The number behind the squeeze
JLL's Global Construction Costs Index 2026 puts data center construction cost escalation at an annual 3 to 6 percent, roughly double the 2 to 3 percent that was typical before 2020. JLL separately forecasts global tender prices for construction broadly, not data centers specifically, to rise 3.1 to 4.2 percent in 2027, including 3.2 to 4.2 percent in North America. The firm's underlying point is that higher asset values and strong AI demand still leave individual projects exposed to this inflation, since passing the cost through depends on project-specific conditions, not a market-wide average.
Matthew Giddings, JLL's executive director of technical services for data centers and critical environments, put it directly: "Higher asset values alone cannot close that gap," and "strong demand supports the market, but it does not make every project viable." That is a notable statement from a firm whose business depends on the sector staying healthy. It reads as a warning that the AI buildout's headline demand numbers are masking real variance in which individual projects actually pencil out.
Where the cost is actually coming from
The report names labor, not materials, as the dominant pressure. Electrical and mechanical trades, particularly electricians, high-voltage specialists, pipefitters, and advanced cooling technicians, create the largest cost and schedule risks on a modern data center build. Controls, commissioning, and startup teams are also bottlenecks, and competition for utility and substation crews is intensifying as every hyperscaler and neocloud competes for the same finite pool of specialized workers. Long lead times for critical equipment, transformers, switchgear, generators, cooling systems, compound the problem: developers who order early to protect schedule can get burned if a tenant's technical requirements change before the equipment arrives.
This labor dynamic explains why remote or newly developing data center markets face a double cost penalty. Beyond the base construction inflation, JLL notes these markets often require housing, transportation, and travel incentives just to attract the specialized trades a modern facility needs. A site with cheap land and abundant power can still end up expensive to build if it cannot draw the electricians and commissioning specialists the project requires, and that gap is widening rather than closing as demand for those trades grows faster than the labor pool.
Which projects actually absorb this and which don't
JLL identifies four categories of projects most exposed to this cost pressure: speculative campuses built without an anchor tenant, developments with uncertain power delivery timelines, projects locked into fixed-rent agreements with little room to pass through cost increases, and older facilities being repositioned for AI workloads where the retrofit costs can quietly erode the underlying economics. None of these are hypothetical categories. Every one describes a real, common deal structure in the current market, which means a meaningful share of the pipeline announced over the past two years carries this exposure right now.
The flip side is informative too: projects with secured power, a committed anchor tenant, and a credible delivery path can pass costs through to tenants because the underlying demand is locked in rather than speculative. That is exactly the distinction a CTO evaluating a colocation deal or a build-to-suit proposal should be pricing into the negotiation, because it tells you which party actually holds leverage once ground breaks and costs start moving against the original estimate.
What this means at the negotiating table
If you are negotiating a new lease or build-to-suit agreement today, assume the developer's cost base is less stable than the quoted rate implies, and structure the deal accordingly. JLL's own recommendation to developers, phased construction tied to confirmed power and customer commitments, is useful information for a tenant too: a developer using phased construction has more flexibility to absorb cost shocks without passing them to you than one that committed to a fixed full build before securing financing certainty. Ask which phase your capacity falls into and whether that phase's cost basis is already locked.
Push for rent structures that separate a base rate from a cost-indexed escalator tied to a transparent, auditable benchmark, rather than accepting an opaque annual increase the developer sets unilaterally. If a developer resists specifying how escalators are calculated, that resistance itself is a signal about how much of the current cost pressure they expect to pass through to you later, under terms you will have far less leverage to contest once you have already moved workloads into the facility.
The build versus lease decision this reopens
For any enterprise large enough to be weighing a dedicated build against a colocation lease, this data shifts the calculus toward phased, flexible commitments over either extreme. A full self-build now carries the same labor and equipment lead-time exposure JLL describes for developers, without a landlord to share the risk. A long, fixed-rate colo lease signed today risks locking in a rate that does not reflect where construction costs will be in year three of the term, in either direction.
The middle path, a shorter initial term with structured renewal and expansion options tied to actual delivered capacity rather than promised capacity, gives you the ability to renegotiate as the labor and equipment market shifts rather than betting your infrastructure budget on where JLL's cost index happens to be the day you sign. Given that the index itself is forecasting continued escalation into 2027, locking in flexibility now is worth more than it would have been two years ago, when the pre-2020 rate of cost growth made a long fixed term a reasonably safe bet rather than the gamble it has since become.
