SpaceX IPO
Tails, SpaceX Fails, Heads We Lose the Earth
As a starting point, it’s important to remember that SpaceX has grown beyond its name. Yes, it is a rocket company responsible for launching the Falcon 9, Falcon Heavy, and Starship series rockets.
It is also an internet service provider, through Starlink, an AI company as home of Grok’s parent, xAI, as well as a social media platform since it owns Twitter… which is now X.
The successes and failures of each of these major divisions defines the success of the company as a whole, and before you decide if you should buy SpaceX, you need to decide if this mismatch of business models combines into one sensible whole .
This is a free story. If you like it, please consider supporting our work by joining our Patreon.
subscribe now
As part of the filing, SpaceX reported on each programs’ status as of March 31, 2026, the close of the first quarter.
As of that date, SpaceX had ~640 launches with 85% using 1 or more previously flown booster. Across these missions they’d launched 80% of the mass to orbit, including 78 humans. To me this number seemed low until I realized that astronauts stay on the ISS for a long time, so not a lot of humans are launched.
A major part of SpaceX’s launch program is its own Starlink satellites. Currently there are over 9600 satellites of what the FCC has approved to be a 12,000 satellite constellation. At the end of Q1, Starlink had 10.3 million subscribers in 164 countries.
For comparison, LA has a population of 12.7 million and London has a population of 10.4 million (via Wikipedia). People to subscribers isn’t a one for one comparison however, so let’s look at a couple big internet service providers here in the US. Comcast serves 51.2 million subscribers and Charter serves 29.6 million.
It is in this context that I want to look in detail at what is involved in the Starlink network.
Starlinks last for about 5 years before their orbits decay and they burn up in the atmosphere. They are launched in batches of 20–29 satellites per launch, depending on the version. Let’s assume that SpaceX replaces ⅕ of their network every year, or about 2400 satellites, with an average of 24 satellites per launch. That would require 100 launches per year just to maintain their network. Launches cost about $74 million each, per Satbase, which works out to $7.4 billion a year in launch costs. Hardware is another $500 million to a billion a year. With 10.4 million subscribers, that works out to $770/year per subscriber just to build and launch the satellites.
And yes, I’m ignoring Starship as a potential launch vehicle because it is so far behind.
The need to constantly replace satellites isn’t going to go away, so SpaceX will need to grow its subscriber base to increase its profits over time. Let’s assume 20% year-on-year growth as a starting point. That gets them to 22 million subscribers by 2030, which is still small compared to Comcast, but it brings their per user infrastructure replacement costs down to $360 per subscriber per year.
With a max speed of 500 Mbps and a listed typical speed of 65–260 Mbps (roaming), Starlink can’t compete with fiber or cable internet. Its speeds are also similar to cell service, but with a $175/month unlimited roaming plan, Starlink is also much more expensive than my AT&T cell plan (plans).
So we have to ask, can they find the customers?
The kinds of folks who are getting Starlink are truckers, motorhome riders, people who need backup to their traditional internet service, people who are remote or off the grid, as well as planes, and boats.
There are about 2 million US long-haul (heavy/tractor-trailer) truck drivers, & 6000–10,000 airplanes (estimated from peak traffic of 5,500 planes in flight).
Your average truck driver in Peru won’t be able to afford Starlink, and your average truck driver in the US will use cellular.
I can see Starlink’s subscriber-base saturating fast.
They do have one other important market — the military.
And I don’t know how to factor that in as a potential growth market. Making things more complex are Kuiper, OneWeb, Bluebird and other megaconstellations that are starting to launch.
We need satellite communications companies. They are necessary for everything from forestry to mining to emergency response to being an off-the-grid gremlin — something I may have dreamed about as I read the prospectus.
Ultimately, Starlink is a smaller telecom company than Comcast, with slower speeds, higher subscription costs, and infrastructure that has to be replaced every 5 years instead of every bad hurricane.
This isn’t a great model, and their customer base has the potential to saturate quickly.
But what about other parts of the company? xAI, with on-orbit data centers, is billed as the future, but does the model make sense? (Twitter/X is a subdivision of xAI.)
As of March 31, xAI had 350 million daily posts, and ~550 million active users. I presume this includes every shit post of “Hey Grok, what’s the ____ on this” and every request to make a girl in a photo naked. Or a guy. It’s sexist, but will do both. xAI also has 1 gigawatt of Nameplate Compute Draw, which means that if every CPU and GPU are maxed out on their processing, the system will draw 1GW. A typical household in the western world averages 6–12 kWh of power consumption. That means the data center uses the same power, at maximum, as 125 households in 1 hour. This is a lot of power, but it is small compared to other companies, with AWS constructing roughly 4 GW of data center capacity in just 2025.
But xAI isn’t looking to be a ground-based compute network — they are aiming for space, with plans to deploy 100 GW of compute to orbit via 100kW per ton satellites that require 1000s of launches per year.
Folks, data centers in space don’t make sense. The thermodynamics don’t make sense.
