Value chain · CLM-0466
In plain terms
NASA's report prints the computer code for its cost-learning formula, and running that code gives back the report's own module and servicer costs. The one-line formula written in words just above the code is mistyped, so anyone working from the words rather than the code would get a badly wrong answer.
The claim as recorded
NASA OTPS prints its "Crawford model" learning-curve calculation as Python code: each unit costs the first-unit cost times the learning rate raised to log2 of the unit number, and the programme cost is the cumulative sum over units. Re-running that calculation reproduces the report's RD1 module manufacturing cost ($6.91 billion against the published $6.9B) and servicer cost ($11.4 billion against the published $10.9B), within 1 and 5 per cent. The report's prose statement of the same formula, printed one paragraph above the code, does not match the code: taken as printed it gives about $15 million for RD1's modules rather than $6.9 billion.
Scope and limits
The servicer figure is 4.6 per cent above the published value; the source of that gap is not identified. The code is printed for the module case; the servicer case is said to use the same model at an 85 per cent learning rate (PDF p.65), and the code for it is not printed separately. The prose-formula defect is the source's, recorded here so that the prose is not used in place of the code.
What it assumes
The first-unit costs and unit counts are the report's own inputs. The continuous Wright integral T(N) = C1 x N^(1+b)/(1+b), b = ln(learning rate)/ln(2), agrees with the code's discrete sum to within 0.02 per cent at the module unit count (1.463 million); at 17 servicer units the two forms differ by about 5 per cent.
Where it comes from
This is the project's own working, shown in full:
PDF p.62 (printed 45): "we developed a python script to apply the learning curve", with the prose formula quoted in the claim; PDF p.62-63 (printed 45-46): the code,learning_rate = 0.75,units_produced = list(range(1,6000001)),this_unit_cost = first_unit_cost * (learning_rate ** (np.log(units_produced) / np.log(2))),cumulative_cost = np.cumsum(this_unit_cost). PDF p.61 (printed 44): "Module Manufacturing Cost ($) $6.9B" RD1, "Crawford model learning curve applied to initial cost and number of units". PDF p.65 (printed 48): "Servicer Manufacturing Cost - Assemble ($) $10.9B" RD1, "Using the Crawford model, 85% learning curve". Re-run (lcoe-model.py, CALC-053): $6.91e9 over units 1 to 1,463,000; $1.14e10 over units 1 to 17. The $15 million is our arithmetic on the prose formula as printed (1,000,000 x 0.75 x log2(1,463,000) = about 15.4 million).
Basis: FY2022 USD, nominal
Related records
- Answers question
- QST-041 not published in this release
- Supersedes
- CLM-0406 not published in this release
Cite this
energyaperture.com/c/clm-0466
CLM-0466, Energy Aperture, as of 2026-09.
Connected records
Other published records linked to this one. A list, not a judgement.
- Also from SRC-012
- CLM-0006None of the five main studies mentions the international radio-frequency exposure limits, or any safety standard, or a restricted area...CLM-0123NASA's own 2024 study puts space solar power at 12 to 80 times the cost it expects in 2050 from power stations on the ground, a set it...CLM-0124NASA also found that if several things improve at once, rather than just one, space solar power does become cost competitive.CLM-0125On carbon, NASA found space solar power within the range of five renewable ways of making power on the ground (wind, solar, solar thermal...CLM-0144A NASA study found that launching everything into space is by far the biggest cost of a space solar power system, and that most of those...CLM-0192The full list of assumptions behind the headline result of NASA's 2024 study, that space solar power would cost far more than ground...CLM-0193To get to a competitive cost, NASA's own 2024 analysis needs launch costs to halve, solar cells to get much more efficient, satellites to...CLM-0240A from-scratch cost model of the "Innovative Heliostat Swarm" design gets almost the same electricity cost as NASA's own study: about 62...More on the source page.