As demand for AI infrastructure accelerates, Saskia van Gendt, Chief Sustainability Officer at Blue Yonder, explains why developers must build greater resilience into procurement and project planning.

The unprecedented investment in AI data centres does not mean new capacity can be delivered quickly. Supply chains are already under enormous pressure: the International Energy Agency (IEA) recently reported that deployment “is increasingly coming up against a range of physical bottlenecks, limiting the rate at which data centres can expand in the near term.”

The IEA also notes that availability has tightened for “energy technologies such as gas turbines and transformers, as well as for advanced chips and IT components.” McKinsey adds that lead times in North America can reach 80 weeks for medium-voltage switchgear and 50 weeks for transformers, underscoring how quickly demand is straining industrial suppliers.

According to Deloitte, supply chain disruption is right up there among the biggest challenges facing data centre infrastructure build-outs, with 65% of the power and data centre executives it surveyed identifying it as a concern.

This acute dependency is compounded by sustainability challenges because the technologies intended to reduce the environmental impact of data centre development rely on many of the same constrained supply chains. For example, shortages of critical raw materials like nickel, silicon and rare earth elements can limit solar and other renewable energy production.

The physical bottleneck

These market-wide pressures are now appearing at multiple points in the data centre supply chain, from equipment manufacturing to installation and commissioning. Any one bottleneck can create difficulty, but data centre build projects are highly interdependent, so a single delay can quickly affect the wider programme.

Any data centre build is dependent on a network of manufacturers, component suppliers, logistics providers and infrastructure operators. Much of the equipment must arrive in a particular order, and a delay in the arrival of a critical component could disrupt the broader construction schedule.

At present, demand has grown so quickly that, as McKinsey puts it, “industrial equipment suppliers were not built to respond to today’s volatile data center demand.” The challenge is not linear. A data centre developer may see a delay at a direct supplier, while the underlying constraint sits several tiers deeper in that supplier’s own network.

As a result, data centre developers need a broader view of supply chain risk, not just whether their immediate suppliers can meet an agreed delivery date. The objective should be to identify potential problems early enough to adjust procurement or construction plans before disruption reaches the build.

The sustainability bottleneck

Much of this work is now taking place amid growing political and public concern about the environmental impact of building more data centres. Concerns about energy consumption, water use, land use and pressure on local infrastructure are increasingly influencing planning debates and the conditions attached to new developments.

While the policy response remains uncertain, the physical systems needed to reduce the environmental impact of AI infrastructure depend on industrial supply chains already under pressure.

For instance, any difficulties affecting the availability of grid equipment can delay the connection of cleaner generation. Battery storage and more efficient cooling systems also depend on specialist equipment that may be subject to long lead times.

Grid infrastructure is also needed to connect renewable generation and expand electricity networks, creating competition for the same components across the wider energy system. Separate IEA analysis found that procurement can take two to three years for cables and up to four years for large power transformers.

Where a project faces a delayed grid connection, developers may also face pressure to use another power source. The timing of supply chain delivery can therefore affect both a project’s completion date and its environmental outcome. Even when organisations have the capital and technical ambition to reduce their footprint, the physical means to do so cannot always be deployed quickly enough in the short to medium term.

Resilient by design

So what can be done to mitigate the impact of these bottlenecks?

The answer starts with building resilience into planning and procurement processes. That means identifying the equipment and supplier dependencies most likely to affect a project’s critical path, then understanding where a delay would have the greatest effect on construction timelines.

Supply chain risk insight should be closely linked to the construction plan, so that a change in availability can be assessed against the wider project schedule. This is where scenario planning can be particularly powerful as it can show the potential consequences of a delayed delivery before it disrupts work on site.

Developers and suppliers can use that insight to coordinate capacity and procurement decisions earlier. AI itself can help project teams assess how an emerging constraint in a complex supplier network could affect the wider project.

However, scenario planning is only effective if teams have also considered their mitigation strategy before a delay materialises. Securing an alternative source or adjusting the construction schedule can take time, so potential constraints should be identified well before they affect the build.

Communication is key, and developers need regular, shared updates from suppliers so that plans reflect current conditions rather than the assumptions made at the start of a project. Similarly, capacity and procurement decisions need to be coordinated with the wider plan, particularly where an equipment delay could affect the ability to connect cleaner power or deploy more efficient infrastructure.

This is why resilience cannot be added once a delay has emerged. It needs to be built in from the outset, because a project can only be considered resilient when the people managing it can recognise changing conditions and respond accordingly.