Prologis Ventures → Confidential
Confidential

Restricted access

This document is confidential. Enter the access password to continue.

Prologis Ventures → Confidential

Hyperscaler firm baseload power — deliverability matrix

Ten firm baseload technologies scored across permitting, economic, and stakeholder dimensions, with seven regional markets, anchored to disclosed deals. Click any card to expand. Click numbered superscripts to jump to sources.

Grid benchmark = Data Center All-In Delivered Cost. Not retail industrial rate. For new GW-scale loads in major markets, all-in delivered runs $110-150/MWh: retail industrial ($70-95) + substation upgrade amortization ($8-15) + transmission spur ($3-8) + capacity market charges ($5-20) + demand charges934. This is the realistic alternative against which PPA deals (TMI/Crane ~$98-115, Talen-AWS ~$69) should be measured.

The delta, decomposed — why a PPA price isn't a utility rate

If a utility quotes power at X and a hyperscaler signs a PPA at Y, both are $/MWh — but they are not the same product. The gap Y − X is not a markup. It is the sum of everything the retail rate leaves out, plus the contract terms a hyperscaler is actually buying.

Y PPA price − X utility retail rate = deliverability stack +$30 to +$55 + firm-power · speed · tenor premium − scale · existing-asset · avoided-charge discount
The short version. The delta only looks like a premium because X is usually quoted as the retail industrial rate — the price of plugging into capacity that already exists. A new gigawatt does not plug into capacity that already exists. Re-baseline X to the all-in delivered cost and most of the delta disappears: the disclosed nuclear PPAs price at, or below, the true grid alternative.

Step 1 — from the retail rate to the real grid cost

The retail industrial rate assumes the grid can already serve the load. Connecting a GW-scale load triggers four cost layers the sticker price excludes. Together they are the deliverability stack, and they convert the headline rate into the cost a hyperscaler would actually pay to take grid power. Bars show midpoints of the ranges in the benchmark callout above934.

Utility retail industrial rate (X, as quoted)
$80
+ Interconnection & substation upgrades
+$12
+ Transmission spur
+$6
+ Capacity market charges
+$13
+ Demand charges
+$19
= Data center all-in delivered cost (the real X)
≈ $130
Talen–AWS Susquehanna PPA (Y)
~$69
Constellation TMI / Crane PPA (Y)
~$98–115
$0$40$80$120$160

Both disclosed PPA bars (blue) end left of the all-in delivered total (amber) — the hyperscaler is buying firm power below its true grid alternative, not above it.

Step 2 — from the grid cost to the PPA

Against the all-in figure, where Y lands depends on what kind of power the contract delivers. Some terms justify a premium; others pull the price down. The net of the two is the second half of the delta.

Premium — what a PPA buys that the grid doesn't (pushes Y up)

  • Firm & 24/7 carbon-free: a grid blend is neither firm nor clean; a nuclear PPA is both1.
  • Speed-to-power: a contracted asset sidesteps interconnection queues of 5–7+ years in PJM14.
  • Price certainty: a 15–20-year fixed price vs. a merchant grid rate that escalates with the stack above13.
  • Scarcity at signing: signing into the oversubscribed 2027 pipeline costs more than signing into a balanced market33.

Discount — what pulls Y below all-in grid cost

  • Scale & counterparty: GW-scale offtake from one investment-grade buyer is bankable and cheap to finance3.
  • Existing-asset economics: restarts and uprates skip new-build capex — the reason Talen–AWS clears near $694.
  • Avoided grid charges: a behind-the-meter PPA skips the capacity and demand charges from Step 1 entirely — it removes the layers, rather than paying them.

The delta, worked through

The same PPA looks expensive or cheap depending only on which X it is measured against. That choice — not the contract — is what creates the apparent delta.

Disclosed PPA (Y)PriceΔ vs. retail rate (X ≈ $80)Δ vs. all-in delivered (X ≈ $130)
Talen–AWS Susquehanna34 ~$69/MWh −$11 — roughly at par −$61 — deep discount
Constellation TMI / Crane1 ~$98–115/MWh +$18 to +$35 — looks like a premium −$15 to −$32 — still a discount

Estimates are analyst-derived, not disclosed contract prices14 — directionally robust, not precise to the dollar. Bottom line: Y − X is large and positive only when X is the retail sticker. Re-baseline X to the all-in delivered cost and the delta collapses to roughly zero or negative.

PPA pricing drivers — supply versus demand

PPA prices move with the balance between firm baseload supply and hyperscaler demand. Both supply easing and demand softening are deflationary; both supply scarcity and demand strength are inflationary.

Inflationary pressure (prices rise)

  • Supply scarcity: Transformer lead times 24-30 mo (pre-2020) → 5 yrs current34. Gas turbine OEMs (GE Vernova, Siemens, MHI) booked out — capex ~195% above 2019 by YE 20262134. Interconnection queues 5-7+ yrs in PJM14.
  • Demand strength: ~$650B hyperscaler AI capex 2025-202634. Sightline tracks 190 GW across 777 projects since 2024; 31.2 GW announced for 202733.

Deflationary pressure (prices fall)

  • Supply abundance: VoltaGrid-Oracle 2.3 GW Stargate + Vantage 1 GW + Caterpillar-Joule 4 GW Utah171832. Mainspring $258M Series F scaling LGen production36. GE Vernova 29 aero turbines = ~1 GW for Stargate17. SMR FOAK deliverable 203023. Fervo EGS drilling cost down 49% in 18 mo10.
  • Demand softening: Sightline forecasts 30-50% of 2026 pipeline slips33; 26% of 2025 already slipped. GPU depreciation / writedown risk if the cycle compresses34.
Pipeline context — Sightline Climate Q1 2026 Data Center Outlook33 click to expand chart▾

Data center capacity by year and project status, from Sightline Climate's Q1 2026 Outlook33 (Feb 24, 2026; 190 GW / 777 projects >50 MW). The 2027 bulge — 31.2 GW announced, only ~6 GW under construction — is a roughly 5x oversubscribed pipeline. That gap is the current source of inflationary pressure on PPAs.

PPA WTP & BTM LCOE, 2018-2035

Three scenarios labeled by driver. PPA anchored to: TMI/Crane1, Susquehanna3, Hyperion5, Stargate17, Google-Kairos23, xAI Colossus28. Grid line from EIA9 + capacity market data. Tech LCOE from Lazard8 + NREL ATB19.

Click any legend item to isolate that curve. Click again to restore.

Deliverability scorecards — 10 firm baseload technologies

Each card scores 10 dimensions in three groups: Permitting, Economic, Stakeholder. Striped bars indicate config-dependent spread. Click any card to expand regional drill-downs.

Reading the bars. Solid bars = base/typical config. Striped bars = high-end with added config (SCR for RICE/aero, H2 for SOFC/LGen, order-book maturation for SMR, restart vs new for nuclear). Manufacturing capacity = can OEMs build enough units; Site scalability = can units be sequenced to power a multi-GW campus — two different things. Political bar: fills from the center — blue left = Democratic-leaning, red right = Republican-leaning, both sides = bipartisan.

Modifiers (not standalone baseload)

Technologies that don't deliver 24/7 firm power alone but modify the economics of paired baseload techs. Click to expand.

Sources 36 cited · 22 directly verified

✓ = URL fetched and/or content corroborated across multiple independent outlets during this session. • = canonical institutional URL (government / established data provider), high-confidence but not re-fetched this session.