Solmar Insights
Announced data center projects across the United States now demand an estimated 780 gigawatts of power, exceeding the country’s all-time recorded peak grid load of 759 gigawatts hit in July 2025. Despite robust demand for new compute infrastructure, critical power limitations and persistent grid interconnection delays are set to keep much of this pipeline from moving forward at scale.
Key figures
780 GW in announced U.S. data center projects
759 GW all-time U.S. grid peak load (July 2025)
40% of AI data centers power-constrained by 2027
Announced pipeline surpasses grid
The scale of announced data center projects now rivals, and surpasses, national grid capability. According to data cited from Physical Superintelligence (PSI), the U.S. pipeline for new compute facilities totals roughly 780 GW. In comparison, the U.S. grid set a demand record of 759 GW in July 2025, illustrating that planned expansion demands more electricity than has ever been supplied simultaneously in U.S. history.
This imbalance underscores the scope of both digital infrastructure ambition and the challenges awaiting investors, hyperscalers, and developers. While project announcements continue at a brisk pace, most will struggle to achieve timely completion, primarily due to transmission and interconnection constraints rather than generation shortfalls. Few markets, aside from China, have remedied such bottlenecks, highlighting the significance of grid access in the U.S. buildout context.
Transmission backlogs and interconnection delays
Interconnection queues and transmission congestion have emerged as the leading obstacles to adding meaningful compute and power capacity. As new projects attempt to secure grid access, most face years-long waits. Market participants cited in the report note that grid transmission backlogs have eclipsed capital and permitting as the primary challenges to development.
These interconnection bottlenecks mean only a segment of the proposed 780 GW is likely to materialize by the end of the decade. Even projects with approved financing and advanced designs are subject to extensive wait times for grid hook-ups, resulting in a new urgency to explore alternatives to traditional grid-supplied power. Such delays continue to influence site selection, development timelines, and ultimately, the cost structures for operators and their institutional partners.
Rise of behind-the-meter solutions
The combination of multi-year interconnection queues and surging demand has prompted a significant shift toward on-site, behind-the-meter power generation. By 2027, projections point to 40 percent of existing AI data centers facing power operating constraints. This pressure is driving facility operators and developers to deploy behind-the-meter assets, including backup generation and renewables co-located on site, as a workaround to transmission roadblocks.
These investments allow for incremental expansion of compute capacity despite grid constraints, though questions remain as to their scalability and economic viability at the largest campus scales. Nonetheless, the trend for AI and high-density compute facilities to rely on on-site energy is expected to accelerate through 2029, as global data center infrastructure approaches an impending power wall within the next five to seven years.
Scrutiny of energy efficiency claims
As the race for power intensifies, market skepticism around vendor energy efficiency claims has grown. The dynamic described in the report suggests a ‘market for lemons,’ where unverified efficiency percentages are regularly disregarded by buyers, effectively discounting promised performance improvements to zero. In this environment, operational data transparency and rigorous verification become prerequisites for contract negotiations and procurement decisions.
Vendors are now increasingly expected to substantiate their performance metrics by matching predictive claims to withheld operational data sets, such as unused halls or periods prior to public release, before being accepted by institutional buyers. This marks a shift toward evidence-driven claims validation in an industry seeking to balance efficiency gains with ever-increasing power demands.
Impacts for data center investment strategy
Developers and investors now face a dramatically altered landscape in both acquisition and greenfield buildout. With overall demand for digital infrastructure continuing to rise, access to power and interconnection capacity has overtaken other inputs as the most acute chokepoint for U.S. data center growth. The value of grid positions, fast-track sites, and existing interconnection rights has increased correspondingly across the sector.
For institutional buyers, this situation calls for heightened diligence in evaluating project viability. Successful transactions increasingly hinge on the ability to secure near-term grid access, or on reliable behind-the-meter generation strategies for major new facilities. Asset strategies now weigh not only financial and operational attributes, but also the realities of capacity reservation and multi-year utility queues as limiting factors in portfolio returns and risk management.
What this means for buyers
Grid and data center capacity across the U.S. are subject to new constraints due to 780 GW of announced projects and a record 759 GW grid peak. Institutional buyers must recognize that power availability and interconnection delays, not funding or construction, will limit which facilities reach operation before 2029. Site selection and investment this quarter should focus on locations with established or expedited grid access, or viable behind-the-meter generation plans.
Reporting via the original publisher
