Space · Wednesday, 26 August 2026

01 · Briefing · what happened

Washington wants 1,000 launches a year. Nothing on its shopping list is a rocket.

Space 3 min 18 sources

A new US space policy sets a target five times today's launch rate - and asks for pads, airspace, radio spectrum and published schedules. The same week, the fix for a crowded sky turned out to be something no instrument can supply.

1,000+

US launches and reentries a year

the 2030 target, about five times last year's rate [1][3]

48%

of the public object catalogue SpaceX's sensors can see

and nothing under a metre across [10]

1.7 million

satellites currently proposed

against roughly 15,000 in orbit today [11]

flights by a single Falcon 9 booster

a reuse record, set on 25 August [7]

At a glance

The White House set a target of more than 1,000 US launches and reentries a year by 2030, up from under 200 last year. [1][3]
Almost nothing the policy orders is a rocket - it asks for pads, airspace, radio spectrum, landing ground and published range schedules. [1][2][5]
The US has four federal launch sites, and only the Florida and California ones can take a heavy rocket. [2]
Too few approved landing zones have pushed companies to bring their capsules home in Australia instead. [1]
Three careful outlets counted last year three ways - 176, 178 and just under 200 - because attempts, territory and ownership are different questions. [1][2][3]
SpaceX flew its 100th orbital mission of the year on 22 August, then reflew one booster for a record 37th time. [6][7]
It now has more than 11,000 Starlink satellites working in orbit, and plans to retire Falcon 9 once Starship flies regularly. [7][8]
Starship has still never reached orbit - its 14th test flight is the one meant to. [9]
A European study modelled the 1.7 million satellites now proposed against roughly 15,000 flying, and found astronomy substantially degraded. [11]
SpaceX is giving collision warnings away free while operating about three quarters of the world's steerable satellites - and its own sensors see under half the public catalogue. [10][15]
Rides are the scarce thing: two rideshare firms bought whole Falcon 9 rockets, and one service to high orbit is booked out to the end of 2028. [12][13]
SpaceX said it will spend at least $100bn on a Louisiana spaceport built only for Starship. [14]

Forces in play

Demand for rides High↑

two rideshare firms bought entire Falcon 9 rockets to get slots, and one service to high orbit is sold out to end-2028 [12][13]

Ground capacity Building↑

four federal launch sites, only two able to take a heavy rocket, and so few approved landing zones that capsules fly home to Australia [1][2]

Crowding overhead High↑

1.7 million satellites proposed against 15,000 flying; observatories say the streaks already spoil their pictures [11]

Operator secrecy Easing

30 operators handed over the positions and planned moves of 700 spacecraft in a single week, to a free warning service [10]

In play The White House — set the 1,000-a-year target and the 90- and 180-day deadlines SpaceX — flies most of it, and now runs the free collision-warning service The Pentagon and NASA — own the federal ranges and must publish their schedules European Southern Observatory — measured what the traffic does to astronomy Small-satellite operators — cannot get rides, and are now the ones supplying the tracking data

How it unfolded

2013 the previous US policy, when launches ran about a tenth of last year's rate [2]
Aug 20 the new policy is signed: 1,000 launches and reentries a year by 2030 [3]
Aug 22 SpaceX flies its 100th orbital mission of 2026 [6]
Aug 24 SpaceX offers operators free collision warnings, if they share their own data [10]
Aug 25 a booster flies a record 37th time; a $100bn Louisiana spaceport is announced [7][14]
Within 180 days the Pentagon and NASA must set scheduling rules for the federal ranges [1]

Where this points

Watch whether those range schedules actually get published on the 180-day clock - the policy says plainly that its own targets depend on money Congress has not yet voted. [1][4]

Full briefing

Nothing on the list is a rocket

The memorandum signed on 20 August replaces a policy last touched in 2013 [1][3]. It sets one number: US ranges able to support more than 1,000 launches and reentries a year by 2030 [1]. Read what it actually orders and the rockets are missing. Federal land surveyed within 90 days for a new site where things can come back down [1]. Scheduling rules for the government’s own launch ranges within 180 days [1]. Priority airspace corridors, and a plan to fold space traffic into air traffic control [2][4]. Radio spectrum, which appears in a policy of this kind for the first time [5].

That absence is the point. The vehicles are not the constraint. There are four federal launch sites in the country and only the Florida and California ones can handle a heavy rocket [2].

Three outlets, three numbers, one year

Ask how many launches America managed last year and careful reporting gives three answers. Reuters says 178 [3]. Ars Technica counts 176 orbital attempts from US soil [2]. SpaceNews says just under 200 by vehicles owned by US entities [1]. Nobody is wrong. Attempts, territory and ownership are three different questions, and the industry has no settled convention for which one “a launch” means. All three agree on the shape - roughly ten times the 2013 figure [2][3]. So the target is about five times a rate that is already a record.

The part that cannot be ordered

More traffic goes up into a sky that is already busy, and here the policy runs out of things it can build. A European Southern Observatory study published this month modelled the 1.7 million satellites now proposed, against roughly 15,000 in orbit today [11]. Each one that crosses a telescope’s view leaves a bright line, and the study found the damage substantial [11].

SpaceX’s answer, offered free at a conference on 24 August, is instructive [10]. Its Stargaze system spots other objects using the star cameras aboard its own satellites [10]. It tops out at about 16,500 of them, roughly 48% of the US Space Force’s public catalogue, and is blind to anything smaller than a metre [10]. The rest has to be handed over. Operators submit their own positions and, crucially, the moves they are planning [10]. A move that has not happened yet leaves nothing for any telescope to find.