Rather than running the numbers, let’s look at an on orbit example. The ISS, with its massive solar panels, generates 80–100 kilowatts of power, and has radiators that are similar in area to those solar panels. Now imagine that every ton of data center requires this area of solar panels and radiators. … And they are getting launched year after year. They are looking at roughly a million tons to orbit — that’s a million tons of steel, aluminum, carbon fiber, rare earth minerals, and more.
This is about the same mass as the US yearly soybean harvest getting launched into space… each year.
And all this mass will burn up in the Earth’s atmosphere. This material can’t be recycled. It is simply lost as it becomes one more thing changing the Earth’s atmosphere in ways we don’t understand.
Per SEC filings, SpaceX’s mission is “To build the systems and technologies necessary to make life multiplanetary, to understand the true nature of the universe, and to extend the light of consciousness to the stars.
My deeply sarcastic inner voice responded: “Of course we’re going to spread humans to other planets and consciousness to the stars. Building and maintaining this company will destroy the Earth, and living in a computer just might be easier than living on Mars.” My inner voice needs a whiskey.
Sarcasm aside, we have to ask, to what end are we doing this? To what end is SpaceX planning this many launches? Ultimately, they are hoping that if they build it, the audience will come, and they will become profitable.
But in Q1 2026, SpaceX’s costs exceeded their revenue by $2 billion (using their Segment income). If this rate stays constant, they will lose more money in 2026 than NASA will spend on science.
Let me put this as clearly as I can. NASA’s entire FY26 Science Mission Directorate budget is $7.25 billion and SpaceX is on target to lose $8billion.
Do you believe that Grok and xAI will need to scale by a factor of 100 year on year? Do you believe the customers and their pocketbooks will be there to fund these launches?
And do you believe that Starship will be ready to begin launching commercial satellites by the thousand by the end of the year?
The FAA has so far approved a potential 44 Starship launches a year. This IPO requires “thousands of launches” (p. 9). Musk, on various things, has said he foresees 10k launches a year (example), which is thousands. This will require new launch sites/pads/atmospheric badness.
Humans are bad at understanding large numbers, so let’s break this down. 10,000 launches a year is 27–28 launches per day. Let’s assume these are ballistic launches that go up, deploy satellites, and come back down, with a 6 hour duration from propellent loading to landing. Musk claims he’ll have 8 hour turn around on Starship, so let’s call it 18 hours, fully optimized, from landing to back on a pad.
These are crazy town numbers, but let’s just see what these crazy town numbers imply?
If each Starship can launch once a day, and we assume 20% of the fleet is being refurbished or upgraded at any given time, that’s a fleet of 36 starship. That is totally reasonable when you compare that to the number of 777 aircraft in service (example, American Airlines).
But Starship uses a lot more fuel than flight.
Each Starship launch uses about 1,030 tons of methane, a gas that can come from many sources, but primarily is derived from Natural gas. The US exports about 100 million metric tons per year of Liquid Natural Gas. To make the math simple, let’s assume that 95% of the liquid natural gas can be converted to methane. That’s 95 million tons a year. With 10,000 launches a year, Starship alone will use 10.3 million tons of methane, and will need to acquire the equivalent of 10% of US exports for launches. Methane is a powerful greenhouse gas.
SpaceX is looking to use a powerful greenhouse gas to build an on-orbit infrastructure that simply burns up in the atmosphere after a few years without any chance of recycling.
And all this discussion ignores the number of Starships that will need to be built and then stored on orbit awaiting a Mars launch window to get that million-person colony on Mars build… You can learn about that in this past episode.
Long story short, SpaceX is looking to support 10–20 thousand starship launches per year using a significant amount of natural gas, a noticeable amount of steel production and somewhat horrifying amounts of rare earth minerals… all of which will end up in one form or another in the Earth’s atmosphere.
Do I think they’ll succeed? No.
Can our planet survive if they do succeed? Also, no.
As a planet, we can’t afford for them to do what they want to do.
The data centers in space, the mega constellations, temporary internet satellites, 1000 Starships to Mars… this adds up to a depletion of Earth’s resources & atmospheric harms.
Here’s the thing, though: if this IPO fails, it could start the collapse of the AI bubble. It’s going to happen. The move to token-based pricing made that crash all the more likely. When that market crashes, it’s going to take the global economy with it, & that economy already sucks due to wars. SpaceX has gotten special dispensation to be added to index funds before it is a proven stock — this jeopardized my retirement accounts and yours… if you’re lucky enough to have one.
If the AI bubble bursts, potentially because of SpaceX being unrealistic in the ways I’ve detailed here & others, we’re going to face a crash under a president who frankly doesn’t care about people and has dismantled the social welfare system.
One man’s dream of Mars is my economic nightmare.
And it is all a problem of scale. Sometimes, we need to dream big things… at smaller scales.
I want SpaceX to get their shit together enough to launch Starship a few times a year and get us going to the Moon as they were supposed to do in 2025. I want what they promised, not what they’re building.
We can’t turn back time, however.
And we can’t afford to cave to the sunk cost fallacy.
Friends don’t let friends buy SpaceX.
That’s all I’ve got.
This content was originally written for the EVSN podcast. Watch video episodes and check out additional content at EVSN.tv.
