Written by GPT-5.6 Sol under Leo's direction. Human-directed Workbench essay, 15 September 2026.
I got into this by asking when Starship was launching.
Then I asked how long it would take to fly to Brownsville.
Then I laughed at the airport code BRO.
A few questions later I was staring at launch economics and thinking, oh. Oh, God. I get why people keep pouring billions of dollars into this.
My day-to-day emotional investment in spaceflight is close to zero. I’ll happily sleep through booster telemetry. Mars colony discourse can continue without me. The breathless civilizational language around every test flight wears me out pretty quickly.
But the underlying industrial case is much better than I had given it credit for.
The click came when I stopped thinking of the rocket as the product.
The rocket is an input
A rocket by itself is an absurd object.
You spend years designing engines, tanks, avionics, plumbing, software, launch sites, test stands, factories and recovery systems so that a gigantic tube can accelerate another object to orbital velocity.
Viewed as spectacle, the whole industry invites a very understandable response:
Cool. Expensive. Why?
Viewed as an input, it starts behaving like compute, bandwidth, electricity, container shipping or semiconductor fabrication.
Cheap launch makes other things possible.
More satellites. Larger satellites. More frequent replacement. Different communications networks. Better Earth observation. More scientific instruments. Bigger telescopes. Orbital servicing. Space stations. Experiments whose economics looked stupid when every kilogram was precious.
NASA describes Falcon 9 as the first orbital-class reusable rocket and says reuse lets SpaceX refly the most expensive parts of the vehicle, driving down the cost of access to space. NASA's current launch-services roster also includes New Glenn, Starship and Terran R, each carrying a different bet on reuse and scale. (NASA)
Once I saw the rocket as a capital good, the investor logic became much less mysterious.
Semiconductor fabs earn their keep by creating streams of chips. Data centers earn theirs by selling compute.
A reusable launch system has the same basic ambition: build an expensive machine, use it many times, spread its cost across a huge amount of useful output.
The business question becomes much more concrete:
How much useful orbital work can this machine do over its life?
It is a much better question than whether the launch looked cool.
Reuse is the entire economic seduction
Throwing away an expensive machine after one trip makes the machine dominate the economics.
Using it again changes the arithmetic.
Commercial aviation feels obvious because nobody expects an airliner to disappear after Toronto to London. The purchase price gets spread across years of flights. Maintenance, fuel, crews, airport fees and financing remain, but the vehicle survives the trip.
Rockets spent most of their history on the disposable side of that line.
Now several companies are trying to move orbital launch toward fleet economics.
Falcon 9 already lives much farther into that world than the old expendable model. NASA calls Space Launch Complex 40 a high-cadence launch site and says it had handled more than 310 Falcon 9 missions by early 2026. (NASA)
Blue Origin's New Glenn first stage is designed for at least 25 flights. (NASA) Rocket Lab is developing the reusable Neutron. Relativity is developing the reusable Terran R. Stoke is developing the fully and rapidly reusable Nova.
The headline maiden flight is among the least interesting parts.
I care much more about flight 12.
Then flight 30.
Then whether the same hardware flew again after a week, what had to be replaced, how many technicians touched it, how much inspection it needed, how often an engine comes off, how many vehicles the factory can finish in a year, and how the price behaves when the company is launching all the time.
At that point the semiconductor analogy came back and punched me in the face.
Rockets may be entering their repetition era
Semiconductors used to be weird too.
Early computers were huge, expensive, fragile machines built in tiny quantities. Then transistors, integrated circuits, photolithography and industrial semiconductor manufacturing created repeated production on an insane scale.
Repetition changes what engineers can know.
Make one precious object and every failure is a catastrophe plus one data point.
Make thousands or millions of related objects and the process itself becomes a learning machine. Yield becomes visible. Failure modes become statistical. Suppliers improve. Testing improves. Production steps get revised. The next generation inherits a giant pile of practical knowledge.
Rocketry has always had testing, production learning and fleets. The catch is that orbital vehicles historically flew in tiny numbers compared with chips, cars or aircraft components, and much of the expensive hardware vanished after use.
Reuse plus higher cadence changes the information flow.
Fly a booster. Recover it. Inspect it. Find what wore. Change the part. Fly it again. Build another. Compare them. Keep going.
The thing I had been reading as "these programs seem embarrassingly unpredictable compared with semiconductors" started looking more like an industry trying to reach the stage semiconductors already reached decades ago.
The OECD's new Space Economy at a Glance 2026 estimates global space-related revenues at $550 to $600 billion in 2025, already approaching the roughly $620 billion scale it cites for the global semiconductor industry in 2024. (OECD)
The comparison deserves restraint because launch remains far younger and lower-volume than chipmaking.
But Jesus Christ, okay.
This is already a large industrial economy.
The competition is weirdly healthy underneath a huge leader
SpaceX has an enormous operational lead in reusable orbital launch.
Yet the field underneath it contains genuinely different technical bets.
Blue Origin has New Glenn. Rocket Lab has Electron in service while developing Neutron. ULA has Vulcan. Relativity is developing Terran R. Stoke is developing Nova.
The U.S. Space Force's broader National Security Space Launch Lane 1 contract pool now includes SpaceX, Blue Origin, ULA, Rocket Lab, Stoke, Relativity and Impulse Space. (Space Systems Command)
Seven providers sitting inside one procurement pool span very different levels of maturity. The useful signal is that the customer wants options and several teams have cleared enough gates to compete for work.
Commercially concentrated, technologically crowded.
This is a pretty interesting place for an industry to be.
A dominant company has already proved that reuse can work at serious cadence, and a collection of rivals now has a concrete target to beat, copy, specialize around or leapfrog.