Ars Technica noticed the policy reaching for the same lever on the ground. Range schedules today go public days or weeks ahead; the memo tells the Pentagon and NASA to publish them regularly [2].

Away from the traffic

Quietly, NASA’s Pandora began science this week [16]. It is a small satellite built to separate a planet’s atmosphere from the flicker of its own star, using a spare infrared detector left over from Webb [16]. A new 34-metre dish opened at Goldstone in California, the sixth in a project to expand the network that talks to more than 40 spacecraft [17]. And Ursa Major, a rocket-engine maker, agreed to go public at $2.3bn to build missiles instead [18].

02 · Lesson · why it matters

Why you can find where a thing is, but never where it means to go

Instruments only find what has already happened. A plan has not happened yet, so the only way to know one is to be told.

How it works

Instruments find where a thing is
But a planned move has not happened yet
So it leaves nothing behind to detect
Past a certain crowding, safety needs those plans
The only way to get them is for everyone to say
And saying means agreeing to be seen

The twist

No instrument can detect an intention, because it has not happened yet - so past a certain crowding, safety stops being a matter of looking harder and becomes a matter of everyone agreeing to say.

Where you've seen this

Driving

your indicator tells other drivers the one thing no camera could see - what you are about to do

Air traffic

aircraft broadcast who they are and where they are going, rather than waiting to be found by radar

Hospitals

near-miss reporting only works because staff volunteer the accidents that did not happen

Roadworks

a published schedule prevents jams that no amount of traffic monitoring can

The catch

Declaring costs something - it shows rivals your plans - so these systems run on a bargain, not a machine, and whoever everyone declares to ends up with a view of the sky that nobody else has.

Full lesson

Forty-eight percent

SpaceX has the best privately built eye on low orbit. Every Starlink satellite carries star cameras, meant for working out which way the satellite is pointing. Turn them outward and they catch other objects drifting past. Thousands of satellites, watching in every direction, all the time.

It sees about 16,500 things. That is roughly 48% of the public catalogue kept by the US Space Force, and nothing smaller than a metre across.

Now here is the part worth slowing down on. Even if that number were 100%, it would still only answer one question.

The question no telescope answers

Knowing where a satellite is does not keep it from hitting anything. What keeps it from hitting anything is knowing where it will be in eight hours. And the largest term in that answer is not physics. It is a decision somebody has already made and not yet carried out.

A planned engine burn emits no light. It reflects no radar. It has no position, no temperature and no signature, because it has not happened. It exists as a line in a schedule inside a company.

Build a bigger telescope and you learn nothing about it. Build a thousand and you learn nothing about it. The information is not faint. It is absent.

Which is why the fix was not a machine

Two things happened this week and they made the same move.

A company offered every satellite operator free warnings about close approaches - on one condition. You have to hand over your own positions and the moves you are planning. Thirty operators did, for seven hundred spacecraft, in a week.

And a national policy aimed at multiplying launches told the government’s own launch ranges to start publishing their schedules, which today go public only days or weeks ahead.

Neither of those builds a sensor. Both ask someone to say something. That is what a system does when it has run out of things it can detect.

Saying is not free

This is why disclosure systems are always harder than they look.

Your positions are an asset. Your planned moves are an intention. Together they tell a competitor where your fleet sits, how often it has to dodge, and how much fuel you are burning to do it. A published launch schedule tells rivals when you fly, and every time you slip.

So each participant is asked to accept a small real cost for a large shared gain. That is a bargain, not an installation. It holds while enough people think the trade is worth it. It comes apart quietly when they stop - not with a failure, but with people filing less, and later, and vaguer.

Where the picture ends up

There is an arrangement underneath this that is easy to miss because it looks like plumbing.

Somebody has to be the place everyone declares to. This week that place is a free service run by the company that already operates about three quarters of the world’s steerable satellites.

Free is not a trick. It is the cheapest way to become the place. But the result is a view of orbit assembled out of everyone else’s disclosures, held by one of the parties in it. That is not a scandal. It is what tends to happen when something everyone needs has no owner, and one participant is willing to pay to run it. The question of where the data lands looked technical. It was a choice.

You have done this today

You did it if you drove. Nothing about a car announces a left turn. The physics of a moving vehicle carry no hint of it. So somebody had to invent a lever whose whole purpose is to tell strangers what you have decided. Then everyone had to agree to use it, at the cost of showing their hand early.

Aircraft do it by broadcasting who they are and where they are going, rather than waiting to be found. Hospitals do it by asking staff to report the accidents that did not happen. In every case the useful information was not hiding. It never existed outside somebody’s head until they put it somewhere.

Nobody has the whole thing

The satellites doing this dodging are the ones carrying the map on your phone, the forecast on your window, the signal that times the payment when you tap a card.

And no one seat sees the situation. The company sees under half of it. The Space Force keeps a catalogue with gaps. Each operator knows its own fleet and nobody else’s. The astronomers in Chile see only the streaks left across their pictures.

The picture exists only while it keeps being assembled from things people choose to say. Every one of them holds a piece they have some reason to keep. That is worth remembering the next time a system looks like it is being watched. Quite often, it is only being told.

03 · Lab · your turn

The Declaration Desk

Rehearse choosing between looking harder and telling others what you plan, when telling costs you something.

04 · Hope · carry this

Thirty operators handed over their own flight plans in a single week, for nothing in return. When the only way to be safe together is to be seen, people turn out to be willing.