CALC-057

Break-even search, and the floor that survives free launch

Feeds CLM-0256 and CLM-0468. SHARED - two claims, one section. Answers QST-004.

Script. lcoe-model.py SECTION 8, section_breakeven() (line 834), with find_breakeven_launch_cost() (794, a 60-iteration bisection) and companion_package() (817).

Why it was needed. A one-lever-at-a-time sweep cannot answer "what would it actually take". Two questions need a search rather than a sweep: what launch price reaches terrestrial parity when the other favourable changes are granted, and what floor remains when launch is free.

Inputs. CALC-053's and CALC-054's, plus the companion package, which is fixed while launch cost is solved for:

Working, transcribed.

-- Break-even launch cost to reach 0.05 $/kWh, companion package held (CLM-0256) --
20-yr hardware life, 0.70 module learning rate:  $27.4M/launch  =  $233/kg to LEO
15-yr hardware life, 0.70 module learning rate:  $14.1M/launch  =  $119/kg to LEO
10-yr hardware life, any tested learning rate :  NOT REACHABLE

  [the $/kg follows CALC-053's route: $27.4M x (1-0.15) / 100,000 kg = $233/kg]

-- Floor with launch cost driven to $0/launch (CLM-0468) --
across hardware lifetime 10-30 yr and module learning rate 0.65-0.75:
                                    0.0969 to 0.1722 $/kWh
-- Floor of that floor, companion package also applied --
                                    0.0410 to 0.0641 $/kWh  [full 12-cell grid; the script prints only the 5-cell subset, 0.051 to 0.064]

Result. Two floors, and neither reaches the comparator. Zero launch cost - a physical impossibility, tested only to establish a bound - still leaves RD1 at roughly 0.10-0.17 $/kWh, because a manufacturing-cost floor remains in every case. Adding every other favourable assumption simultaneously reaches 0.0410-0.0641 $/kWh across the full twelve-cell grid, with SEVEN of twelve cells BELOW 0.05 (corrected 2026-09-06; the earlier 0.051-0.064 was the five-cell subset the script prints, being the cells where its break-even search fails).

Conclusion. Independently confirms, via a floor-search method different from the report's own one-at-a-time sensitivity method, that launch cost alone cannot reach terrestrial-competitive LCOE. Broadly consistent with the report's own "Competitive" combination of five simultaneous changes reaching 0.03-0.08 $/kWh (their Table 26): the same qualitative structure, multiple correlated improvements needed, no single lever sufficient.

Where this fails. A break-even point under a specific, favourable, assumed bundle of simultaneous improvements, not a forecast that the bundle will occur. The "not reachable" cells - 10-year lifetime at any tested learning rate - show the result is highly sensitive to which improvements are bundled.

ONE REPAIRED DEFECT IN CLM-0468.

1. The claim's headline originally said the zero-launch floor was 0.05-0.17 $/kWh while its own support field carried 0.0969-0.1722. The true minimum is 0.0969, which rounds to 0.10. The error made the floor look lower, and therefore SBSP look cheaper, than the model supports. 2. The claim said the companion-package floor could go below 0.05 $/kWh. A first check read only the 5-cell subset the script prints (0.051 to 0.064) and called that wrong. The full 12-cell grid gives 0.0410 to 0.0641, with seven cells below 0.05, so the claim was right on this point (corrected 2026-09-06).

The first is corrected in the transcription above.

Cite this