Source Report

FERC Interconnection Queue Backlogs Severely Delay Grid-Connected Data Centers

PJM Interconnection's queue backlog—exacerbated by FERC Order 2023's shift to cluster studies—forces data centers into 4-5 year waits for grid access, as serial processing of 286 GW of queued generation (mostly renewables) clogs approvals; this creates a supply-demand mismatch where PJM utilities forecast 55 GW new large load by 2030 but only half the generation needed, driving capacity prices to FERC caps of $329/MW-day and shortfalls like the 6.6 GW miss in the 2027-2028 auction.[1][2][3]
- National queues hold ~2,300 GW across 10,300 projects as of end-2024 (down from 2,600 GW peak in 2023 due to withdrawals, not faster processing), with median time-to-operation at 5 years; PJM (286 GW), MISO (311 GW), CAISO (523 GW) worst hit, pausing new requests until 2026.[3][4]
- Data centers amplify this: PJM's 2025 capacity auction saw data centers drive 40% of $16.4B costs; ERCOT's large-load queue hit 233 GW (70% data centers) in 2024, up 300% YoY.[5]
For new entrants, this means prioritizing co-location (e.g., FERC's Dec 2025 PJM order for behind-the-meter rules) or off-grid sites in SPP/MISO, as grid-tied projects risk 5+ year delays costing $3.5B+ in PJM alone from stalled renewables.

Permitting Timelines for Behind-the-Meter Gas Favor Quick Deployment Over Grid Reliance

Behind-the-meter (BTM) gas generation permits via state processes (e.g., Virginia's Permit By Rule for <150 MW solar/gas hybrids) enable 18-24 month timelines versus 5-7 years for grid upgrades, as data centers self-supply via on-site turbines to bypass FERC queues; federal FAST-41 now covers >100 MW/$500M projects, compressing reviews to months, but air permits remain state hurdles (3-12 months).[6][7]
- Examples: Meta/Entergy's 2.3 GW Louisiana gas turbines (Dec 2024); Crusoe's 360 MW Texas turbines (Jan 2025); Sharon AI's 250 MW Texas with CCS (Jan 2025)—all BTM, avoiding queues.[8]
- PJM's Feb 2026 filing proposes 50 MW BTM threshold with 3-year transition; FERC's RM26-4 eyes 20 MW+ loads by Apr 2026, but states retain retail oversight per NARUC resolution.[9]
Competitors must target power-rich states (TX, GA) for BTM, as federal reforms accelerate but can't override local air/land use (e.g., NY/VA moratoriums), risking 1-2 year delays without site control.

Diverse On-Site Generation Sources Scale Rapidly for Data Centers, Challenging Patel's View

Dylan Patel (SemiAnalysis) asserts "scaling power in the US will not be a problem" via hyperscaler CapEx (~$1T in 2026) and BTM diversity, but deployments show reciprocating engines (INNIO's 2.3 GW VoltaGrid), aeroderivatives (Boom Supersonic/Crusoe's 29 turbines for 1.2 GW Abilene), Bloom fuel cells (1.5 GW installed, 2 GW production by 2026), and ship engines scaling to GWs—yet grid queues (2,300 GW backlog) and PJM's 6 GW shortfall contradict, as 30% of planned 56 GW BTM is 2025-announced.[10][11][12]
- Recips: Caterpillar's 4 GW Utah + 1 GW Hunt; INNIO's 92x25 MW packs.[11]
- Aeroderivs: Crusoe's jet-based for Stargate (1.2 GW).[13]
- Bloom: $5B Brookfield deal, AEP's 1 GW; 400 MW to data centers now.[14]
Patel's optimism holds for BTM (1/3 of sites fully off-grid by 2030 per Bloom survey), but new entrants face equipment backlogs (gas turbines 3-4 years) and $350M/site costs, favoring incumbents with scale.

Utility and Grid Operator Reports Highlight Near-Term Bottlenecks

PJM's 2026 Load Forecast warns of 5.3% annual growth (data centers 94% of it), with 6.6 GW 2027 shortfall and $333/MW-day prices; MISO delays 2022-2025 queues; ERCOT's 233 GW large-load surge prompts PUC rules—contrasting Patel, as NERC's 2024 LTRA flags data centers exceeding 120% growth by 2027 in key footprints.[15][4]
- PA PUC (Mar 2025) probes data center grid impacts; NARUC urges FERC preserve state retail jurisdiction amid RM26-4.[16]
- Berkeley Lab: Queues cost $7B in PJM auction alone.[2]
Entrants should hedge with hybrids (e.g., SPP's HILLGA for co-located), as pure-grid risks blackouts (e.g., VA's 1.5 GW data center trip).

Analyst Assessments Split: Short-Term Crunch vs. Long-Term Resolution

Goldman Sachs (Feb 2026) raises DC demand 220% by 2030 (709 TWh), but E3 warns over-forecasts risk stranded assets; McKinsey sees 3x capacity by 2030 amid constraints, while Patel bets on $1T CapEx—utilities like PJM/MISO report 2027-2028 shortfalls, but BTM scales (46 projects/56 GW).[17][18]
- Bearish: Grid Strategies: Delays cost billions; Rystad: 100 GW pipeline vs. 15 GW deficit.[19]
- Bullish: Bloom: 1/3 off-grid by 2030; SemiAnalysis: BTM race wins.[20]
For competition, bet on BTM/gas in TX/SPP (short queues) over PJM/CAISO; Patel's view fits post-2028 if reforms land, but 2026-2027 crunch demands off-grid now (high confidence on queues, medium on BTM scale-up).

