The world's energy system is quietly rewiring itself. The dominant new energy sources and loads are natively direct current (DC): solar PV generates it, batteries store it, and data centers and EVs consume it. Yet the distribution infrastructure connecting them still runs on alternating current (AC), the architecture that has prevailed since the 1890s.
That mismatch was once an acceptable inefficiency. Every AC-to-DC conversion loses a small percentage of energy, but the toll is so modest that it barely mattered in the past. The rapid scaling of AI computing power (known familiarly as AI compute) is changing that calculus. To keep up with demand, hyperscale campuses will need to operate at gigawatt scale, magnifying the cost of unnecessary conversions. These facilities force every watt through multiple conversion stages before it reaches a server—stages that exist not because they are necessary but because the infrastructure predates the loads it now serves.
In four sectors—data centers, industrial plants, commercial buildings, and residential buildings—the shift toward greater DC integration is already underway. The destination is not an all-DC grid, but a hybrid architecture: AC remains the backbone of most transmission and regional distribution, while DC increasingly powers high-density applications, mostly behind the meter. Across sectors, the urgency, the economics, and the window for action differ.
A paper published by BCG in collaboration with ABB, quantifies the economics in areas where decisions about adopting new electrical architecture are already live, and where they soon will be:
- For AI data centers, DC is rapidly becoming the required architecture. It meets the power density demands of next-generation AI hardware, fits more compute behind a scarce grid connection, and turns efficiency gains directly into revenue.
- For automation-intensive industrial plants, the business case for hybrid architecture is positive from the first day of operation, driven by the DC-native loads that most factories already contain.
- For large commercial buildings, DC pays off by creating a backbone for major loads such as elevators, climate-control systems, and energy storage while leaving tenant-facing systems on AC.
- For residential buildings, the business case for hybrid architecture is not yet compelling, but policy shifts on the calendar may improve the economics.
The barriers to faster adoption are systemic rather than technical, which means that this is more of a coordination problem than an engineering issue. Equipment and automation suppliers, hyperscalers, industrial operators, and large developers that commit capital and set specifications over the next two to three years will shape the transition rather than inherit it. Their decisions will influence the cost curves, voltage defaults, and standards that define the next decade. The question that each organization must answer is whether it will help set the standards or merely accept them as developed by others.
