SDCs and Commercial Uses of Space
- WSP Rhodes
- Jul 12
- 21 min read
Forgive my hiatus, I’ve been dealing with some personal issues.
If you’ve been paying attention to the news recently, you might have heard a few famous tech CEOs, particularly Elon Musk, Jeff Bezos, and Sam Altman, discussing the possibility of putting data centers in Earth orbit. As the explosion of AI technology has necessitated more data centers to process the required data, it has put strain on energy grids and water resources in order to power and cool these newly constructed processing facilities. Proponents of this idea argue that these orbital data centers could be powered independently by solar panels and would take up no land area. This is easily the most cartoonishly impractical idea I have ever had to talk about here, showing a profound misunderstanding of physics and economics. The best case scenario is that this idea is one more ploy to attract investors into a technology that still isn’t profitable, and either way it is hard proof that being the CEO of a sci-tech company is not the same thing as being a scientist. I wanted to talk about this in large part to inform you of the realities of this fundamentally absurd concept, but I thought it would also be a good way to talk about the concept of space commercialization as a whole. As improvements in reusable rocket technology bring launch costs down further (Falcon Heavy has 10% the launch costs of the space shuttle), economic uses for space are becoming a serious topic of discussion. In theory, finding such uses could be a boon for humanity in general, leading to more plentiful energy, greater resources, and higher quality manufactured goods, not to mention making space exploration cheaper as there would be economies of scale for rocket launches. How true all this would be in practice is…a good subject for a blog post.
Thermodynamics
One of the major problems that data centers pose to our environment is the waste heat they produce. Computers naturally give off heat, so the largest data centers today can produce 100 million watts of waste heat, equivalent to energy needed to heat 100,000 homes. This requires water to act as a coolant, so these data centers use as much water as 1,000 homes. The obvious solution would be to use this waste heat for something useful, such as power generation or district heating, and some data centers have tried this. But this is low-grade waste heat, meaning it can’t be easily recycled. Boiling water for steam needs something to be heated to 100℃ (212℉) and most district heating systems need to be at least 65℃ (150℉) to work properly. The hottest GPUs only get up to 60℃ (140℉). Now, it is possible to turn a lot of 60℃ hot water into a smaller amount of 65℃ water, and there are data centers that have had success in doing so. But doing this requires a heat pump which requires electricity, which defeats some of the purpose of recycling waste heat. And since reusing waste heat isn’t so easy, most data centers don’t bother. They pump the hot water they generate into cooling towers, which evaporates their cooling water off little-by-little so it can’t be reused. With this being such a problem, putting data centers in the cold vacuum of space does seem to be an attractive idea.
Slight problem; space isn’t cold, at least not in the way you think it is. Quite the opposite.
The most effective way of cooling yourself is to touch something that’s colder than you are. This could mean conduction, where you touch a cold solid and your body heat defuses into it, or it could mean convection, where you touch a cold liquid or gas and the moving molecules absorb heat and carry it away from you like a cooling conveyor belt. For both of these to work, you need to touch something cold, be it a cold surface, cold water, or cold air. But space is a vacuum, there is nothing to touch. As such, vacuum is an incredibly effective insulator and keeping things from overheating is a major engineering challenge for satellites, space stations, and spacesuits alike. You might be familiar with this if you’ve ever owned a thermos. Thermoses (the generic term being vacuum flask) consist of two metal containers, one inside the other, with most of the air sucked out of the space between them. Without air, heat can’t easily defuse through the walls of the flask, keeping heat in or out for an extended period of time (heat can still defuse in through the lid, which necessarily connects to the outer layer). This is also why the Apollo astronauts didn’t die on the Moon despite the lunar daytime temperature being an average of 120℃ (250℉). With no air between the astronauts and the superhot lunar surface, the only way for heat to defuse into their bodies was through their boots and occasionally their gloves, which are easier to insulate than their whole bodies.
There is a way to lose heat in the vacuum of space; blackbody radiation. All objects warmer than absolute zero (no molecular movement whatsoever) give off some amount of light. Vibrating molecules will randomly give off photons of light, and in doing so will vibrate slightly slower as they give a little bit of their energy to make said photon. The hotter the molecules, the higher the frequency of the light it can give off. (More on frequencies of light here) The human body glows with infrared light, too low a frequency to be visible to the naked eye, but enough to make you visible on a thermal camera. Once an object gets up to roughly 500℃ (~1000℉), it begins to give off visible light, glowing red hot. As it gets hotter, it glows not just with the lowest frequency of visible light, but all frequencies of light on the visible spectrum, becoming white hot. If an object reaches the temperature of the Sun, it gives off the whole electromagnetic spectrum from radio waves to gamma rays. Without air, spacecraft get rid of their waste heat using radiator panels. Heat pumps* (like those in a refrigerator) are used to pump the heat of the spacecraft (be it from electronics, human bodies, or heat from the Sun) into these large, thin panels. The most efficient shape at radiating away waste heat is one with a high surface area relative to volume, so the heat of these extremely thin panels efficiently radiates away into space as infrared light.
Image of the International Space Station’s radiators (the white panels). Notice how they are faced perpendicular to the solar panels (black). This prevents sunlight hitting them and giving them additional heat to radiate away.
