SPACEX LAUNCH GROWTH — 3 OCTOBER 2026

Open SpaceX_Launch_Growth.html in a browser. It works offline and has a
2027–2030 year slider. Both panels show historical observations, a smooth
exponential fit, and its forward continuation. The figures are annual rates,
not cumulative totals.

SCOPE AND THE 705 COUNT

The requested website is https://www.spacelaunchschedule.com/.
The SpaceX agency page's downloaded headline said 715 successful flights.
Downloading all 833 SpaceX archive pages returned 718 marked Success,
11 Failure, 4 Partial Failure, 96 Date/Time TBD, 3 Go For Launch and 1 In-Flight.
These page totals are inconsistent, apparently because of caching/update lag.
The original page/version showing 705 was not supplied and cannot be recovered
with certainty. No records have been invented or reclassified to force a match.

The requested 705-row file is explicitly a chronological snapshot: the FIRST
705 flights currently marked Success, from 28 September 2008 through
21 August 2026 (15:14:22 UTC). The chart uses that snapshot. The separate
718-row file retains all successful entries in the downloaded archive, through
1 October 2026. It is not claimed to be a fully current inventory of every
completed SpaceX launch: the schedule still marks Transporter-18 In-Flight and
NROL-97 Go For Launch, while GCAT records completed flights. They are excluded
from the source-defined success selection and retained in the status file.

Success follows the website's classification. Its successes include short
Starship atmospheric hops and the Dragon in-flight abort test. Partial failures
such as CRS-1 are not silently promoted to full success. Landing failure and
payload launch failure are different outcomes; the former alone does not
exclude an otherwise successful launch.

FILES

SpaceX_Launch_Growth.html: interactive graph, offline, with expandable methods.
SpaceX_First_705_Successful_Launches.csv: one row per selected flight, its UTC
  date, estimated payload mass, source URLs, catalogue match and notes.
SpaceX_All_718_Archived_Successes.csv: all currently archived Success records.
SpaceX_Archive_Statuses.csv: factual extract for all 833 archive records.
SpaceX_Annual_705.csv / SpaceX_Annual_718.csv: annual aggregates, with 2026
  explicitly marked incomplete.
SpaceX_Exponential_Projections.csv: unrounded curve values, 2026–2030.
SpaceX_Fit_Window_Sensitivity.csv: alternative fitting periods.
analysis.json: machine-readable aggregates, coefficients and projections.
verification.json: graph/slider checks at desktop and phone widths.
source/: reproducibility scripts and chart source.
raw/: downloaded GCAT data and the factual Space Launch Schedule extract.

MASS DEFINITION

Metric tonnes = kilograms / 1,000. The metric is estimated payload carried to
space, not the launch vehicle's wet mass, maximum advertised carrying capacity,
or mass still in orbit today. Satellites, Dragon spacecraft and their cargo,
and demonstration mass simulators are included. Rocket structure and propulsion
are excluded. Test hops below space altitude contribute zero payload tonnes.
Near-orbital/suborbital Starship cargo is included, so this is not strictly
"tonnes delivered into durable orbit".

For Falcon missions, the figure is GCAT's OrbPay field, including the catalogue's
spacecraft and cargo conventions. GCAT warns that OrbPay is autogenerated and
some special cases may be imperfect. These are catalogue figures, often estimates,
not verified mission-by-mission weighings. Classified and rideshare payloads
can be particularly uncertain. No missing payload was replaced by rocket capacity.

For Starship, GCAT's launch-table OrbPay is 120 tonnes on several tests: that is
the vehicle itself, not its carried cargo. The analysis instead sums separate
payload objects in auxcat and satcat100k, excluding the Starship object:
  Flights 3–6: no separately carried payload -> 0 tonnes.
  Flights 10 and 11: 8 simulators x estimated 700 kg -> 5.6 tonnes each.
  Flight 12: 22 objects x estimated 700 kg -> 15.4 tonnes.
  Flight 13: 20 Starlink objects x estimated 2,000 kg -> 40 tonnes.
  Flight 14: 26 Starlink objects x estimated 2,000 kg -> 52 tonnes
    (only in the 718-flight file, after the 705 cutoff).
GCAT flags these individual satellite/simulator masses as estimates. Flight 12
is especially uncertain: Next Spaceflight reports 37.5 tonnes, versus GCAT's
15.4; this analysis consistently retains the primary GCAT estimate and flags
the conflict. Neither version has been promoted to a precise measured mass.
Reference for that discrepancy: https://nextspaceflight.com/launches/details/8002/

