Cushman & Wakefield counts 285 GW data center pipeline as local constraints shape growth

Cushman & Wakefield's Americas Data Center Update, covering the first half of 2026, counts 37.7 gigawatts of data center capacity under construction across the Americas against 50.3 GW operating and a planned pipeline of 285 GW. Connect CRE reported the figures, and the number carrying the most consequence for capital is not a gigawatt count at all: it is 91.7%, the share of the construction already precommitted. The buildings going up are mostly spoken for. The power that runs them is not, and neither is the transmission that carries it, which is why the residual risk on an active data center project now sits with parties a real estate underwriter rarely negotiates with.

The report frames the shift as a market-by-market proposition rather than a regional one, naming four forces: power allocation, transmission constraints, infrastructure funding obligations and regulatory environments. Together they produce what the brokerage calls a "more varied growth environment than before." Demand is not what is in question. Artificial intelligence, cloud computing and enterprise digital transformation will keep pulling load through the Americas, according to the report; whether infrastructure readiness, regulatory certainty and power availability let that load land where an operator wants it is what has stopped resolving at the national level.

Operators keep returning to Virginia, Atlanta, Chicago and Dallas, and the report's explanation deserves attention from anyone holding capacity in those places: "the advantages these data center ecosystems provide remain difficult to replicate." The advantages it lists are network density, hyperscaler concentration, talent pools and established infrastructure. Ten markets count more than a gigawatt of capacity under construction, and four of those sit in Texas, a concentration the report records without explaining.

Secondary and tertiary venues are drawing operators as well, among them Cheyenne, West Texas, Alberta and Pennsylvania. The report does not rank what draws each one, though the forces it identifies as governing development point to availability as the offset: a market that can hand over power sooner can compensate for thinner network density and a shallower talent pool, at least for workloads that tolerate the distance. That logic cuts both ways for an acquirer. If the anchors' advantages are hard to replicate and their grids are constrained by allocation and transmission, then energized capacity already standing in those markets is supply a later entrant cannot quickly match, which argues for buying in the anchors and building where the grid still has headroom. The report stops short of drawing that conclusion; it names the forces and leaves the reader to finish the sentence.

For allocators, that dispersion changes what a diversified private real estate sleeve actually buys. A fund holding campuses in four markets holds four utility queues and four regulatory postures, and the correlation between those positions looks lower than the sector label implies. That argues for underwriting at the market line rather than the sector line, and for asking a sponsor specifically which utilities and which states stand behind the pro forma.

The planned pipeline sits behind the precommitment figure rather than inside it. The 91.7% applies to the 37.7 GW under construction, not to the 285 GW of planned development, and 285 GW against 50.3 GW operating works out to a pipeline roughly 5.7 times the installed base. If the same precommitment rate held across all active construction, something in the neighborhood of 3 GW would still be unspoken for, and that slice is where a cooling in tenant demand would show up first. The rest of the pipeline depends on power and on tenants, which is why the report treats it as planned.

Near term, execution crowds out everything else. With 91.7% of active construction contracted, the live questions are whether the utility energizes on schedule, whether transmission upgrades and substations land inside budget, and which party pays for them. The report's inclusion of infrastructure funding obligations among its four governing forces points the same way. In a market with a mature grid that spending is largely settled before a sponsor arrives; in a newer venue it likely falls closer to the development budget, carried in the project's return rather than recovered in rent. The report does not disclose pricing or terms behind the precommitment rate, so how much of that exposure a given sponsor has already contracted away cannot be read from it.

For private real estate capital the practical consequence is that data center returns are becoming local and less tied to the sector's aggregate growth rate. A sponsor underwriting a campus is also underwriting a specific utility's interconnection queue, a specific state's regulatory posture and that market's share of allocated power, which is why a report at this level of aggregation keeps those forces at headline level. The brokerage's own summary lands there: "the Americas data center market is becoming more complex and geographically diverse than ever before," and "each market will manage growth in a way that best suits its own interests." The next update, likely covering the second half, is where the 285 GW becomes measurable, in how much of it has entered construction and how much of the construction still has no tenant attached.