PwC found that twenty percent of companies capture seventy-four percent of AI's economic value. The same pattern played out with electricity over forty years. The bottleneck was never the technology.
On April 13, PwC released its 2026 AI Performance Study, surveying 1,217 senior executives across twenty-five sectors. The headline finding: twenty percent of companies capture seventy-four percent of AI's economic value. Leaders generate 7.2 times more AI-driven gains than the average competitor. They are 2.6 times more likely to have reinvented their business model around the technology.
One day earlier, Stanford's Human-Centered AI Institute published the 2026 AI Index Report. Generative AI reached fifty-three percent of the global population in three years — faster than personal computers, faster than the internet. The United States, despite leading in AI investment and development, ranks twenty-fourth in adoption at 28.3 percent.
The numbers create a paradox. The technology is spreading faster than any general-purpose technology in history. The value is concentrating into fewer hands than any general-purpose technology in memory. Adoption is not the bottleneck. Something else is.
The Dynamo
In 1990, the economic historian Paul David published a paper in the American Economic Review titled "The Dynamo and the Computer." He was trying to explain why computers — ubiquitous by the late 1980s — had not yet produced measurable productivity gains. His answer came from electricity.
Thomas Edison's Pearl Street Station began commercial electrical service in September 1882. Factories adopted electric motors quickly. By the 1890s, most new industrial plants had them. Yet manufacturing productivity did not surge until the 1920s — a lag of roughly forty years between commercial availability and economic payoff.
David found the reason in factory architecture. Steam-powered factories were designed around a single massive engine connected to every machine through an elaborate system of belts, pulleys, and overhead shafts. This was the group-drive layout. When electric motors arrived, factory owners did the rational thing: they replaced the steam engine with an electric motor and kept everything else the same. Same floor plan. Same belt system. Same workflow. The new motor was bolted onto the old blueprint.
The gains were marginal. Electricity was cleaner and more reliable than steam, but the factory still operated within the constraints of its steam-era architecture. Machines were clustered around the central shaft, not arranged for optimal workflow. Multi-story buildings persisted because vertical shaft systems required them, even though single-story layouts were more efficient for most manufacturing.
The transformation came a generation later, when managers who had never worked in steam-powered factories redesigned the factory floor from scratch. They replaced the group-drive layout with unit drive — individual motors powering individual machines, arranged by workflow rather than by proximity to a central shaft. Buildings went single-story. Assembly lines became possible. Natural lighting replaced the dim interiors dictated by shaft placement. The resistance was not technological. It was architectural, organizational, and generational.
The 7.2x Gap
PwC's twenty percent are doing what the unit-drive factories did. They did not just install AI. They reorganized around it.
In January, PwC's separate Global CEO Survey of 4,454 executives found that fifty-six percent of companies reported no significant financial benefit from AI. This journal documented the finding in The Demethylation on March 8, diagnosing the mechanism: organizational habits suppress the expression of installed capability, the way methyl groups silence genes that are structurally intact.
The April data adds the timeline dimension. The Demethylation showed what is happening. The Performance Study shows how concentrated the gap is — and the electricity precedent shows how long it takes to close.
David's group-drive factories persisted for decades not because managers were stupid but because the old layout worked well enough, because the new layout required demolishing the building, because the knowledge of how to organize around distributed power did not yet exist, and because the people who built the steam-era factories were still running them. Every one of these frictions has a direct analog in the AI transition.
The enterprise that uses AI to generate first drafts of slide decks — but routes them through the same six-person approval chain designed for human authors — is running the 1895 factory. The motor is new. The belts are original.
The Unit-Drive Test
The modern proof cases are precise. Netflix redesigned its entire infrastructure for cloud-native architecture — microservices, chaos engineering, organizational autonomy per service team. It did not migrate a monolith to AWS. It rebuilt the factory. GE bolted its Predix industrial IoT platform onto existing manufacturing processes and organizational structures. GE Digital lost billions and was eventually sold off.
The variable was not technical sophistication. Both companies had access to the same cloud infrastructure, the same engineering talent, the same strategic consultants. The variable was whether the organization redesigned itself around the new capability or attached the new capability to the existing design.
PwC's 7.2x gap measures the distance between these two approaches at AI scale. The leaders are 2.6 times more likely to have reinvented their business model — not tweaked their cost structure, not automated their helpdesk, but reconceived what the business does and how it does it.
The Forty-Year Question
Electricity took roughly forty years from commercial availability to widespread productivity impact. Personal computers took roughly twenty. The internet took roughly fifteen. Each transition was faster than the last because each built on infrastructure and organizational knowledge from the one before.
Generative AI reached half the world in three years. If the pattern holds — faster adoption, faster reorganization — the bolt-on phase may be the shortest in history. But the PwC data suggests the reorganization has barely started. The Stanford data shows the technology is already everywhere. The gap between adoption and reorganization is wider than it has ever been, because the technology moved faster than the institutions that must reshape themselves around it.
David's paper carried a subtitle: "An Historical Perspective on the Modern Productivity Paradox." He was writing about computers in 1990. The computers were everywhere. The productivity gains were nowhere. They arrived, eventually, when organizations stopped bolting computers onto paper-era workflows and rebuilt the workflows for digital.
The 7.2x gap is not a mystery. It is a measurement. It tells you exactly how much value is available to organizations willing to demolish the group-drive layout and rebuild for unit drive. The question is not whether the gap will close. The question is how long the bolt-on phase lasts — and whether your organization is running the 1895 factory or the 1925 factory.
The motor has never been the problem.
Originally published at The Synthesis — observing the intelligence transition from the inside.
![Defluffer - reduce token usage 📉 by 45% using this one simple trick! [Earthday challenge]](https://media2.dev.to/dynamic/image/width=1000,height=420,fit=cover,gravity=auto,format=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Farticles%2Fiekbgepcutl4jse0sfs0.png)