JOIN AND CHECKS

Launch timestamps come from each schedule page's Event startDate, not its
WordPress publication date (which can differ by years). Payload catalogue
records were matched within rocket family and within one hour of the timestamp,
then checked for one-to-one launch tags. Most differences are seconds. The
largest is USSF-259: the schedule says 01:07:56 UTC, GCAT 01:29:18 UTC on
17 September 2026. Both name the same mission; both times are retained.
Every selected flight has a mass value or an explicitly documented zero.
No timestamp, URL or matched launch tag is duplicated in the success selection.
The 705-flight snapshot sums to about 8,675.951 estimated tonnes. The 718-flight
extract sums to about 8,897.559 estimated tonnes. Display rounding does not
affect the fits or CSV values.

EXPONENTIAL MODEL

Each metric is fitted separately to nine complete years, 2017–2025:
    log(y) = a + b * (year - 2017)
    y = exp(a + b * (year - 2017))
Ordinary least squares is applied in log space; years are equally weighted.
This minimises squared log errors, not squared errors in tonnes or launches.
No retransformation/smearing correction is applied: the curve is the direct
exponential trend, not an estimated arithmetic conditional mean.

Launch count: a=2.529775094663778, b=0.317135483618764
  Implied annual growth: 37.3189%; log-space R-squared: 0.917965.
Payload: a=4.465641535287408, b=0.414233737675521
  Implied annual growth: 51.3211%; log-space R-squared: 0.956734.

The window starts in 2017 to focus on regular reusable operations rather than
the sparse early years. All earlier years are still shown as observations.
2011, a true zero-launch year, remains in the historical data; it is outside
the fitting window. 2026 is incomplete and shown separately as a hollow diamond.
The 705 and 718 extracts have identical complete-year fitting data, so the
chosen fits are the same for both.

YEAR       LAUNCHES/YEAR       ESTIMATED TONNES/YEAR
2027       299                5,475
2028       411                8,285
2029       564               12,537
2030       775               18,971

INTERPRETATION

These are mechanical continuations of historical growth, not forecasts of
achievable launch capacity. They assume growth keeps compounding and do not
model Starship ramp-up, launchpad limits, demand, reliability or regulation.
The 2026 pace is BELOW the historical exponential curve: through the 705 cutoff,
101 flights imply roughly 158/year and 1,415.955 tonnes imply about 2,212/year.
The fitted 2026 curve is 218 launches and 3,618 tonnes. The scenario therefore
requires acceleration from the current pace, not simply continuation of 2026.

Fit-window sensitivity is material. A 2020–2025 fit implies roughly 1,316 launches
and 22,801 tonnes for 2030. The 2015–2025 fit implies 834 and 20,312. These are
alternative trend windows, not confidence bounds. No spurious precision is
intended by the raw CSV decimals. A statistical confidence ribbon would not
capture uncertainty about capacity, technology, future demand or source masses.

SOURCES AND CREDIT

Space Launch Schedule:
https://www.spacelaunchschedule.com/agency/spacex-spx/
https://www.spacelaunchschedule.com/wp-json/wp/v2/launch-json?categories=126&per_page=100
Mission-specific URLs appear in the CSV. Extracted factual records only are
redistributed; complete schedule webpage HTML is not included.

GCAT data from Jonathan C. McDowell, General Catalog of Artificial Space Objects,
Release 1.8.8, https://planet4589.org/space/gcat/ (data update 2 October 2026).
GCAT is released under the Creative Commons Attribution licence (CC BY).
Original catalogue data are included with attribution; derived estimates,
filtering and Starship mass treatment are described above.
https://planet4589.org/space/gcat/tsv/launch/launch.tsv
https://planet4589.org/space/gcat/tsv/derived/launchlog.tsv
https://planet4589.org/space/gcat/tsv/cat/auxcat.tsv
https://planet4589.org/space/gcat/tsv/cat/satcat100k.tsv
Definitions: https://planet4589.org/space/gcat/web/launch/lcols.html
Object mass flags: https://planet4589.org/space/gcat/web/cat/cols.html

Chart uses D3.js v7.9.0, Copyright 2010–2023 Mike Bostock (ISC licence), embedded
in the standalone HTML so the graph does not need a network connection.

REPRODUCING THE NUMBERS

With Python and NumPy installed, run: python source/analyse.py
It reads the included raw/ files and writes output/ aggregates, fits and CSVs.
The downloader script records the public schedule archive; refreshing it later
can change source classifications, counts and dates. Preserve this dated snapshot
for reproducible comparisons. The chart fragment includes its own frozen data.
