Article

The new age of contractual risk allocation in data center development

AI-driven data center demand is shifting contract risk, liability caps, and insurance gaps across development projects.

How hyperscale and AI-driven demand Is reshaping contractual liabilities across the development stack.

Risk in data center development used to flow in one direction and come to rest in a predictable place. The AI build-out has broken that logic. Demand-driven scarcity has inverted the old rule that risk settles with the party best able to control it: Today, the equipment maker, the utility, the specialist contractors may be most able to control delays, and yet they may have the most leverage to refuse it. Exposure is accumulating, unfunded, on the developer-operator’s balance sheet, and the durable fixes are no longer purely contractual. They can be financial.

This article is for the parties who now live inside that problem. Developer-operators, owners and sponsors, and the suppliers and contractors can be asked to absorb more than they once did. The central argument is simple: A contract clause is only as good as whether the exposure it creates can actually be borne or transferred. That makes the contract and the insurance program a single design problem, not two.

What is actually changing

For much of the last decade, data center development contracts followed relatively predictable risk allocation principles. Developers typically assumed responsibility for delivering a facility, contractors usually managed construction risks, and vendors supplied equipment under defined performance standards. While disputes certainly arose, the commercial framework generally reflected traditional construction and real estate development practices. But over the past 18 months there has been a step change within the ecosystem, as hyperscale and AI demand has turned speed to power into the scarce resource for which the market is competing. The shift has produced three emerging trends across the contract stack:

1. Liquidated damages are becoming larger and more frequent.

Delayed delivery provisions that once reflected modest rent abatements are, in some instances, now evolving into substantial daily penalties tied to customer business impacts including termination rights. Contractors and suppliers can be asked to backstop those obligations through corresponding delay damages in their own agreements.

2. Performance guarantees are expanding beyond traditional construction metrics.

Power availability, commissioning milestones, energization schedules, equipment lead times, and operational readiness dates can now be contractual obligations with significant financial consequences if missed.

3. Property and casualty risk is moving off the insurance program and into indemnities.

Some developers are increasingly seeking indemnification and liability protections from contractors and vendors for damage events that historically may have been borne through project insurance programs. Questions regarding builder's risk coverage, consequential damages, subrogation waivers, and insurance limits may be central negotiation points rather than boilerplate provisions.

Underlying all three is a single economic fact: for many hyperscale and large neocloud tenants, delays in facility delivery can translate into millions of dollars in lost revenue, stranded capital investments, missed customer commitments, and competitive disadvantages in the race to deploy AI infrastructure. As a result, these tenants can seek to transfer those significant risks contractually to developers. And some of the most aggressive demands may now come from newer, less-established counterparties: fast-growing AI and neocloud tenants whose credit on the way out may be thinner than their leverage on the way in. Owners and sponsors should be testing both sides of that asymmetry, not just the headline rent.

The Emerging Liability Cap Disconnect

Perhaps one of the most significant challenges facing the industry today is the growing disconnect that can occur between customer-imposed liabilities and the liability caps accepted by the parties responsible for delivering the project.

An example of this disconnect involves the termination right of a lease agreement between a datacenter developer and tenant for the failure to complete and commission a spec built data center on schedule or within a narrow grace period (90 days). This risk carries termination of the entire lease and rent revenue associated with the project. The developer can be unable to negotiate or obtain meaningful delay indemnification or reimbursement from the contractor or critical equipment vendors for delays potentially leaving the risk solely on the developer.

The conventional model is a chain: risk accepted upstream is “flowed down” to the general contractor, the EPC firm, the equipment suppliers and the commissioning providers. Flow-down language is necessary, but the chain image can mislead. A development project is not a chain but a stack of overlapping contracts that may respond to the same event in different and uncoordinated ways and at each interface a portion of the recovery can leak out through a cap, an exclusion, a different trigger or a defense.

Consider an example of a single, increasingly common event: the primary transformer for a campus slips its delivery date by several months, the utility energization date moves with it, and the tenant’s service-commencement obligation is missed. That one event strikes the whole stack at once, and the recoveries do not add up. The tenant’s delay remedy accrues against the developer from the missed date, indifferent to which component slipped. The contractor invokes an excusable-delay defense on the long-lead equipment. The transformer supplier’s liability is limited to repair, replacement or equipment value, with downstream delay and consequential losses excluded. And the insurance program, built to respond to physical damage, may not answer to a pure supply-chain or interconnection delay at all.

These leak points are not drafting errors; each counterparty is behaving reasonably in isolation. The gap is structural, and it comes from a mismatch of measures.

Upstream, Hyperscale and AI-focused tenants are increasingly negotiating delay remedies that can reach tens of millions of dollars on large campuses. In some cases, developers are accepting substantial liquidated damages, milestone penalties, rent credits, or service commencement obligations that bear a direct relationship to the customer's anticipated business losses.

