Launch Is Not Solved Yet — Notes on Andy Lapsa and the Threshold Nobody Has Crossed

In the south there are many divers, for they live with the water daily. At seven they can wade, at ten they can float, at fifteen they can dive. Is diving something one arrives at carelessly? They must have come to know the way of water.

—— Su Shi, On the Sun as Analogy (Northern Song, 1078; my own rendering)

What This Episode Is About

Valley of Depth’s 2026-09-08 episode features Andy Lapsa, CEO of Stoke Space. That week they had closed a billion-dollar round, bringing their total to about $2.3 billion; that same morning the NASA administrator dropped by their Moses Lake test site, so they fired two engines for him. The host asks a sharp question: how did an industry go from “everyone assumes launch is getting cheap” to “everyone is panicking about launch”?

What stayed with me isn’t the rocket. It’s how he sorts problems into kinds. Before leaving Blue Origin in 2019, the first thing he did was put roughly 150 companies into a spreadsheet, trying to find which one would win so he could go work there. What came out of that exercise was the reason he started his own instead: only one thing ends up mattering, and everything else lives at the margins.

The Points Worth Keeping

The size of the launch market equals how many rockets leave the pad each year, and not one drop more. That’s his answer to investors who say launch has a modest ceiling. The whole space economy is capped by that layer — not for lack of demand, but because rockets worldwide are already going up as fast as everyone can execute. The layer that’s stuck sets the ceiling for everything above it.

Flight rate plateaus in the teens, for companies and for nation states alike. If you throw the vehicle away, how often you fly equals how fast you can build. Plot SpaceX’s rate and you see the same flat line in the teens — until they land a first stage on a barge, and within a couple of years it takes off; roughly 165 flights last year. He credits that stage recovery with creating the industry, but he won’t call it a 70% solution. Going from a hundred flights a year to thousands means the upper stage has to come home too.

The threshold nobody has crossed: bringing a vehicle back and flying it again without inspecting it, without refurbishing it. He says that in the present tense. It comes down to heat shields. Stoke uses a metallic shield with cooling channels in its surface, pumping the liquid hydrogen already on board through it, with that circuit integrated into the rest of the second-stage engine plumbing. His line: you can walk up and hit this thing with a baseball bat and not care. The contrast is ceramic tiles — mature technology, flown by NASA for decades — but materials that survive thousands of degrees passively end up brittle: micrometeorites, bird strikes, a dropped wrench, hard to bond. The origin of his approach isn’t exotic. It’s how you cool the inside of a rocket engine, and an engine’s environment is about twenty times more intense than reentry.

Beyond cost and availability there’s a third pillar: reliability. Every calendar year since 1980, between 2% and 8% of all orbital attempts worldwide have failed. That’s roughly the lemon rate in cars — except you can’t tow a rocket back to the shop. His inference: without flying proven hardware again and again, you won’t push past that number. The industry now has the data — SpaceX first stages have gone hundreds of flights without missing mission, and the handful of failures to orbit were second-stage anomalies fresh out of the factory. So full reuse changes reliability’s order of magnitude, not just its price, and that’s what makes human flight a serious conversation.

He bets on engineering, never on science. If your plan needs a breakthrough — materials science included — you’re praying for a miracle, and that isn’t a company he’d raise outside capital for. So they chose full-flow staged combustion: hard, attempted a few times, done successfully once before them, but known to be possible. Why pick the most complex engine? Because reuse costs you mass. Saved propellant, landing gear, heat shields — each comes out of payload. You need elbow room in the architecture, and the only way to buy it is the highest fuel efficiency humans know how to build.

The new 15-ton class was reverse-engineered from the market, not chosen on a whim. Their fixed infrastructure was deliberately oversized, so scaling doesn’t mean starting over. Fifteen tons is the industry’s de facto standard, which reaches government payloads and telecom constellations. Go bigger than the market supports and you get high operating costs with empty fairings. There’s a counterintuitive point buried here: for constellation customers, stockpiling hundreds of satellites on the ground to fill one big rocket is waste, because the metric is time from factory to revenue, and satellites launched several planes at a time spend months phasing into final orbit. If launch is available on demand, filling one plane at a time is the efficient bite size.

Going Further

”This theme has been talked about for a decade and delivered nothing — is it early, or is it dead?”

This is where I get stuck most often. Money burned, story told, years gone, nothing on the books. I used to judge it by how much time had passed, which turns out to be no criterion at all.

