Americas power and energy transition cost metrics: Bernstein argues that the cheapest nominal MWh is not necessarily the lowest-cost reliable power solution

Overview

This Bernstein note explains why levelized-cost measures should not be used alone to compare power-generation and storage technologies. It argues that a power system increasingly shaped by large-load growth, congestion, interconnection constraints and tighter reserve margins requires evaluation of firm, deliverable power at a specified node and date.

Core views

Bernstein’s central argument is that LCOE is a useful but incomplete measure. LCOE divides discounted lifetime costs by discounted lifetime electricity output, incorporating capital expenditure, operating and maintenance costs, fuel and financing assumptions. It therefore answers the narrow question of the average lifetime cost to build and operate an asset per MWh. It does not establish the time or location of delivery, availability during system stress, reliability contribution, or what the asset earns. LCOS similarly measures discounted lifetime storage costs per MWh discharged, including charging energy, losses, cycle life, degradation and augmentation, but does not value the flexibility for which storage is deployed. The omission matters more as power systems face large-load growth, interconnection queues, transmission congestion, higher renewable penetration and tighter reserve margins. Bernstein argues that the relevant question for a 24/7 large load is the cost of firm delivered power at a named node by a named date, rather than the lowest average MWh cost. Intermittent generation is not a like-for-like substitute for dispatchable supply: storage, transmission or backup capacity may be required to provide equivalent reliability, and these firming and integration costs are absent from standalone LCOE. Storage economics are also not fully represented by LCOS because merchant batteries can capture energy arbitrage, capacity payments and ancillary-service revenues by adapting charge and discharge schedules to market conditions. The report details several other limitations. LCOE treats all MWh as equal, even though midday solar generation in CAISO can coincide with negative wholesale prices; additional solar can reduce the value of existing and incremental solar output through value deflation or self-cannibalization. Published values can also overstate delivered output when curtailment or incomplete interconnection lowers realized capacity factors. Rankings are sensitive to capital structure and discount-rate assumptions: in Bernstein’s Lazard-based model, utility-scale solar has LCOE of about $38/MWh at 60% debt financing, while reducing debt to 20% raises its LCOE to a level broadly comparable with a new-build CCGT. LCOE also averages costs across a 20–30-year life, concealing differences in leverage, cash-flow timing and financing risk, and excludes carbon, land-use and water-use externalities unless explicitly modeled. Bernstein proposes a scoreboard rather than a single replacement metric. It includes hybrid LCOE, which adds storage to intermittent generation; VALCOE, which adjusts cost for energy, capacity and flexibility value; system LCOE, which includes balancing, backup and grid-reinforcement costs; ELCC-adjusted capacity value; and, where warranted, full dispatch modeling. ELCC translates nameplate MW into dependable MW and is the appropriate denominator for capacity-cost comparisons. In an illustrative gas CCGT-versus-utility-scale-solar comparison, standalone LCOE is $51/MWh for gas and $40/MWh for solar. The comparison shifts to $51/$64 on hybrid LCOE, $49/$53 on VALCOE, $51/$55 on system LCOE, and $53/$77 on ELCC-adjusted capacity value. Using PJM final 2026/2027 ELCC class ratings and an illustrative $150/kW-year Net CONE, solar carries roughly a $37/MWh firming adder versus less than $5/MWh for CCGT. The practical conclusion is not that LCOE should be discarded. Bernstein views it as a screening tool and a valid way to track cost trends within the same technology over time. For cross-technology decisions, however, the report advises checking embedded capital-structure and discount-rate assumptions and adding measures of capacity, grid integration, market value, flexibility and delivery constraints. For US data-center build-outs, it highlights MW per acre, noise, emissions, air pollutants and water use as additional determinants of technology choice.

Analysis framework

Bernstein starts by defining LCOE and LCOS and identifying their cost inputs. It then tests the metrics against real-system requirements—timing, location, reliability, grid integration, financing and market value—and uses a gas CCGT versus utility-scale solar example to show how adjusted measures can alter apparent cost rankings. The note uses Lazard’s LCOE+ framework as a foundation, supplemented with PJM capacity assumptions and Bernstein analysis.

Methodology notes

• Other

  Levelized-cost and system-value comparison using LCOE, LCOS, hybrid LCOE, VALCOE, system LCOE and ELCC-adjusted capacity value

  The report compares lifetime unit costs with measures that add storage, grid, capacity and system-value effects, showing why a low standalone generation cost may not equal low-cost reliable delivered power.

Key data

• Utility-scale solar LCOE under 60% debt financingapproximately $38/MWhBernstein’s Lazard-assumption-based model; a lower 20% debt share raises the cost to broadly comparable with new-build CCGT.
• Illustrative standalone LCOEGas CCGT $51/MWh; utility-scale solar $40/MWhLow-case LCOE from Lazard’s 2026 report.
• Illustrative hybrid LCOEGas CCGT $51/MWh; utility-scale solar $64/MWhAdds storage costs to solar.
• Illustrative VALCOEGas CCGT $49/MWh; utility-scale solar $53/MWhUses value factors of 1.05x for CCGT and 0.75x for solar.
• Illustrative system LCOEGas CCGT $51/MWh; utility-scale solar $55/MWhIncludes moderate-penetration balancing, grid-reinforcement and profile costs.
• Illustrative ELCC-adjusted capacity valueGas CCGT $53/MWh; utility-scale solar $77/MWhUses firming costs and PJM ELCC assumptions.
• Solar firming adderroughly $37/MWhBased on PJM final 2026/2027 ELCC class ratings and an illustrative $150/kW-year Net CONE; CCGT is below $5/MWh.

Impact & implications

The report says technology selection and power-project underwriting should focus on reliable, deliverable power and investable returns, not just average generation cost. This is especially relevant where capacity needs, congestion, large-load demand or renewable penetration make timing and system integration economically material.

Risks

• Standalone LCOE can misstate economic competitiveness by excluding timing, curtailment, firming, grid-integration and capacity costs.
• Technology rankings can change materially with discount-rate and debt-versus-equity assumptions, particularly for capital-intensive assets.
• Average lifetime costs can obscure project cash-flow timing, leverage and financing risk.
• LCOS can understate the economic value of flexible storage in volatile markets by averaging away arbitrage, capacity and ancillary-service opportunities.

What to watch

• Capital structure and discount-rate assumptions embedded in any LCOE comparison.
• ELCC ratings, capacity-market assumptions and the cost of firming intermittent supply.
• Curtailment, interconnection status, transmission constraints and realized capacity factors.
• Technology-specific delivery constraints for data-center demand, including MW per acre, noise, emissions, air pollutants and water use.