Space · Wednesday, 26 August 2026

01 · Briefing · what happened

The most reliable rocket ever built just flew its last Starlink mission from Florida, on a booster's 37th flight

Space 1 min 31 sources

Falcon 9 set a reuse record and closed out an era in the same launch. SpaceX is retiring it from its busiest job while it is winning, and spending $100bn on the thing that replaces it.

flights by one booster

a company record, set on the final Florida Starlink mission [1][2]

260

Cape Starlink missions

flown by Falcon over seven years, ending here [1]

$100bn

stated Louisiana investment

against $15bn spent on all of Starship so far [4]

2029

earliest launches there

with construction starting in 2027 [4]

At a glance

On 25 August a Falcon 9 lifted 29 Starlink satellites from Cape Canaveral before dawn, throwing a jellyfish plume over New York City. [1]
It was the last Falcon 9 Starlink mission that will ever fly from Florida. [1][2]
The same flight was the rocket's 100th of 2026 and the 37th launch of that particular booster - a new reuse record. [1][3]
In seven years Falcon has flown roughly 260 Starlink missions from the Cape. That job is now over. [1]
From here, Starlink out of Florida flies on Starship - a vehicle that has never yet carried an operational Starlink batch. [1]
The Cape workforce has already been moved onto Starship preparations, which is why the change is happening now rather than when Starship is ready. [2]
Falcon is not being retired. Vandenberg in California keeps launching Starlink, and has already flown 44 of this year's 76 Starlink missions. [2][7]
The wider record is not a rocket in trouble: that booster landing was the 653rd Falcon booster landing, and the 220th on that one droneship. [3]
The day before, SpaceX announced Starbase Louisiana - 125,000 acres in Vermilion Parish, ten Starship pads, and a stated $100bn of investment. [4][5]
Construction begins in 2027 and the first launches are no earlier than 2029, so the replacement is four years out from the site meant to carry it. [4]
For scale: in its May prospectus SpaceX said it had spent more than $15bn on the entire Starship programme to date. The new site is quoted at nearly seven times that. [4]
The announcement gave no detail on how the $100bn would be spent, or over what period. [4]
SpaceX has also been buying its way up the stack - it acquired xAI in February in an all-stock deal valuing the combined business at $1.25tn. [8]
The record flight was watched as an event in itself: a rocket-reuse record announced in advance and covered live. [6]

Forces in play

Falcon at its peak Easing

100 flights this year, a 37-flight booster, 653 landings - and being stood down from its main job [1][3]

Starship's bill High↑

a $100bn site announced against $15bn spent on the programme so far, with no schedule for the difference [4]

The handover gap Building↑

Florida stops flying Starlink on Falcon before Starship has flown an operational Starlink batch [1][2]

Where the work sits Steady→

Vandenberg carries the Starlink launches meanwhile, having already flown 44 of 76 this year [2]

In play Booster B1067 — flew 37 times, the last one closing an era The Cape workforce — moved onto Starship, which is what forced the timing Vandenberg — keeps flying Falcon Starlink, and carries the gap Louisiana — sold a $100bn number at a podium, with no schedule attached

How it unfolded

2019-2026 Falcon flies about 260 Starlink missions from Cape Canaveral [1]
This year 44 of 76 Starlink missions already move to Vandenberg [2]
25 Aug the last Florida Falcon Starlink flight, on a record 37th booster launch [1][3]
2027 construction begins at Starbase Louisiana [4]
2029 at the earliest first launches from the new site [4]

Where this points

The thing to watch is not Starship's next test but the first operational Starlink batch it carries from Florida: until that flies, the Cape has given up a working capability and not yet replaced it.

Also today

9 more stories on this beat.

