The AI Transition Tracker

Is the West on the recovery path or sliding toward decline? Three hard numbers, measured against the J-Curve whitepaper’s required paths — every value sourced, every estimate flagged.

Read the whitepaperOpen the ledger

Welcome to July 8, 2026: our energy gap is 111 TWh/yr (≈ 2.0 NYCs).

U.S. Data Center Energy Demand vs. New Generation Supply (2024–2035). Computed live from the ledger every day — data-center demand and new-generation supply are each a probability-weighted sum (MW × delivery probability) over the tracked projects, converted to TWh/yr with sourced capacity factors per modality (see the J-Curve whitepaper’s methodology). Published demand projections from EPRI, Goldman Sachs, US DOE, McKinsey, and BCG are the thin dotted lines inside the shaded min–max envelope (each named at its right edge on wide screens); the pink fill is the supply–demand gap, with the two red callouts sized live from the current ledger. Past 2030 (the shaded “extrapolation” region), announced projects thin out, so the raw ledger would otherwise flatten into an artifact of visibility, not a forecast — instead the lines there are extrapolated: demand at the median growth rate of the published projections, supply at the paper’s sourced later-decade addition rate, both anchored to the ledger’s last value. Situational Awareness’s AI-compute path is omitted here: it is on an absolute basis, not the new-since-baseline scale every other series shares. The right axis converts to NYC-equivalents (1 NYC = 56 TWh/yr).Source: Demand & Supply Workbook — live ledger mirror of the J-Curve whitepaper’s Figure 1; underlying rows in the ledger. Updates daily as the owner approves ledger proposals.

The three numbers

Energy growth

US new firm-equivalent generation supply added

18TWh/yrest

0.25× required pace — 18 vs 73.1 TWh/yr required now

Required path — US new firm-equivalent generation supply added

J-Curve whitepaper §1.2 (papers/jcurve-whitepaper.md): 'Net new firm-equivalent supply from the entire US generation portfolio is modest—roughly 18 TWh added in 2025, rising to 150–250 TWh per year later in the decade'. Band 150–250 TWh/yr. Firm-equivalence per EIA Electric Power Monthly table 6.07.B.

as of Dec 31, 2025 · J-Curve whitepaper §1.2 (firm-equivalent additions), EIA Electric Power Monthly, EIA-860 generator inventory

Compute growth

US data-center electricity demand

320TWh/yrest

0.61× required pace — 320 vs 521 TWh/yr required now

Required path — US data-center electricity demand

J-Curve whitepaper §1.2 and Figure 1 (papers/jcurve-whitepaper.md): bottom-up pipeline ≈175 TWh in 2024 (1,521 tracked facilities @ 0.85 CF); ~1,000 TWh by 2030 (industry projections span 290–1,050 TWh). 2035: ΔP ≈ 1,329 TWh/yr incremental powered AI energy (§3.3, §5.1).

as of Jun 16, 2026 · LBNL 2024 U.S. Data Center Energy Usage Report, IEA Electricity 2024 (data centres)

Value-to-society growth

AI economic value per kWh of compute

0.65$/kWhest

0.80× required pace — 0.65 vs 0.81 $/kWh required now

Required path — AI economic value per kWh of compute

J-Curve whitepaper §5.1 (papers/jcurve-whitepaper.md): q₀ ≈ $0.50/kWh in 2025 (band $0.3–0.8; frontier-lab revenue ÷ ~275 TWh AI electricity) = analysis/config.AI_Q0_USD_PER_KWH; trend-restoring q₂₀₃₅ ≈ $12/kWh at g_Q ≈ 37%/yr; g_Q band 25–45%/yr = analysis/config.AI_PRODUCTIVITY_GROWTH_BAND.

as of Jun 16, 2026 · J-Curve whitepaper §5.1 (q derivation), Frontier-lab revenue (OpenAI + Anthropic run-rates)

Signals

Good news

Bad news

Snapshot as of Jul 8, 2026 · generated Jul 8, 2026.