---

Recent Findings Supplement (March 2026)

Interconnection Queue Backlogs Persist Despite Reforms, Challenging Data Center Timelines

Lawrence Berkeley National Laboratory's Queued Up 2025 edition reveals US interconnection queues shrank to 2,300 GW total capacity (1,400 GW generation + 890 GW storage across ~10,300 projects) by end-2024—the first decline after years of growth—but this stemmed more from record 700+ GW withdrawals (77% of entrants fail) and fewer new requests than efficiency gains; median time-to-operation hit 4+ years for recent projects, exposing data centers to 3-7 year delays as FERC Order 2023 cluster studies and readiness rules take hold unevenly.[1][2]
- Queues exceed US installed capacity (~1,300 GW); natural gas queued capacity surged 72% YoY to 136 GW amid data center demand, while solar/storage dropped 12-13%.
- PJM (largest RTO, 55 GW data center load forecast by 2030) processed 170 GW requests since 2023 but faces backlogs tying project schedules to broader queues.[3]
For data center entrants, this means hedging via multi-queue filings or off-grid power, as grid-tied approvals lag AI deployment cycles by years—non-obvious risk: withdrawals signal speculative overload, inflating studies for viable projects.

FERC and Grid Operators Accelerate Large Load Rules Amid PJM Shortfalls

PJM's Jan 16, 2026 Board letter—post-FERC Dec 2025 order finding its tariff "unjust"—proposes "Bring Your Own New Generation" (BYONG) and Expedited Interconnection Track (EIT) for ≥250 MW resources paired with data centers (≥50 MW loads), slashing timelines to 10 months via higher deposits ($10-20k/MW), state permitting pledges, and 100% upgrade cost responsibility; caps 10/year, targets Aug 2026 rollout with FERC filing, plus backstop procurement for reliability gaps (e.g., 2027/28 auction 5.6% short).[4][5]
- DOE Oct 2025 Section 403 letter spurred RM26-4 rulemaking for ≥20 MW loads (data centers routinely 200-300+ MW), emphasizing cost causation; FERC Dec order mandates PJM clarify co-location terms by mid-2026.
- MISO's ERAS fast-tracks gas-heavy projects (75% of requests); PJM forecasts 82 GW peak rise over 15 years, data centers 40% of growth.[6]
Entrants gain faster paths if self-supplying generation, but states/utilities push back on federal overreach—implication: hybrid co-location becomes moat, shifting costs from ratepayers ($9-23B PJM hit).[7]

Behind-the-Meter Gas Deployment Surges, Bypassing Grid Bottlenecks

Data centers announced 56 GW across 46 behind-the-meter (BTM) projects (90% post-2025), with 16-30 month timelines vs. 3-7 year grid waits; Cleanview analysis confirms permits/equipment orders (GE Vernova, Caterpillar) for many under construction, enabling islanded ops via reciprocating engines (24-30 mo), aeroderivatives (18 mo), and fuel cells (16-18 mo).[8]
- Wärtsilä Jan 2026: 429 MW (24x 50SG engines) for US data center plant; INNIO/VoltaGrid: 2.3 GW (92x 25 MW packs); Caterpillar: 4 GW Utah campus + Monarch.
- Aeroderivatives (e.g., OpenAI Stargate's 29x GE LM2500XPRESS, ~1 GW) and ship-retrofitted turbines (ProEnergy/Mitsubishi, 48 MW/unit) deploy in Texas for fast-start (10 min).[9]
BTM circumvents queues via state permitting (e.g., Texas SB6 cost-sharing), but lacks utility commission timelines—new: EPA Dec 2025 hub speeds air permits, allowing site work pre-approval (months saved); scales to GW via modular stacking. New entrants must front $2-5M/MW capex, favoring hyperscalers.

Diverse Generation Sources Scale Rapidly for BTM, Validating Patel's Optimism

Patel-esque bullishness (e.g., Dylan Patel/SemiAnalysis: power solvable via speed-onsite) holds as Bloom fuel cells hit GW deals—AEP $2.65B/1 GW (exercised Jan 2026), Brookfield $5B global "AI factories"; Bloom 2026 report: 1/3 data centers fully off-grid by 2030 (up 22% forecast), 50%+ campuses >500 MW by 2035.[10][11]
- Recips/aeroderivs: Wärtsilä/Jenbacher for Meta/Stargate (onsite gas microgrids); efficiency edge (Bloom 60-65%, recips heat-tolerant).
- Vs. reports: PJM data centers drove $9.3B auction spike (63% price rise), but BTM offsets (e.g., Nebius 85% self-gen); ERCOT/Texas lead with 40 GW DC forecast.[12]
Patel's view proven: diversity de-risks via BTM (Texas/Wyoming fast-paths), but ratepayer pushback (e.g., $16-37/mo hikes) demands BYOG mandates—competitors without GW-scale financing lag.

Policy Shifts Enable BTM but Spark Ratepayer Backlash

Jan 2026 White House/13 governors principles + PJM tariff overhaul enforce data center upgrade payments, deterring speculation; states (VA moratorium to 2028?, GA/OK bans) + FERC RM26-4 target April 2026 rules.[13]
- Utility filings: PJM IMM attributes $21B capacity costs to unbuilt DC forecasts; Grid Strategies: 166 GW peak growth by 2030 (55% data centers, possibly overstated 25 GW).
- Announcements: DATA Act (Feb 2026) exempts off-grid DCs from FERC; EPA hub accelerates BTM air permits.[14]
Reforms support Patel by prioritizing BTM/BYONG, but challenge via cost allocation fights—new players need state buy-in, risking delays in constrained hubs like PJM/CAISO. Confidence high on data (LBL/FERC primary), medium on state filings (sparse post-9/13).