much energy the facility intends to use for each phase as well as its designed full capacity. SINES DC, for example, aims to have 1.2GW of capacity. With this knowledge, their PPA with Siemens Energy prioritized long-term energy affordability. By thinking so far ahead, Start Campus presented an attractive investment for companies like Microsoft, which is investing $10B USD into SINES DC.  

2. Going SF6-free 

Any data centre is reliant on its switchgear, especially medium-voltage (MV) switchgear, for the continuous operation demanded by customers. However, traditional switchgear sets use sulphur hexafluoride (SF6) as an insulating gas.  

SF6 is also the world’s most potent greenhouse gas, with roughly 23,500 times the impact of CO2. Consequently, removing SF6 from switchgear is becoming a common tenet in supply chain ESG requirements.  

Unsure of SF6-free equipment availability or performance, however, some data centres have made a costly and disruptive switch during operations.  

In reality, SF6-free MV switchgear performs to equivalent technical standards and is available from the same suppliers data centres already patronize. Using it also generates specific sustainability evidence to present to potential customers. Before SIN01 ever came online, SINES DC was built with all-SF6-free equipment, including 2 MV metal-clad switchgear sets and 13 PremSet installations.  

3. Designing the right kind of UPS architecture  

Alongside switchgear sets, Uninterruptible Power Supply (UPS) architecture is critical for optimizing a data centre’s operations. Perhaps unsurprisingly, UPS systems are one of the single largest factors when calculating PUE.  

Many data centres have been cautious when selecting and designing UPS architecture, often over-provisioning for extra energy capacity or over-focusing on 2N redundancy to meet uptime requirements. These redundancies come at the expense of less efficiency and more energy consumption, which inflates PUE scores in turn.  

As rack density requirements increase for AI workloads, UPS systems must now prioritize accommodating what’s next. Two major developments have simplified this challenge. First, UPS systems’ eco modes have come a long way from when we wrote about them in 2011. They can now reduce heat loss more efficiently, directly improving PUE. At Eco DC in Sweden, using Galaxy VX UPS systems in ECOnversion Mode has led to 99% energy efficiency and a 1.15 PUE at full 2N redundancy. Second, more modular designs mean that data centres can grow capacity with demand instead of over-provisioning from the start.  

Redundancy and efficiency shouldn’t be trade-offs. They’re the same design decision made well.  

4. Integrating digital monitoring 

The most common mistake in data centre design is focusing on the physical at the expense of the digital. True digital intelligence requires integrated monitoring being built into the data hall from the start, with unified data architectures enabling cross-system optimisations. For example, connecting each part of the centre’s infrastructure enables predictive maintenance, reducing unplanned downtime and therefore increasing efficiency. With real-time insights, adjustments can be made quickly and often remotely.  

Across our customers, we’ve seen that facilities with fully integrated digital infrastructure from the beginning consistently outperform retrofitted facilities; they offer up to 30% higher infrastructure utilization, 30% fewer false alarms, and 60% faster deployment timelines, all of which directly appeal to potential customers.  

5. Planning for certifications 

In the pursuit of OCP Ready™ Hyperscale certification or sustainability certifications like LEED, facilities that start these processes after commissioning find they are working with one hand tied behind their backs, because the evidence these certifications require is accumulated during the build process and they often can’t change key design decisions to accommodate the requirements.  

The key is thinking ahead and having answers to questions like the following in the design phase: 

• What rack density will you need to support in Years 3–5?  
• Is your power redundancy architecture genuinely 2N-capable?  
• Can your cooling system scale without a full redesign? 

These questions should inform today’s specifications to streamline certification processes. OCP Ready Hyperscale certification tells potential customers your facility is ready for present and future demands; LEED certification represents your sustainability commitment to a globally-recognized standard.  

Although SINES DC’s SIN02 is still under construction, it has already earned OCP Ready™ v2 Hyperscale certification alongside the operational SIN01. Start Campus already had the evidence it needed from the construction of SIN01 to showcase why SIN02 was equally deserving of the certification. And with an LEED Gold BD+C v4 certification at SIN01, customers can be confident that energy optimisation is and remains a priority.  

Being hyperscaler ready and energy efficient don’t have to be mutually exclusive for data centres. The five key design decisions above cover both goals. A facility designed for world-class PUE is, by design, ready to accommodate hyperscalers’ standards.  

The data centres being built today, the ones that will define whether or not projections of skyrocketing energy demand come to pass, are making these decisions before a go-live commitment forces them to compromise. The window to get these decisions right is at the very beginning.  

To learn more about our support of leading sustainable data centre projects, please visit our Start Campus story.