Blue digital background with glowing binary code and data streams
Insights

The Deal Behind-the-Meter: Vertical Integration Reaches the Grid

Overview

Completed in March 2026, Alphabet’s acquisition of clean energy developer Intersect is the first instance of a major technology company acquiring a renewable energy development platform. The deal illustrates a broader trajectory in which power availability has become a limiting factor in new data center development, prompting hyperscalers to shift from procuring energy to internalizing its production.

As hyperscalers deepen their integration into energy markets, the acquisition raises questions about cost allocation—namely, whether infrastructure built to serve hyperscaler demand will ultimately raise costs for customers remaining on the shared utility system. Existing regulatory frameworks evaluate energy markets and digital infrastructure competition in separate jurisdictions, and neither is designed to assess vertical integration that spans both. 

Background

In the 1990s, large industrial customers pushed the Federal Energy Regulatory Commission (FERC) to dismantle vertically integrated utility monopolies that controlled generation, transmission, and distribution simultaneously—and used that control to favor their own generation and block competitors. FERC responded with open-access rules that opened generation to competition while maintaining regulated transmission and distribution, thereby creating a framework based on shared grid infrastructure. 

That architecture now faces a challenge for which it was not designed. Data centers are the fastest-growing source of electricity demand on the US grid, yet the interconnection process governing how new generation and load connect to the shared system has become the binding constraint on expansion. Only 14 percent of the generation and storage capacity requesting grid connection between 2000 and 2018 reached commercial operation by 2023; by the end of that year, roughly 2,600 gigawatts—more than twice the country’s total installed generating capacity—remained actively queued across major US grid operators.

Hyperscalers have historically secured energy through long-term contracts with independent generators, routing power through the shared grid. Faced with multiyear delays, they are now shifting toward direct asset ownership. This approach is not entirely new. In 2014, Apple acquired a small hydroelectric project in Oregon near its data center campus to supplement supply at that facility. Amazon structured a special purpose vehicle called Oregon Solar 1 to acquire the Sunstone solar project, the state’s largest planned solar development, in a bankruptcy auction. Each case involved acquiring a discrete asset within the existing grid framework. 

More recent developments have bypassed the grid entirely. Meta’s Ohio data center campus includes $1.6 billion in on-site natural gas generation, developed in partnership with Williams, a major pipeline infrastructure company. In 2024, xAI deployed unpermitted gas turbines at its Colossus facility in Memphis after the site’s grid connection could supply only a fraction of the power it needed to begin operations.

Alphabet’s $4.75 billion acquisition of Intersect is structurally different. Meta and xAI secured power for individual facilities. Alphabet purchased the capacity to do so repeatedly and at scale: a multi-gigawatt pipeline of solar, wind, and storage projects designed for co-location with its own data centers, along with the team to build more. When generation and consumption occupy the same site and electricity never enters the public grid, it operates “behind the meter”—outside the wholesale market FERC regulates, outside the retail relationship state commissions oversee, and outside the cost-sharing framework that funds the grid everyone else depends on. 

Why Is This Important?

Data center operators accounted for 43 percent of all clean power purchase agreements signed in 2024. The top five hyperscalers have contracted over 40 gigawatts of US renewable energy, representing two-thirds of the total corporate renewables market. Alphabet's acquisition of Intersect goes beyond contracting: It converts a development platform that would have served multiple buyers into proprietary infrastructure for a single firm. Amazon recently outbid Puget Sound Energy, a regulated utility serving a customer base that is half low-income, for a solar facility. The competition now extends to existing supply, where utilities face well-capitalized competitors. 

These shifts redistribute costs across the broader grid. When large customers secure dedicated generation and reduce reliance on shared infrastructure, fixed costs must be recovered from a smaller base of remaining users. FERC twice rejected Amazon’s proposed co-location at the Susquehanna nuclear plant in Pennsylvania, an arrangement estimated to shift up to $140 million annually in transmission costs onto other ratepayers across PJM, the nation’s largest regional grid operator. The Intersect transaction sidesteps that precedent by operating primarily within the Electric Reliability Council of Texas, which falls outside FERC’s wholesale market jurisdiction. No federal agency has challenged vertical integration between technology and energy markets.

In September, this series examined a shift in technology mergers and acquisitions toward capability-driven deals that bypass traditional merger review. That logic now extends to physical infrastructure. But where licensing agreements and team transfers acquire capabilities within existing market structures, vertical integration into energy production restructures the market itself. 

What Happens Next?

In June 2025, Texas—the second-largest US data center market—enacted SB 6, tightening requirements for large-load interconnection and introducing regulatory review of co-located generation. The law requires behind-the-meter loads above a specified threshold to pay retail transmission charges but delegates the specific rate-setting methodology to the Public Utility Commission of Texas, with no fixed rulemaking deadline. 

PJM moved more directly in its territory, proposing a 50-megawatt threshold in February 2026 that would limit the use of behind-the-meter netting for large loads on its grid. State legislatures are also active: more than 190 state-level bills addressing data center energy use were introduced in 2025 alone. 

Federal direction diverges. At a White House event in March 2026, Amazon, Google, Meta, Microsoft, OpenAI, Oracle, and xAI signed the nonbinding Ratepayer Protection Pledge, committing to cover the costs of new generation required for their data centers. The pledge frames self-supply under a "build, bring, or buy" framework that treats the three options as equivalent. They are not.

Buying (procuring from the grid) and bringing (contracting dedicated supply) operate within shared infrastructure; building bypasses it. Industry analysis from Cleanview identifies 46 planned US data centers, representing roughly 56 gigawatts—about 30 percent of planned US data center capacity—that intend to generate their own power. Ninety percent of those projects were announced in 2025.

The US electricity system spent decades separating generation from distribution. As hyperscalers reshape how their facilities are powered, the regulatory frameworks governing that system must keep pace.