Downstream, recovery is measured by something else entirely. General contractors typically seek liability caps tied to a percentage of contract value. Critical equipment manufacturers often limit liability to the value of the affected equipment. Design professionals maintain professional liability coverage that may be significantly lower than the developer's exposure to its customer. Many vendors expressly exclude consequential damages, business interruption losses, and delay-related damages altogether. The result can be a widening risk gap.

Developers may find themselves contractually responsible for customer liabilities that far exceed the recoverable amounts available from the parties whose actions ultimately caused the delay or failure. A transformer manufacturer facing a multi-year lead-time issue, a generator supplier impacted by supply chain disruption, or a contractor missing a commissioning milestone may have contractual exposure measured in millions, while the developer's customer exposure may be multiples of that amount. And because a single negotiation typically does not see the whole stack, a single negotiation may not close the gap.

This can force developers to move beyond simple flow-down contracting. Sophisticated organizations are increasingly focused on true risk alignment: testing downstream liability caps against actual customer exposure, identifying critical-path vendors that warrant enhanced remedies or security, evaluating project-specific insurance solutions, and trying to tie contractual remedies to risks each party can realistically control. The aim is not to win every clause, but to try to cause whatever risk is accepted upstream to be borne or transferred somewhere downstream. Which raises the harder question: borne or transferred where?

Borne or transferred where? The three-way constraint

There are three primary answers. Exposure a developer accepts can be retained on the balance sheet, transferred down the contract stack, or transferred to the insurance and capital markets. Sound risk allocation is simply the deliberate routing of each exposure to the destination that can actually hold it and the defining feature of this market is that all three routes are constrained at once.


Retain	Transfer by contract	Transfer by insurance
Hold the residual exposure on the balance sheet and price it into the deal.	Push the exposure down the stack through indemnities, liability caps and delay damages.	Move the exposure to the insurance and capital markets through cover or contingent products.
Constrained by: appetite, lender and rating-agency tolerance, and the sheer scale of capacity-linked exposure.	Constrained by: supplier leverage in a scarcity market. The party best able to control the risk may now have the most leverage to refuse it.	Constrained by: insurability. Pure delay with no physical-damage trigger sits largely outside traditional cover.

Two of those constraints are widely underestimated. The first is the leverage inversion: in a market short of transformers, switchgear, generators and skilled trades, the party best able to control a delay may also the party best able to refuse the risk or price it punitively. For suppliers and contractors this can be an opportunity as much as a threat. Those who can demonstrably take and manage a specific risk, backed by the right security and cover, will win work and command margin. The conversation is about who is genuinely positioned to hold what risk, not about winning a clause.

The second is insurability, and it is where contract and cover can often fail to meet. Delay-in-start-up cover, the policy usually relevant to delivery penalties, generally responds only to delay caused by insured physical damage. Pure schedule slippage, supply-chain disruption or a missed interconnection date with no physical loss typically falls outside the trigger, and sub-limits and waiting periods can narrow it further. So a delay remedy that looks acceptable on the page may only be truly acceptable if the resulting exposure can be retained, transferred or insured, and which of those holds depends entirely on how the clause is drafted. Drafting and placement are the same exercise.

Looking Ahead

The most important question may no longer be whether risk can be transferred. The question is whether risk is being allocated to the party best positioned to manage it and whether the portion that cannot be transferred has been honestly priced and funded. That can reframe contractual risk from a legal task into a capital-allocation discipline, and raises questions many organizations cannot yet answer crisply:

— Across the entire stack, what is your single largest uninsured delay exposure and which counterparty’s failure could trigger it?

— If your most critical long-lead supplier slipped two quarters tomorrow, how much of your customer-facing liability could you actually recover, and how much would land on your balance sheet?

— Are your delay remedies drafted in a way that your insurance program can respond to — or have clause and cover been designed by different people, on different days, against different assumptions?

— Which risks are you accepting simply because a powerful counterparty asked, rather than because you are the party best placed to manage them?

The developer-operators and owners and suppliers who lead the next phase of this build-out may not be the ones who negotiate the hardest clauses. They may be the ones who can answer those questions — because they treat the contract and the cover as one design, and price the residual risk deliberately rather than discovering it in a claim.

About this perspective

Marsh’s digital infrastructure contract advisory team works at the interface of contract drafting and risk transfer — stress-testing liability and delay provisions across the full development stack and helping to align them with the insurance and capital solutions that can determine whether those provisions are bankable. For a confidential review of how a project’s contracts and program fit together, contact the team.

Contact us to discuss how to better manage data center development risk.

Rows of data storage units are present in a large facility at night. The lights from the units create a blue glow, showing the technology in use. The setting highlights the scale of storage.

Report