3 private positions, held through vehicles whose terms bar naming the company — recused
Commitment
sha256 fc5792fdae164aadbd7280cacf15d828808eb91fe2b593f426bcc1ec2025a593 · positions record version 2026-10-09
Intent, 90d none
How positions are declared, why some are not named, what a recusal means, and how the sealed record works: disclosure. Where a position I may not name sits in a company this issue would otherwise forecast, I publish no forecast on it and say so.
The contract
I am Kevin Neuburger. I invest in space, defense, energy and manufacturing, and I advise companies deciding what these economic sectors mean for them. This is a brief, every two weeks, on four sectors — space, defense, energy and manufacturing — that the same capital, the same founders and the same cost curves now connect. It opens with space, because that is where the evidence is deepest. It exists because most coverage of these sectors is a relay: a company announces, the announcement is reproduced with adjectives, and nobody returns to it. There is no scoreboard, so being wrong costs nothing — and analysis that costs nothing when it is wrong is worth what you pay for it.
One rule governs this one. Every forecast here gets a probability, a resolution date, and a falsifier — the specific observation that would prove it wrong. On the date, it is graded in public, right or wrong. The record is public and cannot be edited after the fact; how that is enforced is in the register's rules.
I am not neutral and this is not journalism. I am an analyst with money at risk, writing down what I think will happen, on the record, with dates.
The call
2026-09-003 · Space
SpaceX lists a per-kg Starship price to Earth orbit on spacex.com before 2028
19% · Monte CarloResolves 2028-01-31
2026-10-23 Shown as 20% from 23 Oct 2026, when every entry moved to the nearest 5%. W1 published it as 19%; the stored number, 0.19, is unchanged and is the one graded.
Disclosure: I hold SpaceX.
Claim, falsifier, and reasoning
ClaimBefore 2028-01-01, SpaceX lists on its own website (spacex.com, including the rideshare page) a Starship launch price per kilogram to Earth orbit, or a Starship rideshare price with a per-kg rate to Earth orbit.
FalsifierAs of 2028-01-01, spacex.com (including the rideshare page) lists no Starship launch price per kilogram to Earth orbit and no Starship rideshare price with a per-kg rate to Earth orbit, assessed against the live site and Internet Archive snapshots of it. Not counted: prices to the Moon, Mars or other destinations beyond Earth orbit (including the Mars cargo rate); per-kg figures computed from a launch price and a payload mass; customer regulatory filings; executive statements or posts; cost targets; press reports; private contract prices.
Monte Carlo · inputsFour judged factors in the model P(YES) = R × K × L + (1 − R × K × L) × O, drawn as correlated Beta distributions: Starship credible enough to sell 2029 seats (R), rideshare moved to Starship (K), a per-kg price listed in time (L), any other route (O). The reasoning, the sources and the runnable model are published with the entry.
How I got 19%
ModelP(YES) = R × K × L + (1 − R × K × L) × O
R = 0.45By mid-to-late 2027, Starship is credible enough that SpaceX will sell 2029 seats to outside customers.Evidence. First orbital flight on 28 Sep 2026 (Flight 14; 26 Starlink V3 satellites deployed, early return after an engine shutdown). A ship catch attempt is planned for Flight 15, not yet flown. Musk's condition for winding down Falcon is Starship "flying reliably several times per week" (X post of 22 Aug 2026), far above the current few flights a year. My judgement: his timelines have historically slipped two to three times.
K = 0.50SpaceX keeps rideshare and moves it to Starship, rather than dropping it or leaving it on the few remaining Falcon 9 missions.Evidence. SpaceX reportedly is not booking Falcon 9 beyond 2028 (single-source report, Sep 2026). Industry reports say it may stop all but a few Falcon 9 missions from 2028. Rideshare is small next to Starlink, but SpaceX dominates it.
L = 0.65Given R and K, SpaceX lists a per-kg price on spacex.com before 2028-01-01.Evidence. SpaceX already publishes Falcon 9 rideshare pricing: $350,000 for up to 50 kg to sun-synchronous orbit, then $7,000/kg for 2026 (SpaceX's own price feed, read 11 Oct 2026). The feed's ladder runs to 2027 and does not mention Starship. Timing risk: rideshare sells 12 to 24 months ahead, so a listing for a late-2029 flight could land in early 2028, after the deadline.
O = 0.04Any other route: an early listing for marketing, or a response to competitors.Evidence. Seven years of Starship development with no listing so far.
CorrelationThe factors move together: if Starship works, the rest become more likely. A Monte Carlo with a moderate correlation (0.3 between R and K; 0.15 between each of them and L) gives a mean of 0.187, so 19%. 80% range: 8% to 31%.
SensitivityEach factor moved 0.15 up and down, the others fixed: R 13% to 23%; K 14% to 22%; L 15% to 21%.
Signals to watch
Up
A ship catch on Flight 15.
A reflight of a recovered ship.