A typhoon may have cost China two years on the Moon

Chang'e-7 - an orbiter, lander, rover and hopping probe, aiming to be the first mission to land directly at the lunar south pole and hunt water ice - was delayed from Sunday's window, most likely by Typhoon Narra sitting off Hainan. Chinese sources gave no explanation, and chatter suggests the next suitable window may not open until February or March. [9][10][11]

Why it matters — Polar landing windows depend on sunlight at the pole, not on when a rocket is ready. Missing one is not a delay of days. [12][13][14]
Canada is buying a space capability it does not have yet

Ottawa signed a $2.3bn, 15-year contract with Telesat for secure military communications over the Arctic using its future Lightspeed constellation, alongside funding for a first military spaceport and start-up money for three domestic rocket and satellite firms. Full operational readiness is envisaged well beyond 2035. [15]

Why it matters — Every asset in that sentence is unproven. Sovereignty is being bought as a bet rather than a purchase.
Satellite makers keep raising

Muon Space raised $250m to expand into larger, more power-hungry satellite platforms, at a valuation reported around $1.5bn. [16][17]

Why it matters — The money is moving from launch toward what gets launched, which is what a maturing launch market looks like from the outside.
The small satellites are a target now

A panel warned that small-satellite constellations face growing threats, with a dedicated watch centre in Colorado Springs tracking them. [18] A separate demonstration satellite was booked this week onto Japan's small Kairos rocket. [29]

Why it matters — Constellations were cheap because each satellite was expendable. That logic is different when something is removing them on purpose.
A telescope built from a spy satellite

NASA's Nancy Grace Roman Space Telescope - which will survey the sky far wider than Hubble or JWST can - traces its optics to hardware originally built to look the other way, at the ground. [19][20]

Why it matters — The most expensive part of a space telescope is a mirror somebody has already paid for.
Japan is going to a moon of Mars

Japan unveiled its Martian Moons eXploration spacecraft ahead of a 19 October launch, aiming to bring back a sample from Phobos. [21]

Why it matters — A sample return from a Martian moon is the cheapest way to hold a piece of the Mars system in a laboratory.
Rocket Lab keeps flying the small stuff

Rocket Lab launched its ninth satellite for the Japanese Earth-observation company iQPS on an Electron from New Zealand. [22][23]

Why it matters — While the giants argue about 100-metre rockets, the routine work of putting a specific satellite in a specific orbit went on nine times for one customer.
Washington starts asking where the pads go

The FAA opened a request for information on spaceport development and airspace access, days after the president signed a policy targeting 1,000 US launches and re-entries a year by 2030. [24][31] Separately, SpaceWERX picked 11 firms to scale military space technologies, NASA teamed with Advanced Space on a CAPSTONE follow-on, and a startup took a reusable-satellite pitch to the Pentagon. [25][26][30]

Why it matters — A launch rate is a question about airspace and land, not about rockets - which is why the request for information matters more than the target.
Curiosity climbed a kilometre

NASA's Mars rover marked a cumulative elevation gain of one kilometre on 26 August, in the fourteenth year of a mission designed for two. [27] And a chance look at an old Hubble image produced a finding that may bear on dark matter. [28]

Why it matters — One kilometre of climbing, at the pace of a machine that has outlived its design life seven times over.

02 · Lesson · why it matters

Replace it while it is still winning

The revenue and the people that fund a successor come from the thing it replaces, so the handover has to start while the old one still works - which is exactly when nobody can justify starting it.

How it works

A working asset earns money and proves itself daily
Its replacement earns nothing and proves nothing yet
So every flight the old one takes is one the new one does not
The learning the successor needs only comes from doing the job
Which means the handover must start while the old one still works
And that is exactly when no one can justify starting it

The twist

The right moment to replace something that works is while it is still winning - because you need its revenue and its people to fund the switch. Wait until it starts failing and you have lost the bridge you were going to cross on.

Where you've seen this

A profitable product line

the cash from it is what pays for its successor, and the successor is always worse on the day it launches

An old codebase

it keeps working, so the rewrite never starts, until it fails on a day nobody chose

A skill you are good at

the better you are at it, the more the next skill costs you in forgone earnings, and the more urgent it is

A power station

you must build the replacement while the old one is still generating, or you are choosing between blackouts and a rush job

The catch

The trap on the other side is real too: retire early and you carry two systems, pay twice, and gamble on a successor that has not yet done the job once. SpaceX has kept Vandenberg on Falcon for exactly that reason.

