PJM Interconnection received 811 generation projects totaling 220GW in the first cycle of its reformed interconnection process, managed by an agentic AI system from Google-backed Tapestry. The queue backlog dropped from 300GW to 170GW.
Key facts
- 811 generation projects totaling 220GW applied in first reformed cycle
- Queue backlog dropped from 300GW to 170GW under new process
- Agentic AI system from Google-backed Tapestry manages queue
- PJM covers 13 states and Washington D.C.
- 220GW is 1.8x PJM's current installed capacity of ~120GW
PJM Interconnection, the Regional Transmission Organization covering 13 states and Washington D.C., announced that 811 new generation projects with a combined capacity of 220GW applied to connect to the grid through the first cycle of its reformed interconnection process [According to Data Center Dynamics]. The reformed process will utilize an agentic AI system developed by Google-backed Tapestry to manage the queue, marking one of the largest real-world deployments of agentic AI in critical infrastructure.
Why this matters more than the press release suggests
The 220GW figure represents roughly 1.8 times the total installed generation capacity of PJM's entire current fleet (about 120GW). That the queue backlog has dropped from 300GW to 170GW under the new process suggests the AI system is enabling faster triage of viable projects vs. speculative ones. PJM's reformed process, known as the Interconnection Process Reform (IPR), shifts from a serial first-come-first-served model to a cluster-based approach where projects are studied in groups. The agentic AI system from Tapestry automates feasibility studies and queue management, reducing study times from years to months [per PJM documentation].
Scale and context
The 220GW figure includes a mix of solar, wind, battery storage, and natural gas projects. PJM has not disclosed the exact breakdown by technology type. Google's backing of Tapestry aligns with its broader push into energy infrastructure — Google announced a $5 billion Texas data center for Anthropic in April 2026 [as previously reported], and has been investing in grid interconnection solutions to support its growing data center footprint. The Tapestry AI system represents Google's first major foray into applying agentic AI to physical grid operations, distinct from its cloud-based AI services.
Agentic AI in critical infrastructure
The Tapestry deployment is one of the earliest examples of agentic AI systems being used for real-time management of physical infrastructure at scale. Unlike generative AI chatbots, agentic AI systems autonomously execute multi-step workflows — in this case, evaluating interconnection requests against grid capacity models, regulatory requirements, and queue priorities. The system's ability to process 811 projects simultaneously highlights a shift from human-in-the-loop to AI-driven decision-making in grid operations, a domain traditionally dominated by manual engineering reviews.
What to watch
Watch for PJM's second cycle results in Q3 2026 to see if the AI system maintains throughput. Also watch for Google's next data center announcement — the Texas $5B facility for Anthropic signals continued energy demand. Tapestry's agentic AI deployment could expand to other RTOs like MISO or SPP.
[Updated 01 May via gn_dc_power]
The Google-Anthropic deal now includes a 5 GW compute commitment, pre-selling AI capacity at unprecedented scale, according to Data Center Knowledge. This deepens the link between Google's Texas data center investment and the grid interconnection demands managed by PJM's Tapestry AI system.
