The grid interconnection queue is the binding constraint on the AI buildout — faster than capital, regulation, or fuel

Confidence: 8/10 Supporting evidence:

• 2,600 GW backlog, 5-year median wait — this is physical infrastructure time that cannot be compressed by money alone. Evidence: strong
• 80% of projects withdraw from the queue; even well-funded projects face uncertainty and cost. Evidence: strong
• Co-location (Amazon/Talen Susquehanna) is an attempted workaround; FERC outcome uncertain. Evidence: strong

Challenging evidence:

• PJM queue reform (April 2026) could meaningfully accelerate new interconnections to 1-2 years. Evidence: moderate — too early to verify
• HVDC transmission investment could expand capacity significantly by 2030. Evidence: moderate

Evolution:

• Jun 15, 2026 — Initial thesis at 8/10

Depends on: grid-interconnect-queue Would change if: FERC/PJM reforms demonstrably reduce interconnection time to <2 years at scale

---