AI Meets the Grid: Interconnection Queue Analysis in PJM and ERCOT

Summary

Research report from Carbon Direct (published May 2026; page metadata dated 2026-07-15) analyzing the generation-and-storage interconnection queues in the two US markets most exposed to data-center load growth, PJM and ERCOT. It fills the interconnect-queue-wait-time gap flagged in June and still open at the 2026-07-22 compile — the power thesis previously had no quantified queue. analysis-tier (named research firm with a methodology, not a filing or regulator).

Key figures (as-of report publication, May 2026)

• Interconnection wait time in data-center load zones: 3 to 4 years — the headline number the power thesis was missing.
• Combined queue: >300 GW of generation + storage across ~1,500 projects (PJM + ERCOT, as of the May 2026 report).
• Queue composition: renewables + storage = 77% of PJM's active queue, 87% of ERCOT's queue.
• Natural gas moves faster: gas projects clear interconnection "substantially more quickly" than renewables; post-One-Big-Beautiful-Bill-Act, monthly gas entries to ERCOT's queue rose ~150%, with renewable withdrawals following in both markets.
• Capacity-on-paper vs deliverability: the queued generation is "enough capacity on paper to meet projected data center demand through 2030" — i.e., the shortfall is queue throughput / deliverability, not a raw megawatt shortage.

Context from the same search (NOT from this report — treat as unverified corroboration)

Other 2026 sources surfaced alongside this report (not ingested) put harder single-market numbers on the queue: PJM's first reformed "Cycle 1" (May 2026) drew ~220 GW across 811 projects; ERCOT's large-load queue was ~410 GW as of April 2026, ~87% data centers; PJM's Dec-2025 capacity auction came up ~6,625 MW short. These are directionally consistent with Carbon Direct but from lower-authority outlets — flagged as context, not ingested facts.

On-thesis relevance

Directly quantifies the binding constraint the KB has called "power." The reframing that matters: the bottleneck is queue throughput (3-4 yr waits) and interconnection deliverability, not an absolute generation shortage (the queue holds enough MW on paper through 2030). That distinction sharpens frontier debate #5 ("who owns the power") — behind-the-meter / on-site generation (Tesla Cortex, gas) is attractive precisely because it skips the 3-4 year grid queue, which explains the ERCOT gas-entry surge. It also gives the "36+ projects / $162B stalled" claim in the existing power source a mechanism.

Provenance / caveats

• Analysis-tier: a research-firm report, not a FERC/ISO filing. The >300 GW / ~1,500-project figure is the report's own combined active-queue count as of May 2026; single-market GW figures vary by outlet and by what's counted (generation queue vs large-load queue vs a specific reform cycle).
• "3 to 4 years" is for data-center load zones specifically; some outlets cite 5+ years for the broadest interpretation.

Source: AI Meets the Grid: Interconnection Queue Analysis in PJM and ERCOT, Carbon Direct, May 2026.