Does building a base on the Moon increase the total mass throughput humanity can deliver to elsewhere in the solar system — and at cosmic scales (Tt/yr), what constraints bind first on each side?

← Sources · Report home
epa-o2-supply-chain-2023

EPA Oxygen Supply Chain Profile

EPA Supply Chain Profile team 2023 report cited by: q1-earth-industrial-ceiling
https://www.epa.gov/system/files/documents/2023-03/Oxygen%20Supply%20Chain%20Profile.pdf

Source review

Source Review: EPA Oxygen Supply Chain Profile (March 2023)

Summary

Verdict Count
Consistent 1
Different conclusion 0
Novel supporting 0
Merits investigation 1
Not relevant 0

Claim 1: U.S. industrial oxygen production ~10.3 Mt/yr (2019 data)

Quote (from extract.md): "Total US O2 production: ~10.3 Mt/year" — most recent comprehensive EPA figure, 2019 data. Cited in q1 calc as: assumption #5 (US LOX denominator) and supports claims q1.c1, q1.c7. Verdict: Consistent — but with retrieval caveat. Why: The 10.3 Mt/yr figure is industry-consensus across multiple secondary citations (workspace SDC anchor, EIA-adjacent reports, market research like Mordor/Future Market Insights). The Air Liquide Baytown announcement (3.29 Mt/yr from one new facility in Texas) is internally consistent: scaling one TX facility's capacity at ~32% of current total US capacity matches a $10.3 Mt/yr-class national baseline.

Claim 2 (merits investigation): Industrial O₂ vs merchant LOX vs rocket-grade LOX

Quote: Not directly extracted because PDF fetch failed; this is a Codex-audit-flagged caveat. Verdict: Merits investigation Why: The EPA figure is total industrial oxygen production, not the rocket-relevant subset (merchant liquid oxygen, with storage, transport, and load-rate capacity). The actual rocket-LOX bottleneck is somewhere below the bulk-O₂ ceiling but not quantified in this retrieval. Future re-pass should fetch the EPA HTML or use USGS / Industrial Gas Association breakdown.

Methodological flag

The PDF returned binary streams via WebFetch during pass-01-research. The 10.3 Mt/yr figure rests on the workspace SDC anchor's prior compilation citing the EPA, not a fresh primary retrieval. Tier S claim citation is methodologically weaker than intended. Re-pass should attempt the HTML or alternative-primary-source fetch.

Extract

Abstract

The EPA's March 2023 Oxygen Supply Chain Profile is the canonical U.S. government characterisation of the industrial oxygen sector. The report's load-bearing figure for our analysis (as cited in Workshop/Space/sdc/starship-launch-cadence-constraints.md and corroborated by industrial-gas trade press) places total U.S. oxygen production at approximately 10.3 million tonnes per year (most recent comprehensive figure, 2019 data). Major U.S. producers identified are Air Products, Air Liquide, Linde, and Matheson Tri-Gas. Direct retrieval of the PDF via WebFetch failed (the PDF text streams returned as compressed binary not decompressed during fetch) — the figure is recorded here via the workspace's prior compilation. A pass-through fetch on a different format (HTML rendering of the report's summary tables, or the EPA Oxygen Profile webpage) should be attempted in a future re-pass.

Key claims

  • us-o2-production-2019: "Total US O2 production: ~10.3 Mt/year" (10,335 million kg) — most recent comprehensive EPA figure, 2019 data.
  • major-us-o2-producers: Air Products, Air Liquide, Linde, Matheson Tri-Gas (per EPA + industrial-gas trade press).
  • fetch-issue: The EPA PDF returned uncoded streams via WebFetch; cited via secondary anchor and to be re-fetched on next iteration.

Reviewer notes

Tier S — government primary source. However: direct WebFetch of the PDF failed; the 10.3 Mt/yr figure is sourced via the workspace SDC anchor's prior citation, not via fresh retrieval. This is a methodological flag for the audit pass: a tier-S claim should rest on the primary document, not a secondary internal compilation. The reconcile pass should either (a) fetch the EPA HTML version, or (b) downgrade the claim to tier-A-by-cross-reference if the PDF cannot be retrieved. The figure is industry-consensus (multiple secondary sources converge on 10.3 Mt/yr) so confidence is medium-high even with the retrieval gap.