The episode handed me a better question: does the stuck step need new science, or does it need the same thing done many times? The probability structure differs. The first has no schedule and no partial credit. The second gives you milestones, and when it fails you know which step failed. When Andy says betting on a materials breakthrough is praying, I thought of theses I’ve read that look rigorous until you hit one sentence — “assume this efficiency reaches that level within a few years” — with no engineering path underneath, just a line sloping up.

Running that question across my own holdings, two of them turn out to rest on prayer-shaped unknowns. That doesn’t mean sell. It means position size and patience need a different standard, because “let’s give it another quarter” isn’t a unit of measurement for that kind of thing.

”A billion raised, the administrator visits, the market loves it — do I chase?”

The timing of this episode makes it a good drill. Funding rounds, famous visitors, media volume — what they share is that none of them are falsifiable. None tells you what would count as being wrong.

Meanwhile Andy lays out something you can check. Flight one goes up and out on purpose, not coming back, as a statement about engine performance (he says the engine is sandbagged today). Flight two demonstrates on-orbit capability to build confidence for reentry. Flight three attempts the reentry and the heat shield. That’s ordered and checkable, and he volunteers up front that even a perfect first flight will produce changes before the second.

Listening to that, it struck me he’s handing outsiders a ruler. After the third flight you can look back and score how much of what he said came true — far more informative than tracking the next valuation. Flip it around: a story that can’t produce a sequence like that leaves you with no ruler, only mood.

He also offers his own explanation of the whiplash, and I think it’s the useful part for retail investors. Why did the mood flip? Confirmation bias: people watched SpaceX do it, assumed it was easy, assumed everyone would now do it, and treated it as done. One company doing something and an industry replicating it aren’t the same event. I make the same mistake with technology stories — reading “someone demonstrated it” as “this is settled,” then pricing that future into an entire supply chain.

”I like this trend — which layer do I buy?”

The line I most wanted to write down is that the launch market is exactly as big as how many rockets leave the pad. It compresses a complicated industry into something countable: the throughput of the stuck layer is the ceiling for everything stacked on top.

But the second half matters as much, and he says it himself: being the bottleneck isn’t the same as collecting rent. Oversize the vehicle beyond what the market supports and you’re left with fixed costs and empty payload bays. He also chooses to stay payload agnostic rather than compete with his customers, precisely so the market keeps forcing him to fly more often at lower cost. He’s treating this layer’s scarcity as temporary and execution-dependent, not as a structural toll booth.

So I ask the two separately. Which layer is the bottleneck — that sets the ceiling. How long can that tightness hold, and who can defend it — that sets who earns. Miss the first and you hunt for names on a chain whose ceiling is already sealed. Miss the second and you buy a shortage that evaporates the moment capacity arrives.

Where to Read More

• The Valley of Depth episode of 2026-09-08 with Andy Lapsa of Stoke Space, on any podcast platform
• SpaceX annual launch counts and booster reuse records: public launch statistics, enough to draw the flat-then-vertical curve he describes
• Yearly global orbital attempts and failures: several public launch databases carry the year-by-year record against which his 2%–8% can be checked
• Shuttle thermal tile maintenance and damage history: NASA technical documents and accident investigation reports
• Full-flow staged combustion: public propulsion literature and engineering texts on the cycle and its earlier attempts
• Stoke Space’s own published material on Nova and the Zenith engine

One Thing to Take With You

One idea: “not solved yet” and “cannot be solved” are different animals, and you tell them apart by asking whether the stuck step needs knowledge you don’t have, or needs the same thing done many more times. The difference isn’t difficulty. It’s whether you can name the next step. Knowledge-shaped problems leave you waiting with no progress to observe; repetition-shaped problems give you milestones and let you know where you are.

Here’s something I tried that has nothing to do with markets. Pick one thing you’ve been talking about for three years and haven’t done — changing jobs, learning an instrument, having the conversation you keep postponing, getting your body back. Write down the step it’s stuck on, then judge one thing only: is what’s missing something you don’t yet have (a piece of knowledge, a relationship, an identity), or is it the twenty ugly repetitions you haven’t done? Then look back at what you spent three years preparing for, and see which kind it was preparation for.

When I did this, I found I’d been treating a repetition problem as a knowledge problem for years — reading, researching, waiting to be ready, when the only thing missing was sitting down and producing the first few bad versions.

Raw Markdown

This article is an educational discussion of investment method. It is not advice to buy or sell any individual security, offers no target prices, and does not analyze any current holding. Investing carries risk; make your own decisions or consult a qualified professional.