AI Data Centers Strain Traditional Contract Models, FTI Warns

The explosive growth of AI workloads is exposing critical weaknesses in how data center projects are contracted and delivered, according to a new analysis from FTI Consulting.

European Union data centers consumed record electricity volumes in 2024, with usage projected to climb more than 60% by 2030—equivalent to powering several major European cities. AI-related data center power demand alone is expected to grow 47.9% annually between 2023 and 2027, creating unprecedented strain on infrastructure, supply chains, and the contractual frameworks that govern them.

Why it matters

As AI compute becomes a tradable commodity measured in tokens, service outages no longer represent minor inconveniences but potential revenue losses that cascade through entire value chains—from energy suppliers to tenants to end customers. Traditional service level agreement credits, designed to compensate for contained disruptions with billing adjustments, may prove inadequate when claims involve lost compute revenue calculated against token pricing schedules and facility utilization rates.

The old model breaks down

Conventional data center contracts allocate each risk to whichever party can best manage it: developers handle construction, utilities manage grid connections, suppliers own their timelines. Insurance and liability caps absorb residual exposure.

That approach worked when data centers were relatively small and stable in design. But today's AI facilities introduce interdependencies that span multiple parties simultaneously. Site readiness, grid connection delays, commissioning bottlenecks, and supply chain constraints now affect everyone at once, making clean risk allocation nearly impossible.

The construction process itself has fundamentally changed. Operators increasingly procure individual components—power systems, cooling infrastructure, racks—to maintain flexibility as technology evolves during multi-year build cycles. This fragmentation provides speed and adaptability but transfers integration risk and delivery coordination directly to operators.

A new governance framework

FTI Consulting recommends operators adopt stage-gate governance models that divide projects into phases with formal review points between each. These gates force different workstreams to align before proceeding, improving integration and reducing downstream issues.

At each gate, operators should assess whether critical dependencies remain achievable, whether risk still sits with the party best positioned to manage it, and whether changes elsewhere have created new exposures.

This approach requires robust supply chain management and real-time monitoring of costs, risks, and schedules—capabilities that function as early-warning systems rather than reactive damage control.

Interface risk emerges

As projects split across more contracts and vendors, disputes increasingly arise not from single-contractor failures but from gaps between contractual arrangements. Interface management—coordinating where one vendor's obligations end and another's begin—becomes critical.

The scale compounds the challenge. Major operators now maintain AI data center pipelines worth hundreds of billions of dollars annually, rivaling capital-intensive sectors like energy and infrastructure. At that scale, strong internal delivery capability becomes a competitive advantage, not just a risk management function.

The analysis, first published during London International Disputes Week in September 2026, warns that contracts signed today will govern disputes under conditions dramatically different from those that shaped traditional data center agreements.

The details were first reported by FTI Consulting in an article republished from London International Disputes Week.