The next AI data center might hang on your living room wall.

Brian Jamison, Vice President of Strategic Procurement at PulteGroup, endorsed the partnership in the announcement, stating, “Installing SPAN panels, XFRAs, and backup batteries reduces home construction costs—and enables underutilized residential electrical infrastructure to serve the grid.” The business model behind this deal is highly unconventional. Homeowners do not pay extra upfront for the XFRA installation; the hardware is provided by Span and embedded by PulteGroup during construction. After completion, homeowners pay Span a fixed monthly fee covering electricity and home Wi-Fi. In return, Span reimburses them—based on actual usage—for the electricity and network bandwidth consumed by the XFRA node. A Span spokesperson told Realtor.com, “In some cases, there may be no cost at all.” Yet upon closer inspection, this arrangement makes solid sense for all four parties involved. Homeowners gain lower utility bills plus a hands-off passive income stream; builders acquire a compelling marketing edge (“Homes with Span help offset your utility bills”) and procurement cost savings; Span acts as an intermediary—charging cloud providers for compute power, distributing revenue shares to homeowners, and keeping a margin; and Nvidia, as the GPU supplier, gains access to a new sales channel for its Blackwell architecture—beyond traditional hyperscalers. For cloud customers, it’s also a smart deal: when conventional data centers face shortages of power, land, and community approval, buying distributed compute from Span offers a viable bypass around these bottlenecks. And this shortcut finds willing buyers because—by 2026—the most defining trend across the U.S. data center industry is clear: electricity has become scarcer than GPUs. According to Sightline Climate, roughly 12 GW of data center capacity was originally slated to come online in the U.S. in 2026—but only about 5 GW has actually entered construction. The remaining two-thirds remain stalled at land acquisition, power interconnection, or community permitting stages. PJM Interconnection—the largest U.S. grid operator, serving 13 states and 65 million people—forecasts a 6 GW shortfall in reliability capacity by 2027. Delivery timelines for high-voltage transformers have stretched from 12–18 months to 36–48 months. As reported by Bloomberg, OpenAI’s and SoftBank’s Stargate project still shows no significant physical construction progress—not due to lack of funding, but because power supply, transformer availability, and grid capacity are constraints money alone cannot instantly resolve. Marc Spieler, Global Energy Industry Senior Director at Nvidia, candidly told CNBC that achieving the massive load requirements of a large-scale data center at a single location has already become “difficult.” XFRA takes the opposite approach: leveraging the portion of grid-connected capacity already installed in every home—but currently underutilized.

PJM, the largest U.S. power grid, advances data center power-supply solutions—capacity pricing surges over 1,000% in two years.

According to a Reuters report on Wednesday, PJM members held a non-binding vote on June 30 regarding more than ten “backstop procurement” proposals. One proposal—jointly advanced by data center advocates and major regulated utilities—was selected for advancement. It proposes launching the procurement process on September 10, 2026, with submissions closing on November 20, 2026—aligning with PJM’s previously announced timeline. The voting outcome will serve as advisory input submitted to the PJM Board of Directors; however, final policy direction and specific terms remain subject to the Board’s decision. PJM serves 13 U.S. states and the District of Columbia, delivering electricity to approximately 65 million people. Its persistently strained supply-demand balance now poses direct threats to regional power reliability and economic efficiency. Surging data center demand intensifies supply-demand imbalance, driving up capacity prices Over the past two years, large technology firms and data center developers have filed a sharp surge of interconnection applications with PJM. The large-scale integration of energy-intensive data centers is threatening to exceed regional power supply capacity. This imbalance is directly reflected in pricing. PJM’s capacity price—the payment made to power generators to ensure sufficient electricity supply during peak demand—has surged by over 1,000% since approximately 2024. This dramatic rise signals sharply increased system-wide costs to maintain reliability, which may ultimately be passed on to end users. Backstop procurement proposal advances—with long-term contracts prioritized

Biggest US Grid’s Data-Center Plan Accepted With 5-Month Pause

Sep 29, 2026

FERC accepted PJM Interconnection's power supply plan but suspended it for five months to address cost allocation concerns amid surging AI-driven data center demand. Regulators insist existing customers should not bear costs from new demand while ensuring regional reliability.

bloomberg

2026-09-29

Biggest US Grid’s Data-Center Plan Accepted With 5-Month Pause

2026-09-30

FERC asks grid operator PJM to revise plan to shield homes from data center costs

Biggest US Grid’s Data-Center Plan Accepted With 5-Month Pause

FAQ