And the whole of it

Nobody involved is being reckless or sentimental. The engineers moved onto Starship because that is where the work is; the governor announced a number because that is what a governor does with an announcement; the booster flew a 37th time because it could. Between them they have retired the most reliable rocket ever built from its main job before its replacement has done that job once - and every one of those decisions was locally correct.

03 · Lab · your turn

When To Switch

Choose the year to hand over from a working machine to its unproven replacement, and find that the best year arrives while the old one is still performing well.

04 · Truth · what's really going on

Stripped of the framing

SpaceX has moved its Florida workforce onto Starship, and the end of Falcon 9 Starlink launches from the Cape is the consequence of that rather than a judgement about Falcon - which set a reuse record on its way out.

Why it lands — A round number does the arguing. '$100 billion' is quoted at a podium with no period and no breakdown, and it will be repeated for years as though it were a budget, because a figure that large ends the question of whether the project is serious.

Claimed

What people said. Not yet a fact.

Louisiana's governor

SpaceX will spend $100bn developing Starbase Louisiana, a transformational moment for the state. [4][5]

No detail was given on how it would be spent or over what period. SpaceX itself said in May that it had spent more than $15bn on the whole Starship programme to date. [4]
SpaceX's VP of launch

From here on, Starlink missions out of Florida will fly on Starship. [1]

Said on the day the last Falcon Starlink left Florida, and Starship has not yet carried an operational Starlink batch.
Chinese official sources

Nothing. The Chang'e-7 delay was announced without explanation. [9]

The absence is the claim here. Reporters attribute it to Typhoon Narra sitting off Hainan, which is a reasonable inference and not a confirmation. [9][10]

Verified

What we could actually stand behind.

The 25 August flight was simultaneously the 100th Falcon 9 launch of 2026, the 37th flight of that booster, and the last Florida Falcon Starlink mission. [1][2][3]

How we checked — Three independent outlets covering the same launch, and the retirement was confirmed by a named SpaceX executive on the record afterwards.
Vandenberg has flown 44 of this year's 76 Starlink missions and continues on Falcon. [2]

How we checked — A specific count in trade reporting, and consistent with a separate Vandenberg Starlink launch in the same week. [7]
Chang'e-7 comprises an orbiter, lander, rover and a hopping probe, targeting a direct landing at the lunar south pole. [11][12][13][14]

How we checked — Described identically across four outlets including a general-interest wire and two specialist publications, before the delay.

Nobody knows

Open questions — ours included.

What the $100bn actually covers.

No period, no breakdown, no phasing was given. Against $15bn spent on all of Starship so far, it is either a decades-long figure or a different kind of number entirely. [4]
When Starship will fly an operational Starlink batch from Florida.

No date has been given, and the capability it replaces stopped on 25 August. The gap is the whole story and nobody has sized it. [1][2]
How long Chang'e-7 is really delayed.

Social-media chatter suggests February or March, which is chatter. Polar launch windows depend on sunlight conditions at the landing site, so the real constraint is astronomical rather than logistical. [9]
Whether Canada's Lightspeed capability will exist when it is needed.

Full operational readiness is envisaged 'well beyond 2035' for a constellation that is not built, launched from a spaceport that is not built, on rockets that are not designed. [15]

Who gains

Louisiana's state government — A $100bn figure announced at a podium is a political asset immediately, and the construction it describes does not begin until 2027. [4][5]
SpaceX's Starship programme — Freeing the Cape workforce and pads is what the transition needs, and the Falcon capability being stood down is the resource being moved. [1][2]
Vandenberg — It absorbs the Starlink work, which makes it the operationally important site during the handover. [2][7]

Who pays

Anyone relying on Florida launch capacity — A working, record-setting capability has been stood down at the Cape before its replacement has done the job once. [1][2]
China's lunar programme — A weather delay on a polar mission is not measured in days; the next suitable window may be six months away, on a mission racing others to the same place. [9]
Canadian taxpayers — $2.3bn over 15 years for military communications that are not expected to be fully ready until after 2035, from assets that do not yet exist. [15]

05 · Hope · carry this

The booster that closed the era did it by flying for the thirty-seventh time. A machine designed to be thrown away after one flight came home 653 times before anyone was ready to stop using it.