Loading
Two powerful forces are reshaping the U.S. economy in 2026.
Archive edition · Market data and company circumstances reflect 21 February 2026, when this newsletter was sent.
Two powerful forces are reshaping the U.S. economy in 2026.
One is accelerating at extraordinary speed. The other has been stagnating for decades.
Artificial intelligence is driving a surge in electricity demand. Construction, the industry responsible for building the physical backbone of that surge, has barely improved its productivity in half a century.
Together, they reveal something much bigger about the US economy.
Electricity prices have been rising meaningfully faster than overall inflation. And this trend may persist. Why? Because AI is not just software, it is industrial scale computation.
Data centers already account for roughly 7% of total U.S. electricity consumption, and their footprint is expanding rapidly as AI workloads, cloud computing, and high-performance processing scale up. Over the next five years, these facilities are expected to drive a disproportionate share of incremental power demand, becoming one of the primary structural sources of load growth in the grid.
Data centers are driving rapid growth in US power demand.
The challenge, however, lies on the supply side. Electricity generation capacity cannot be expanded overnight. Building new power plants (whether gas, nuclear, or renewable) requires years of planning, permitting, financing, and construction. Transmission and grid interconnection approvals often stretch across multi-year regulatory timelines.
At the same time, key components such as gas turbines and other specialized equipment are facing supply constraints, while the availability of skilled technical labor remains limited and slow to scale. The result is a structural mismatch. Demand is accelerating quickly, while supply expansion remains constrained by physical, regulatory, and human capital bottlenecks.
Electricity inflation is expected to remain elevated through 2026–27 before easing as supply gradually improves and fuel costs normalize.
Higher electricity prices will:
But the drag from electricity costs is small as compared to the long-run productivity gains AI is expected to generate. This is short-term friction within a long-term transformation.
Let’s take a look at the other half of the equation: Construction.
Since the 1960s, labor productivity in U.S. construction has largely stagnated and, in some studies, has even declined, even though the productivity across the broader economy has marched steadily higher.
That gap is more than a statistical curiosity. It means the industry tasked with building the physical backbone of modern growth — housing, factories, transmission lines, substations, pipelines, and now hyperscale data centers — has not become meaningfully more efficient for decades.
So what is holding it back?
1.2.1 Limited Innovation Adoption
Much of the heavy equipment used on job sites would look familiar to a foreman from the mid-20th century. Tools have improved but the underlying workflow is still labor-intensive, bespoke, and site-specific.
Adoption of industrial machines
Prefabrication and modular construction have long been pitched as the breakthrough yet adoption remains uneven and far below its potential. Unlike sectors such as technology, advanced manufacturing, or pharmaceuticals, construction has captured relatively little spillover benefit from the wider innovation cycle.
1.2.2 Regulatory Friction
Land-use rules, zoning restrictions, height limits, environmental reviews, approval delays, and impact fees have expanded over time, and the cumulative effect is powerful. When approvals take years, projects do not just get delayed; they become riskier, more expensive, and harder to finance.
These frictions raise costs, discourage new entrants, reduce competitive pressure, distort where capital flows, and slow the economy’s ability to respond when demand spikes. In practice, the constraint is not only “can we build?”, it is “can we get permission to build in time for it to matter?”
It is fair to argue that some improvements in housing quality may not be fully captured in productivity statistics. But even after accounting for quality adjustments, the long-run underperformance remains hard to dismiss. The core reality is visible on the ground, construction has not scaled with the modern economy in the way other industries have.
And that matters a lot because when demand accelerates (whether it is housing shortages, grid upgrades, or data center buildouts), the economy’s “build capacity” becomes the bottleneck.
Costs rise faster than plans can be executed. Timelines stretch. And the sectors that depend on physical infrastructure end up constrained not by ideas or capital, but by the slow, expensive process of turning blueprints into reality.
AI is not just a digital trend; it requires a massive physical rollout including new data centers, transmission lines, power plants, cooling capacity, and stronger grids.
But the industry tasked with delivering that infrastructure is structurally slow. Construction productivity has barely improved for decades, and approvals, interconnection queues, and permitting add years to timelines.
That mismatch can quickly result in a bottleneck. AI demand can compound quickly, while physical supply expands step-by-step. When demand accelerates faster than generation and grid capacity can be built, the system tightens and prices do the balancing.
That is why electricity markets are already showing upward pressure. It is the economics of scarcity meeting exponential growth.
The real macro question is not whether AI will lift productivity, it almost certainly will. The harder question is whether the U.S. can build fast enough to power it.
AI adoption is ultimately constrained by physical capacity. If construction productivity improves and bottlenecks ease, supply can scale faster. That would help electricity inflation cool, reduce regional grid stress, and allow AI to diffuse more broadly across the economy.
But if productivity stays weak and timelines remain clogged, energy becomes a structural cost constraint. Power-hungry regions pull ahead, others fall behind, and infrastructure turns into a persistent source of inflation pressure rather than a platform for growth.
This is not just a technology story anymore. It is an infrastructure story.
This “grid versus crane” dynamic sets up a handful of durable, multi-year themes because AI growth is increasingly constrained by physical throughput, not just computing breakthroughs.
The digital economy is now running on physical constraints. Copper, turbines, permits, transformers, and transmission lines matter almost as much as algorithms because they determine how fast AI can actually scale in the real world.
This article preserves the analysis in our weekly newsletter sent 21 February 2026. Market prices, forecasts and company circumstances reflect the time of publication and may have changed.
This material is general information, not personal financial advice or a recommendation to trade. Investing and trading involve risk, including loss of capital.