Competition becomes less philosophical at that point.
Can you launch?
Can you launch again?
Can you recover the hardware?
Can you give the customer the orbit they bought?
Can you do it often?
Can you make money?
Great. Keep going.
Oh, right. The government is one of the customers.
This part cleared up another piece of the puzzle for me.
The space business gets narrated as either heroic private capitalism or giant government boondoggle depending on who is talking.
The real arrangement is much messier and more ordinary.
Private investors supply risk capital. Billionaires supply patient capital. Public markets supply capital to companies like Rocket Lab. NASA buys transportation, landers, science missions and development milestones. The military buys launch and satellite capability. Commercial satellite operators buy rides. Telecom customers pay for services that depend on satellites already in orbit.
NASA's Commercial Crew Program explicitly describes itself as a partnership with American private industry to provide astronaut transportation to and from the International Space Station. (NASA)
So the government can be an anchor customer for a capability a company wants to keep selling elsewhere.
Investors get a possible future market.
The government gets competing suppliers and a service it wants.
The company gets paid while accumulating factories, engines, flight history, staff and know-how that remain useful after the original contract.
This can absolutely produce waste, favoritism, overruns and dumb procurement. Every giant purchasing system can.
It can also bootstrap a market that would be difficult to finance from consumer demand alone.
Rocket Lab made the financial picture click
Rocket Lab is useful because it is public, so there is an actual income statement instead of endless guessing about a private company's books.
In Q2 2026 it reported $234 million in revenue and $2.36 billion in backlog, while continuing to pour money into Neutron and acquisitions. It also said its launch backlog had grown past 90 missions across Electron, HASTE and Neutron. (Rocket Lab)
There is a mature-business / future-business tension sitting right there.
Sell today's launch and spacecraft products.
Use the revenue, investment and contracts to build tomorrow's larger capability.
Lose money at the company level while the expensive development program is running.
Hope the new vehicle eventually turns a giant development bill into an asset that can produce revenue for years.
Suddenly the losses look less mysterious too.
A company can have real customers, real gross profit, real backlog and still consume capital because it is building the next factory, engine, vehicle and product line.
Semiconductor companies do versions of this every time they commit fortunes to the next fab or process generation.
The bet is that the expensive capability becomes productive.
"Why don't we solve hunger first?" is a lousy capital-allocation model
The moral objection deserves a cleaner answer than space people sometimes give it.
Hunger is real. Emergency aid is valuable. The World Food Programme says it needs about $13 billion in 2026 to reach 110 million vulnerable people. (WFP)
A dollar sent there can do something valuable right now.
Civilization also invests in capabilities whose payoff arrives later and whose final uses emerge over time.
Drug research.
Energy.
Computing.
Transportation.
Scientific instruments.
Manufacturing.
Education.
Space.
If every long-horizon investment had to beat emergency relief on immediate suffering reduced per dollar, huge portions of research and capital formation would disappear.
The moral arithmetic gets stranger once private money enters the picture. A billion dollars invested in a rocket company pays engineers, suppliers, machinists, software teams, factories and tests while creating an asset the investor hopes will become more valuable. A billion dollars donated to food assistance performs a different job.
Both can be worthwhile.
Civilization is allowed to feed somebody today and build a better machine for somebody twenty years from now.
The clearest historical example may be GPS. It began as a government space capability. A NIST-sponsored study estimated that civilian GPS generated roughly $1.4 trillion in economic benefits for the U.S. private sector from the 1980s through 2017, across fields including agriculture, finance, telecommunications, surveying and navigation. (NIST)
A person arguing against expensive satellites decades ago could easily have asked for a more immediate use of the money.
The downstream applications were the point.
I can keep the anti-hypebeast part
I can keep rolling my eyes every time somebody says "multiplanetary."
Good.
The hype around space can be exhausting precisely because the real thing is already interesting.
The engineering already gives a rocket company plenty to prove:
more flights;
more reuse;
shorter turnaround;
fewer refurbishment hours;
higher factory output;
successful payload delivery;
real backlog;
better margins;
lower customer prices;
new services that become economical because launch got cheaper.
Give me the boring numbers.
I want to know whether the machine is becoming ordinary.
A company can be technically brilliant and absurdly promotional at the same time. A founder can have grandiose timelines while the engineers underneath him produce excellent work. A launch can be a genuine achievement while the surrounding civilizational narration makes me roll my eyes.
Easy enough.
Discount the sermon. Watch the learning curve.
The miracle eventually becomes plumbing
The strongest pro-space argument, for me, has very little to do with Mars.
It is the possibility that orbital access becomes boring.
GPS became boring.
Satellite television became boring.
Looking at a map on a phone and having a blue dot know where you are became boring.
Transoceanic communications became boring.
Computers became boring enough that we complain when a pocket supercomputer takes two seconds to open an app.
Successful technology gets absorbed into ordinary expectations.
Maybe reusable launch keeps moving in the same direction.
Maybe a few of these companies fail spectacularly. Almost certainly some will. Maybe Starship takes longer than advertised. Of course it will. Maybe New Glenn, Neutron, Terran R or Nova finds a niche nobody currently expects. Maybe government demand carries some companies farther than commercial demand ever would.
The heroic story can stay in the brochure.
The industrial story is enough.
Build expensive machines.
Fly them.
Get them back.
Learn from them.
Make more.
Spread the fixed cost across more useful work.
Let governments buy missions, let investors gamble on the upside, let competitors chase one another, let customers invent uses once the input gets cheaper.
Then, every once in a while, look up and see whether another ridiculous human achievement has quietly become normal.
I can get out of bed for that.