First Tesla alienates their best customers, followed by the DOGE nightmare; now the SpaceX IPO. Let the clown show begin.
Before we start, let’s make one thing clear: we at The Fusion Report are hard-core fans of space technology, and more to the point, we are 100% behind investments in space ventures! With that said, what is with the clown show that has become the SpaceX initial public offering (IPO)? It’s not simply that investment bankers and Wall Street guys slavishly line up to get a piece of the latest IPO that high tech is throwing out there. It’s not simply that high tech’s clown prince, Elon Musk, is likely to piss off his customers again. Whether you think Elon Musk is channeling magic or not with SpaceX and Starlink, the truth is that Tesla (Elon’s earlier company), a company that as recently as two years ago led the planet’s electric vehicle (EV) industry and seemed like it could print money, is an impending Titanic under his leadership. That failure is only to be outdone by the lack of vision and independence of the Tesla board of directors in awarding him a massive compensation package. And let’s be honest, Elon Musk KO’d Tesla all by himself. Whether it was moving the company from California to Texas (in hindsight, not smart but kind of understandable), or backing Donald Trump who hates EVs (an obviously bad move), to running the joke that was the “Department of Government Efficiency” (which really alienated Tesla’s customer base), these were all unforced errors, and they were 100% on Elon.
How Much Is SpaceX Really Worth? (The Market’s Observations)
For the record, we at The Fusion Report are not professional stock analysts, so DO NOT use our observations to rationalize buying or not buying ANY stocks. But I have to wonder, regardless of SpaceX’s overall value, why people are lining up to pay a premium for this stock? For the record, SpaceX opened on Friday (its IPO day) at $150/share, rising as high as $176.52/share, and then closing day one at $160.95/share, a market cap of $2.1 trillion. Yesterday held similar/better numbers, with SPCX opening at $171.74 (up over $10/share from last Friday’s close), and closing at $192.50/share, up 28% over the IPO price. Note however that the long-term market cap is not something that is determined in the first few days after the IPO is launched; rather it usually takes the announcement of the first quarterly numbers before we see what value the larger market assigns.
The company’s fundamentals, while generally very good, have some weaknesses as well. While the Starlink part of the SpaceX business is incredibly profitable (and has huge room for growth), the launch operations part of the business consumes a massive amount of capital. Those capital costs could act as an anchor to drag down SpaceX’s value, even though it owns roughly 90% of the global commercial space launch market. Additionally, the AI and computing part of SpaceX lost $2.5 billion in the last quarter, and xAI and Grok continue to lose significant market share to both Anthropic and OpenAI. This is one of the reasons why Aswath Damodaran, the NYU Stern School of Business professor known as “the dean of valuation,” estimates that SpaceX’s long-term equity valuation will be about $1.3T. And that is before I even get to the issues of putting data centers in space…
Why Data Centers in Space Don’t Make Any Sense
So let’s start with the clear positive of locating data centers in space: nearly unlimited solar energy. With the right orbit (known as a sun-synchronous orbit, or SSO), an orbiting data center can get unlimited solar power 24 hours a day; given the issues with providing electricity to terrestrial data centers, that’s a big deal. Now let’s look at the negatives (and there are a bunch!):
Putting AI hardware into an SSO: Most current estimates place the cost of putting 1,000 kg into an SSO at generally between $4 million and $7 million. An orbital data center is estimated (based on the European Union’s ASCEND study) to weigh roughly 40 metric tons per megawatt of power, with the launch cost between $160M to $280M per megawatt. And in data center terms, a 1 MW data center is pretty small; earthbound data centers are looking at sizes in the gigawatt range.
Space junk: Most people think of space as being wide open; nothing can be further from the truth. There are thousands upon thousands of pieces of debris orbiting the earth of a size greater than four inches that could easily damage or destroy an orbiting data center, as shown in the illustration below. And just in case you think that light blue circle around the earth is the atmosphere, you’re wrong: it’s the density of low-earth orbit (LEO) objects floating around the earth, which is even denser than those further up.
The cost of repairing and upgrading orbital data centers: The useful life of AI computing hardware such as NVIDIA’s H100 is about two years before it is obsolete (i.e., replaced by newer stuff from NVIDIA). Assuming the data center is containerized (see illustration below), you don’t have to assemble the hardware in space, but you still have to essentially replace it every couple of years at that cost of between $4M–$7M per metric ton.
Space weather: Yes, there is such a thing as space weather, believe it or not. The primary issues are solar flares and coronal mass ejections (CMEs), but space weather also includes things like comets, geomagnetic storms, and solar radiation. In addition, satellites in an SSO also have to worry about Solar Proton Events (SPEs) because their orbits are highly inclined and pass over the poles, where the Earth’s magnetic field doesn’t protect them.
Cooling orbital data centers: Space is cold, right? However, there is no atmosphere in space, which means that the heat from data center equipment has to be radiated out into space, which is far less effective than conduction from air cooling or (even better yet) water cooling. That is in addition to the solar power panels required to produce power. A 10 kW, 8-GPU server would require 16 square meters of heat radiator area and 32.6 square meters of solar panels, for a total of (let’s say) 50 square meters per server. A 1 megawatt data center with 98 8-GPU servers would therefore require roughly 5,000 square meters of solar panels and heat radiator, an area roughly the size of a soccer field.
And all of that is required before you even start thinking about how to protect the orbital data center from intruders or other physical issues. In contrast, a ship-borne 1 MW data center could fit into roughly one shipping container, generate power from wind, wave power, and/or tidal effects, and have no problems cooling itself. Most experts believe that a repurposed cargo vessel could support a data center of up to 1.5 GW without problem. While pirates (cyber-pirates?) could be a real issue and you would likely have to fly technicians to a data center vessel with helicopters, it would be significantly cheaper than flying them on a spaceship to an orbiting data center, or protecting against space cyber-pirates.
Conclusion: Orbital Data Centers Are Not a Great Idea
There are a number of use cases where orbital platforms make both logical and economic sense: think of the Hubble Telescope and the James Webb Space Telescope, or manufacturing of drugs that can only be done in a micro-gravity environment. And space-based solar power stations could easily still make sense, though it may be a few years before they do so. However, the difference for most of these is that they are relatively light and do not generate significant waste heat.
While I will bet that future moon or Mars bases will have data centers (most likely underground), it will be because they need compute resources to analyze data from the moon or Mars, and not because either location is a better place to put a data center (everything else being equal) than on the Earth. Similarly, there may be a reason to have data centers in orbit to analyze the data generated by orbital platforms, but it won’t be just as an alternative to Earth-based data centers, whether land-based or sea-based; the economics are just not behind it.
