# External Sources Validation ## Claims Sources Table | Claim | Source | URL | Date accessed | Exact or approximated | Validation Status | How used on page | |-------|--------|-----|---|---|---|---| | 2,300 GW in US interconnection queues | LBNL Queued Up 2025 | https://emp.lbl.gov/queues | Mar 2026 | Approximated (~2,290 GW) | ✅ Confirmed | Queue capacity snapshot | | ~13% completion rate (2000–2019 capacity) | LBNL Queued Up 2025 Edition | https://emp.lbl.gov/queues | Mar 2026 | Exact | ✅ Confirmed | 13% of capacity submitted 2000–2019 reached commercial operation by end-2024. Standardized to 13% per LBNL Queued Up 2025 Edition. | | 4+ year median queue time | LBNL | https://emp.lbl.gov/queues | Mar 2026 | Exact | ✅ Confirmed | Queue delay baseline | | 10,300 projects in queue | LBNL | https://emp.lbl.gov/queues | Mar 2026 | Exact | ✅ Confirmed | Queue size metric | | 408 GW with draft/executed IA | LBNL | https://emp.lbl.gov/queues | Mar 2026 | Stated | ✅ Confirmed | LBNL Queued Up 2025 Edition explicitly states 408 GW | | 62 GW US data center demand 2025 | S&P Global 451 Research | N/A (proprietary; subscription required) | Mar 2026 | Approximated (61.8 GW) | ⚠️ Proprietary | Current DC demand | | 134 GW by 2030 | S&P Global 451 Research | N/A (proprietary; subscription required) | Mar 2026 | Approximated (134.4 GW) | ⚠️ Proprietary | DC demand projection | | $10-12B annual revenue per GW | SemiAnalysis | N/A (proprietary; subscription required) | Mar 2026 | Stated | ⚠️ Proprietary estimate — order of magnitude validated against public announcements, not independently verified | Revenue-per-capacity metric | | 415 TWh global DC electricity 2024 | IEA Energy and AI 2025 | https://www.iea.org/reports/energy-and-ai | Mar 2026 | Exact | ✅ Confirmed | Global DC baseline | | 945 TWh base case by 2030 | IEA Energy and AI 2025 | https://www.iea.org/reports/energy-and-ai | Mar 2026 | Exact | ✅ Confirmed | Base scenario projection | | 1,700 TWh lift-off by 2035 | IEA Energy and AI 2025 | https://www.iea.org/reports/energy-and-ai | Mar 2026 | Exact | ✅ Confirmed | Accelerated scenario projection | | 700 TWh headwinds by 2035 | IEA Energy and AI 2025 | https://www.iea.org/reports/energy-and-ai | Mar 2026 | Approximated | ✅ Confirmed | Constrained scenario projection | | 325-580 TWh US DC by 2028 | Berkeley Lab 2025 (Queued Up 2025 Edition) | https://emp.lbl.gov/publications/queued-2025-edition-characteristics | Mar 2026 | Approximated | ✅ Confirmed | US DC projection range | | 175 GW additional transmission capacity from AI | IEA Energy and AI 2025 | https://www.iea.org/reports/energy-and-ai | Mar 2026 | Stated | ✅ Confirmed | IEA Energy and AI report states 175 GW transmission-capacity unlock directly (Energy Markets & Planning chapter) | | $110B/yr grid optimization savings | IEA Energy and AI 2025 | https://www.iea.org/reports/energy-and-ai | Mar 2026 | Stated | ✅ Confirmed | IEA Energy and AI report states $110B/yr in electricity-system savings directly (Energy Markets & Planning chapter) | | 1.7 workers nearing retirement per young entrant | IEA World Energy Employment 2025 | https://www.iea.org/reports/world-energy-employment-2025 | Mar 2026 | Exact | ✅ Confirmed | Workforce demographics | | OpenAI 2.3 GW onsite gas Texas | SemiAnalysis | N/A (proprietary; subscription required) | Mar 2026 | Stated | ⚠️ Proprietary estimate — corroborated by public reporting | AI facility power draw | | xAI 1.9 GW from Doosan | SemiAnalysis | N/A (proprietary; subscription required) | Mar 2026 | Stated | ⚠️ Proprietary estimate — corroborated by public reporting | AI facility power draw | | 30% of DC sites onsite power by 2030 | Bloom Energy | N/A (company market research) | Mar 2026 | Stated | ⚠️ Proprietary | Onsite power adoption | | Microsoft/TMI PPA | Public reporting | N/A (multiple press reports) | Mar 2026 | Exact | ⚠️ Not linkable | Corporate PPA example | | Google/Kairos Power SMRs | Public reporting | N/A (multiple press reports) | Mar 2026 | Exact | ⚠️ Not linkable | SMR deployment example | | $3.3T annual global energy investment | IEA World Energy Investment 2025 | https://www.iea.org/reports/world-energy-investment-2025 | Mar 2026 | Exact | ✅ Confirmed | Total energy investment baseline | | 60-70% of queue time is document work | Sunya estimate | N/A (internal analysis) | Mar 2026 | Modeled | 🔒 Internal | FERC procedure analysis | | Software 3.0 / centaur model | Andrej Karpathy | YC AI Startup School (Jun 2025), No Priors podcast (Jan 2026) | Apr 2026 | Stated | ✅ Confirmed | Conceptual framework: LLMs as new computing paradigm; centaur = human judgment + AI speed; "remove yourself as the bottleneck" | ## Completion Rate Discrepancy **Status: Resolved** — Page standardized to 13% per LBNL Queued Up 2025 Edition (capacity submitted 2000–2019, commercial operation by end-2024). **LBNL data:** - LBNL reports **13% completion rate** when measured by capacity (MW completed / MW queued) for the 2000–2019 cohort - LBNL reports **19% completion rate** when measured by project count (# completed / # queued) - These figures apply to the 2000–2019 cohort (projects entering the queue in that period) - Later cohorts show lower completion rates due to longer development cycles **Resolution:** Page standardized to **~13%** to match LBNL Queued Up 2025 Edition standard capacity-based metric. The figure applies to 2000–2019 cohort (capacity submitted 2000–2019, reached commercial operation by end-2024). Updated 2026-04. --- ## Validation Key - **✅ Confirmed:** Source data accessed and matches claim within normal rounding; public URL provided. - **⚠️ Proprietary:** Source is paywalled or proprietary research (SemiAnalysis, 451 Research, Bloom Energy); not publicly verifiable. - **⚠️ Not linkable:** Source is podcast/interview/news aggregate; no single URL available. - **❓ Unconfirmed:** Claim not directly validated in accessible source materials; flagged for follow-up or cited as modeled. - **🔒 Internal:** Claim is derived from Energy Decision Stack internal analysis—not from external source. --- ## Notes 1. **LBNL Queued Up 2025** is the primary interconnection queue database. All interconnection-related claims should reference this source when possible. 2. **IEA Energy and AI 2025** is the primary source for global and regional AI electricity scenarios. 3. **S&P Global 451 Research** is the primary source for data center capacity and demand projections. 4. **SemiAnalysis** provides proprietary facility-level power and revenue estimates; verify against public announcements when available. 5. **Bloom Energy, Microsoft, Google public announcements** should be cited by press release date or company announcement, not generic "public reporting." 6. The **60-70% document work figure** is modeled from FERC interconnection procedure timelines and should remain labeled as such.