The first Starship launch from LC-39A.
Falcon 9 Transporter missions disappearing from the 2029-and-later rideshare manifest.
A 2028 or later row in SpaceX's rideshare price feed that names Starship.
Any Starship option appearing on spacex.com/rideshare. That resolves the question yes.
Down
Repeated Starship failures or a long stand-down.
SpaceX confirming that Transporter missions continue on Falcon 9 into 2029 and later.
A 2028 row in the price feed that stays on Falcon 9.
Rideshare quietly discontinued.
Not counted as evidenceA $90M price for a dedicated Starship launch on a 2029 mission appeared in a customer's regulatory filing (reported Mar 2026). It is a customer filing, per launch and not per kilogram, and not on spacex.com, so it does not resolve this question.
For an investor, this decides whether Starship capacity is something you can underwrite at a price, or something that stays captive to its owner and reaches you only as a competitor's cost advantage — and whether any plan priced off cheap heavy lift should be underwritten at today's marks at all. For a corporate strategy team, it decides whether the next decade of launch is an input you can budget for, or a dependency you can only watch.
The argument
Launch cost binds whoever buys launch retail. It does not bind the operator that owns the rocket.
That is the whole argument. Four pieces of evidence carry it, and the first is the one that makes the rest legible: the price of a launch and its cost are different numbers, and almost every claim about cheap access to space quietly swaps one for the other.
One. Price and cost are different numbers, and one launch operator publishes both per vehicle. Rocket Lab discloses price and cost for Electron in its SEC filings. In FY2025 it realised $8.5M of revenue per launch against $4.8M of cost — a 44% gap, same vehicle, same year, both audited. In FY2023 the two figures were $7.1M and $7.0M, nearly identical. The spread is not a constant; it is a policy, and it widened as the vehicle matured.
On Electron's 300 kg capacity that is $28,333/kg charged against $16,000/kg incurred. Hold that ratio in mind, because Rocket Lab is the only launch operator that publishes both sides per vehicle. For everyone else, the number in circulation is whichever side of that gap the speaker finds convenient — and for a vertically integrated operator launching its own payloads, the gap is not an accounting curiosity. It is the business model.
Two. The cheapest published number is going up. SpaceX's rideshare programme — the actual route to orbit for most small satellites — is priced at $7,000/kg for 2026 in SpaceX's own live pricing feed, on a published ladder rising from $5,500/kg in 2023 to $7,500/kg in 2027. That is 2.6× the dedicated Falcon 9 headline figure, and it has risen for four consecutive years. The operators most often cited as the beneficiaries of cheap launch pay the most per kilogram, and more each year.
Nor does the dedicated figure mean what it appears to. Published $/kg is a price divided by a maximum payload, and almost nothing fills the rocket. The headline number is a capacity-normalised abstraction, not a price anyone is charged. A customer buying a dedicated Falcon 9 for a six-tonne satellite pays roughly $10,300/kg on the 2018 list price of $62M, because the unused capacity is not refunded.
Put those two together and the direction of travel is the opposite of the story. The route most operators actually use has risen 36% in four published years. The route they are quoted is a number nobody pays. And the vehicle that is supposed to collapse both has no published price at all.
None of this says launch is expensive, or that it has not fallen. It says the fall is captured, and the published numbers do not measure it.
Three. Starship has no published price per kilogram to Earth orbit. SpaceX's Starship page states a capability — more than 100 tonnes to orbit, fully reusable — and one published rate: $100,000 per kilogram to the Martian surface, for cargo flights no earlier than 2028. For LEO it says only that marginal cost per launch is designed to be lower than Falcon's. No number. The widely circulated targets of $10 or $100 per kilogram could not be traced to a verifiable primary source, and I am not going to print a figure I cannot source.
That matters because the entire downstream literature is built on it. The $200/kg threshold everyone quotes has a traceable origin: Google's 2025 orbital compute paper, which argues that at $200/kg launch amortised over spacecraft life becomes roughly comparable to a terrestrial data centre's energy cost per kilowatt. Google puts its own baseline at $3,600/kg and estimates that reaching $200 needs roughly 180 Starship flights a year on a 20% learning curve, arriving around 2035. Against three Starship flights in the last twelve months, only one of which, Flight 14 on 28 September, reached orbit, that is a flight rate 60 times the current one. Every business plan that becomes viable at $200/kg rests on a number the vendor does not publish.
The cheaper tier is real; the only launch operator that publishes both sides per vehicle shows a 44% gap between what it charges and what it spends. But that tier belongs to whoever owns the rocket. Everyone else buys at the published number, which is where the thesis comes from: the cheap capacity exists and is not for sale.
