The pace of AI transformation is often framed around the pace of model development, but the physical systems that are needed to responsibly scale that progress continue to operate across fundamentally different timelines.

Power and data center supply chains remain critical bottlenecks¹ for deployment and adoption, and the complex ecosystem of constraints has expanded to include transmission infrastructure, interconnection queues, permitting, equipment availability, community engagement, water considerations, and skilled labor, creating additional challenges for build-out.

Goldman Sachs Research now expects global data center power demand to increase by 170% between 2025 and 2030, with more than 60% of that growth coming from within the United States. ² But a constrained global supply chain may serve as a moderating force, with implications for the overall pace of AI growth and requisite capital fueling it.

“Power is probably the forever bottleneck,” says Jason Tofsky, global head of Digital Infrastructure Investment Banking. “There are a lot of other bottlenecks that exist, and to get any one of these projects right, you have to get dozens of things right over a six-month to two-year time frame, including financing.”

From Compute Constraint to Systems Constraint

AI development continues to demand enormous amounts of compute, which, in turn, depends on a complex chain of physical infrastructure decisions. Baseline aggregate AI CapEx estimates hover around $7.6tn of capital between 2026 and 2031 across compute, data centers, and power. ³ Goldman Sachs Research is projecting $420bn of global hyperscaler debt issuance in 2027, representing a more than 60% increase from their FY2026 estimate. Notably, hyperscalers represent only 40% of the AI ecosystem credit raises. ⁴

A grid-connected data center requires approval from regulatory authorities, while a behind-the-meter project may avoid some grid constraints but can face air-permitting requirements and longer environmental reviews. At the local level, traffic, noise, emissions, water use, and the scale of construction concerns can meaningfully affect project timelines.

Building a gigawatt-scale campus in a smaller community can reshape local roads and public-service demands almost overnight.

The projects that move most effectively will likely address these considerations early through credible community-benefit commitments and transparent plans for environmental impact.

The Grid Is Now the Critical Path

The power system itself is also facing a growing set of constraints. The US interconnection queue has swelled to roughly 2,600 GW of generation and storage projects, more than double the installed capacity of the existing grid,⁵ although this figure includes projects at varying stages of maturity and potential overlap, says Tyler Miller, global co-head of Power, Utilities, and Infrastructure Investment Banking. The median time from interconnection application to commercial operation is now roughly five years, with some projects taking longer.

The timeline challenge is not only the size of the interconnection queue, but also the pace at which the physical grid is expanding. US construction of new high-voltage transmission lines fell from an average of 1,700 miles annually between 2010 and 2014 to just 350 miles annually from 2020 to 2023, ⁵ according to Goldman Sachs Research.

Transmission constraints are compounded by permitting and equipment shortages. Miller highlights that environmental reviews can take three to four years, while the lead times for high-voltage substations and generation step-up transformers have stretched to roughly three to five years.

Local opposition adds another layer of complexity. Since January 2026, 86 projects have been canceled or extensively delayed, with 76 directly linked to local opposition. The concerns vary by market, ranging from water and emissions to noise, traffic, and potential pressure on residential electricity bills. The result is a growing interest in hybrid and behind-the-meter solutions.

Repositioning Utilities for an Era of Higher Growth and Even Greater Capital

For utilities, the implications of ongoing changes are profound. What was long viewed as a relatively low-growth, defensive sector is being repositioned as a capital-intensive engine of digital and real economies. Utility capital plans are increasing and expected to reach an aggregate $1.3tn between 2026 and 2030. ⁵

Capturing this opportunity requires a new playbook. Utilities and their large-load customers are increasingly considering structures in which hyperscalers contribute directly to the cost of generation, transmission, and substation infrastructure required to serve their facilities. The objective is much broader than just allocating costs; it is to ensure that new digital infrastructure strengthens the larger power system and delivers tangible benefits to local economies. The central principle is straightforward: “Growth enables growth,” Miller says.

The demand for innovative financing across digital infrastructure and power has only grown more acute, as capital solutions must coordinate across land, development, power infrastructure, equipment, construction, and long-term operations. This new playing field calls for more flexible partnerships among utilities, hyperscalers, infrastructure investors, private-credit providers, insurers, and other long-duration pools of capital. As the bottlenecks evolve, capital solutions must do the same.

This capital formation feeds a self-reinforcing cycle: More capacity enables faster model development and broader deployment, which, in turn, drives more demand. At the same time, model providers, hyperscalers, and chip companies are increasingly moving beyond their original layers in the AI stack.

The Human Capital Constraint

The infrastructure build-out also requires people—and in some markets, competing projects are bidding up wages to attract the same finite pool of workers.

Labor shortages can disrupt even well-capitalized projects that have already secured land, power, and equipment. While finite labor may constrain the pace of deployment, the build-out is also creating well-paying, skilled blue-collar jobs across the country, broadening the economic benefits of AI beyond the technology sector. As a result, the AI infrastructure race has become a race to develop the workforce needed to build and operate this new industrial base.

Power Sets the Pace

If power remains the binding constraint, AI growth will move at the speed of power delivery. Model capabilities advance in months; scaling generation, transmission, and equipment could take years. A likely result is a longer, less linear build-out, with capacity arriving in steps as more generation comes online. Longer timelines delay returns, increase risk, and widen the gap between projects with secured power and those still waiting. The same constraint imposes discipline: A more measured build-out allows demand to absorb investment over time, reducing the risk of stranded capital.

This scenario introduces several important considerations: How much of this timing is already reflected in growth expectations? For markets that can add generation and grid capacity more quickly—for example, China—competitive advantages could emerge.

The AI build-out continues to be as much a coordination challenge as a technological one. Significant advantages lie in addressing constraints together and early: engaging communities before seeking approvals, securing power and equipment ahead of need, developing the workforce, and structuring capital so that new demand funds the capacity it requires.

Endnotes

¹ Powering the AI Era, Goldman Sachs Investment Banking (2025)

² Masanet et al. (2020), EIA, EMBER, IEA, Cisco, Goldman Sachs Global Investment Research

³ Harnessing AI for the Real Economy, Goldman Sachs Investment Banking (2026)

⁴ Source: Dealogic (ION Analytics), Goldman Sachs Global Investment Research. As of September 15, 2026. Excludes data center financings issued by special purpose vehicles.

Hyperscalers include Amazon, Alphabet, Meta, Microsoft, and Oracle.

⁵ Goldman Sachs Global Investment Research