PJM brings in Google and Tapestry to put AI on its interconnection backlog

April 2025 — PJM Interconnection signed a multiyear collaboration with Google and Alphabet's Tapestry to use AI tools to speed reviews of requests to connect new generation to its grid. Here's what happened, what it cost, and what went wrong.

Cost and outcome (as of April 10, 2025 (PJM); Tapestry page undated)

Outcome: a multiyear AI collaboration announced. (PJM)
Backlog: PJM is processing the remaining 67 GW of roughly 200 GW in its transition phase. (PJM)
Tapestry's claim: HyperQ pre-screened site control for 811 Cycle 1 applications. Tapestry's own figure; PJM's release doesn't give it. (Tapestry field notes)
Cost: not disclosed.

What PJM signed

The goal is to "significantly" cut processing times for new interconnection applications so large volumes can be handled quickly and accurately, PJM said. Tapestry will develop and deploy AI tools and models, powered by Google Cloud and Google DeepMind. (PJM)

"PJM is committed to bringing new generation onto the system as quickly and reliably as possible," said Aftab Khan, PJM's executive vice president for operations, planning and security. (PJM)

Tapestry describes itself as a "moonshot for the electrical grid" project. (PJM)

Tapestry general manager Page Crahan said the tools would help PJM "make faster decisions with greater confidence." Amanda Peterson Corio, Google's head of data center energy, called grid evolution "one of the greatest challenges of the AI era." (PJM)

The backlog

The work complements PJM's reformed interconnection process, which began in July 2023. PJM said it would process the remaining 67 GW of transition-phase projects in 2025 and 2026 and start its new Cycle process in early 2026. It has also announced a Reliability Resource Initiative to let selected transition-phase projects enter faster. (PJM)

Founded in 1927, PJM serves all or parts of Delaware, Illinois, Indiana, Kentucky, Maryland, Michigan, New Jersey, North Carolina, Ohio, Pennsylvania, Tennessee, Virginia, West Virginia and D.C. It runs 88,333 miles of transmission lines and says its grid and market operations save $3.2 billion to $4 billion a year. (PJM)

What Tapestry says happened

Tapestry says the reformed process adds compressed timelines and a strict "first-ready, first-served" framework that makes site control, proof that a developer controls the land, a primary gatekeeper. It says manual readiness reviews historically took multiple hours per application, citing its own customer research. (Tapestry field notes)

Tapestry's undated field notes say HyperQ ran 9,312 parallel application reviews and readiness assessments across 811 generation applications in the site-control portion of PJM's reformed Cycle 1, producing a readiness report with page-level citations for grid engineers, who make the final determinations. These are company claims. (Tapestry field notes)

Tapestry reports a median runtime of 6.25 minutes per application and a 95th-percentile time of 28 minutes, and claims "a 20x+ acceleration" against manual review. (Tapestry field notes)

Its footnote says the comparison uses a "conservative lower-bound manual anecdotal historical baseline" of 4 hours (240 minutes). Our check: 240 ÷ 6.25 = 38.4, so the stated "20x+" is below what its own inputs give. Tapestry says the runtimes exclude transmission lag, human validation and final regulatory determinations, and don't represent end-to-end interconnection time. (Tapestry field notes)

What went wrong

The results aren't PJM's. Every HyperQ figure comes from Tapestry's undated page; PJM's release gives none. (Tapestry field notes; PJM)
An anecdotal baseline. Tapestry calls its manual benchmark anecdotal and its comparison "directional." (Tapestry field notes)
Screening isn't connecting. HyperQ covers only the site-control readiness screen; engineers still make the calls. (Tapestry field notes)
What we don't know yet: whether PJM has measured any time savings itself, and the contract's cost.

Our follow-up question for PJM Interconnection

Does PJM confirm Tapestry's Cycle 1 figures, and has AI screening shortened its interconnection timelines?

Sources

PJM Interconnection, “PJM, Google & Tapestry Join Forces To Apply AI To Enhance Regional Planning, Generation Interconnection,” April 10, 2025. insidelines.pjm.com
Tapestry, “Field Notes: HyperQ deployment in PJM's Cycle 1” (undated; vendor). tapestryenergy.com

Disclosure: [not in copy] Editor placeholder