Vehicle
Basis
$/kg to LEO
Source
Falcon 9
list price, SpaceX payload basis
$2,719
NASA, 2018, from $62M ÷ 22,800 kg
Falcon 9
list price, FAA payload basis
$4,654
FAA, from $61.2M ÷ 13,150 kg
Falcon 9 Transporter
rideshare list price, 2026
$7,000
SpaceX pricing feed
Electron
realised price ÷ rated 300 kg, FY2025
$28,333
Rocket Lab 10-K
Electron
internal cost, FY2025
$16,000
Rocket Lab 10-K
Starship
—
no published LEO figure
SpaceX
Two of those rows are the same vehicle at nearly the same price. SpaceX publishes 22,800 kg to LEO for Falcon 9; the FAA's own compendium states 13,150 kg. On 2018 prices of $62M and $61.2M, that single disagreement produces $2,719/kg against $4,654/kg — a 1.7× spread with the price effectively held constant.
Much of this arithmetic is a fight about the denominator rather than the numerator. Capacity figures assume a low circular orbit at a favourable inclination; a real mission to a real orbit gets less. When someone quotes you a dollars-per-kilogram figure, the first question is not what they paid. It is what they divided by, and whether anyone has ever actually flown that.
Four. Kuiper is the control experiment. Amazon committed over $10 billion to launch across four providers — including ten Falcon 9 flights bought from its direct competitor in January 2026. It still launched 396 satellites against an FCC milestone of 1,618 by 30 July 2026, missing by more than four to one, and the FCC waived the deadline and temporarily reduced the priority of satellites launched after it.
Read the behaviour rather than the statements. A firm that commits $10 billion, buys across four suppliers, and takes flights from the competitor it is trying to displace is not price-constrained. It is quantity-constrained, and its deadline came from a spectrum regulator. Its penalty was paid in spectral priority — temporarily, and only on the satellites launched after the deadline — not in dollars per kilogram.
Now the other side of the same trade, from the operator that owns the rocket. Since SpaceX listed in June 2026 its economics are audited rather than inferred, and the number that should end the argument is this: SpaceX agreed to pay $19.6 billion for 65 MHz of EchoStar spectrum — the licences moved into a trust in May 2026 and transfer to SpaceX at a closing that is pending, with payments to EchoStar's lenders running through at least November 2027 — while the gross book value of every Starlink satellite ever built and launched — gross capitalised satellite property, which by SpaceX's own policy includes capitalised launch — was $13.8 billion. Spectrum cost 1.4 times the entire constellation. And on 8 October 2026 it agreed to buy up to 14 MHz of 800 MHz spectrum from Grain Management, subject to FCC approval — more radio licences, not more rockets.
If launch were the master variable for the largest constellation ever flown, its owner would not be spending more on radio licences than on the hardware. Launch binds Kuiper and not Starlink for one reason, and it is not efficiency: Kuiper buys launch retail.
The honest summary is narrower than the usual claim. Launch cost is the master variable for businesses that buy launch retail and whose payload mass is large relative to everything else they spend. That set is real: Kuiper, lunar landers, anyone deploying tonnage they cannot build a rocket for. It excludes the operator that owns the rocket, which is why Starlink works. And it excludes most of what is pitched as launch-gated, where naming the price per kilogram conveniently makes the binding problem someone else's.
What would change my mind
A published Starship price per kilogram to Earth orbit. There is none today. A published price resolves the entry above true, and a company that publishes a price is a company competing on price — that would change this issue's framing, not the estimate on the entry, and it is the single observation I am watching for. Continued silence through 2027, while the rideshare ladder keeps climbing, is evidence for the opposite reading: that the cheap capacity is for Starlink and the rest of us are the residual market. The 2028 rideshare figure is the cheap early test, and it takes a minute to check — the ladder has risen every year from $5,500/kg to $7,500/kg, so anything below that would be the first pass-through of falling cost to customers anyone has seen, and I would revise this issue's framing rather than defend it.
What would damage the argument in a way I have not priced: a Chinese provider publishing a genuinely low price with flights behind it. My case assumes cheap capacity stays captive because its owner has no reason to sell it, and that holds only while there is one supplier.
Close
A Spanish edition ships with every issue — same numbers, same table, same register entry. Calibrated analysis of these sectors is scarce in Spanish.
The register is public and append-only — every claim above stays in it with its original wording and date, and is graded on the day it resolves whether or not that goes well: reckoning-register.
Next issue in two weeks: W2 — What the Space Economy Actually Is, and Where the Cash Is.
Sources
Every figure above traces to a line in the sources for this issue, with publisher and access date. Where two credible sources disagree — as they do on Falcon 9's payload capacity — both are listed and the disagreement is stated rather than averaged away.
One limitation, stated here rather than only in the appendix: SpaceX's own site is client-side rendered and returned no text to automated retrieval; the two SpaceX primary sources used here are the rideshare pricing feed and the Starship vehicle page.
Each issue in your inbox the morning it publishes, a note the day a forecast is graded, and occasional Reckoning news. Free, in English and Spanish. No list sharing, unsubscribe in one click. Privacy.