Corridor
LNG terminal scale
AI campus scale
National stress event
Assumptions (drag to test)
The LNG terminal — from fuel delivery to flexibility asset
The AI data center — national capability that survives shocks
What a single uncontrolled event puts at stake
Foreign investment this backbone unlocks
Direct jobs anchored
Grid-partner posture
About this model — what is verified, what is assumed, what it deliberately does not do
What this is
An illustrative concept prototype of the National AI Energy Backbone — a sovereign decision-support tool, not a built product, project engineering, an OEM guarantee, or a representation of any company's actual plans or endorsement. It exists to make the trade-offs of an integrated LNG → Power → AI design visible and adjustable for Vietnamese decision-makers. The AI campus is modeled on the public Meta "Hyperion" reference (announced scale and dedicated generation), transplanted to a Vietnamese tropical-and-grid setting; nothing here asserts knowledge of Meta's internal plans.
Verified public context for Vietnam [DESK, verified Jun 2026]
- Revised PDP8 targets up to ~22,400–22,524 MW of LNG-fired power by 2030 (~12.3% of capacity), with import infrastructure matched to power plants (Vietnam-Briefing / Watson Farley & Williams, 2025).
- Nhơn Trạch 3 & 4 (1,624 MW), Vietnam's first LNG power complex, entered commercial operation 1 Jan 2026; Thị Vải terminal at 1 MTPA (expandable to 3), Cai Mép at ~3 MTPA from 2025 (expandable to 6) (VnEconomy, Petrovietnam, 2025–2026).
- Decision 2161/QĐ-TTg (29 Sep 2025) opened 100% foreign investment in data centers and AI and targets attracting global hyperscalers; data-localization rules from 1 Jan 2026; DC pipeline already >500 MW, heading toward ~1 GW by 2030 (Hogan Lovells, CBRE, 2025–2026).
- The 2023–24 heatwave-and-drought blackouts hit northern industrial zones with rolling outages and an ~20% demand spike from cooling, raising foreign-investor concern; demand grows >8%/yr (Bloomberg, France24, 2023).
Key assumptions [EST / DOCTRINE / DESK — every one adjustable or documented]
- LNG cold energy: ~830 kJ/kg theoretical recoverable on regasification → ~231 GWh-thermal/yr (~26 MW-thermal average) per MTPA. The recovery-efficiency slider sets the usable share; the default is deliberately conservative. Singapore (SLNG / NUS / Keppel) is the clearest public proof-of-value, not proof that every terminal can cool a campus (ScienceDirect, SLNG).
- Turbine protection: hot tropical inlet air derates gas turbines; the literature reports >10% augmentation from inlet-air chilling. Modeled derate 10–18% of plant output, protected at the validated ambient design point — not a guarantee of nameplate in any weather.
- AI campus: cooling load ≈ IT electrical load at first order; GPU density 375–545 per MW (GB200-class), rack value $2–3M, compute value $3–10.50/GPU-hour (default $7). Recovery tail anchored to the public Google London 2022 incident (18.4 h restoration; 35.25 h long tail).
- Chiller displacement 0.171–0.227 MW-electric per MW-thermal, less a 7.5% auxiliary allowance. FDI ($25–40B/GW) and jobs (8,000–15,000 construction / 800–1,500 operations direct per GW) are macro planning figures from the AI-infrastructure brief, scaled linearly — illustrative, not project forecasts.
Mechanism honesty
- LNG cold energy shown is theoretical gross; usable recovery depends on terminal design, heat-exchanger approach, distance, and load coincidence — approved with caveat / needs project validation. Available only where regasification sits near the cooling load.
- The thermal reserve stores and discharges cooling, not electricity. "Demand handed back" is displaced chiller load — demand reduction, plus permitted plant export — never battery export to the grid.
- The reserve provides cooling autonomy (2.5–4 h), not compute runtime; continuity follows from avoided forced shutdown. Stored cooling is finite — beyond the window the campus goes through an orderly, checkpointed shutdown, not an endless ride-through.
Deliberately excluded
LNG procurement and pricing, terminal and plant engineering, dispatch economics, PPA / tariff design, financing structure, subsea-cable and fiber connectivity, water, CO₂ accounting, and construction schedule. US tax mechanics (§48E ITC) are intentionally omitted — they are irrelevant to a sovereign decision; value here is framed as national capability, investment absorption, and resilience. A production version would add project-specific engineering in each excluded lane.
Reproducibility
Spun out from the Richland Parish reference build (segi_continuity_simulation.html, simulation_plan.md); Vietnam context in the SEGI knowledge base (vietnam_ai_opportunity.md, lng_cold_energy_ai_backbone.md). Self-contained and works offline.
UTILITY