While thermal radiation can cool an object in space, it works far slower than conduction or convection. Depending on the materials, temperature, and surface area of an object, radiative cooling can take two to three times longer than natural convective cooling (letting a warm object sit in still air) and hundreds of times longer than forced convection (using fans to blow air across a warm object, as electronics tend to use). Making matters worse, radiative cooling gets less efficient the cooler an object is, so a GPU that’s only 60℃ is going to radiate that heat away very slowly. Radiators would speed this process up, but remember that satellites and space stations like the ISS don’t have the heavy duty GPUs and electronics that produce the amounts of heat a data center produces. If one were to take the average large data center today and put it in space, it would require 300,000 square meters of radiator panels to regulate its temperature, or about 37 ISSs per data center. Now, the proponents of this idea do seem to be aware that space isn’t cold, and adding radiators to these data centers is part of their plan. I’ve even seen arguments that the vacuum of space is useful for cooling because it can act as an enormous heat sink, meaning that once you’ve gotten the waste heat out of the data center, there’s so much empty room for the heat to go that there’s no chance that it will return to the data center. While this is true, it doesn’t come close to offsetting the effort it takes to expel the heat in the first place. I have no doubt that a space data center with an effective cooling system can be built, but I don’t believe it will ever be cheaper to cool something in orbit than cooling it on Earth. While space data processing might be cost effective for data used by satellites (such as calculating collision prevention maneuvers or processing data from observation satellites where sending the data to Earth becomes a bottleneck), putting our terrestrial digital infrastructure in space would take one of its biggest problems and make it far, far worse. It doesn’t make economic sense to make one of the problems with data centers so much worse unless you can make all the other problems (power, land use) significantly better. And as we’ll see, those problems also aren’t significantly improved by being in space.
On top of the concept-breaking flaw with thermal management, there are a few other engineering problems with space-based data centers (SDCs). One of the big selling points of this idea is that SDCs have access to limitless power from the Sun. This is…technically true, but it understates the logistical hurdles. Solar panels are an effective source of electricity, but you need a lot of them to power something like a data center. The average GPU requires 700 watts of power and a large data center can have up to 100,000 GPUs. The ISS has the largest solar panel array ever launched into space, and it produces 250,000 watts. That means a large data center would require 280 ISSs to power itself. Electronics are also very sensitive to radiation, especially computer systems since small electronic faults can cascade out into systemic failure. Being outside Earth’s atmosphere would expose these data centers to far more radiation than they would on the ground, and having so many interconnected processors would make SDCs especially sensitive to these cascading failures. Also, SDCs could only have wireless connections to anything else on the internet. While you might not notice that wired connections are faster than wireless connections in your day-to-day life (both systems can handle normal internet traffic in less than a second), you might notice a difference if you’re dealing with petabytes of data. And if all of our data centers are in space, that means most of the infrastructure of the internet is on wireless, creating a lot of opportunities for chokepoints. Add on top of that, all this infrastructure (remember, a few hundred ISSs per data center) would have to be launched into space. It costs 1,500 USD to launch a kilogram of weight into Low Earth Orbit ($600 per pound). Not to mention the launches you’d need to service and repair this equipment; the average data center has to replace between 4-8% of its GPUs per year. Again, none of these problems are as big as the cooling problem, but there is clearly no technical advantage to be found in space that’s enough to offset the drawbacks.
Practical Commercial Uses of Space
Mentioning solar panels for SDCs reminded me of another proposal for putting human infrastructure in space. Since the 1970s, there have been several proposals for space-based solar power, building enormous solar panel satellites and beaming the power they generate to Earth. The atmosphere blocks a fair amount of the Sun’s light, and that’s not counting clouds, dust, and weather. The maximum amount of energy a solar panel in Earth’s orbit could collect 144% as much energy as the maximum amount on the ground could collect. There are also orbits one could put a solar satellite into where it would receive 24/7 direct sunlight. Some very optimistic estimates predict that space-solar panels could produce 40 times the electricity that a similarly-sized Earth-based panel would, though more conservative estimates put this at three times the energy. To transfer this power to the ground, these satellites would use a scaled up version of the wireless charging technology that some cellphones and other electronics use today. The satellites would use the electricity they produce to power a microwave laser aimed at a rectenna on the ground that is connected to the power grid. In theory, this could provide for the energy needs of humanity without polluting or even taking up space on Earth.
In practice, there are a few reasons we haven’t done this yet. Let’s be generous and take our optimistic 40-times-the-electricity statistic from before, and assume that one solar panel in space will produce as much power as 40 on Earth. The average residential solar panel weighs about 20 kilograms, so it would cost $30,000 to get it into orbit. That means that if you can manufacture 39 more solar panels for less than $30,000, you’re losing money on the installation alone. And again, that was our optimistic estimate; more realistically, you’d only need three times the ground-based panels to collect as much power as an orbital panel. Whether it’s for data centers or commercial power, space-based solar panels just aren’t that much better than Earth-based panels. While less efficient, Earth-based solar arrays are far easier to build, maintain, repair, and wouldn’t have to contend with damage from radiation and space debris. Also, while space solar farms won’t have the same thermal issues as space data centers since they’re not actively generating heat, they would pick up a lot of heat from being actively pointed at the Sun and couldn’t be air cooled like an Earth-based panel. Not to mention that the technology to beam power to Earth is still very much being developed. The current record for long-distance power transfer is 8.6 kilometers (5.3 miles), and they lost 80% of the power en route. As it stands, this is not yet a feasible way to power the planet.
But that’s not the only proposal for using the unique environment of space to aid in human commercial uses. Today, there are several companies looking into manufacturing goods in Earth orbit, particularly where gravity can be a hindrance to manufacturing. In microgravity, mixtures of liquids and gasses remain much more homogenous since nothing can float or sink (imagine ice in a glass, but without gravity, the ice doesn’t float to the top of the glass, it remains evenly distributed throughout the glass). Without gravity, heat doesn’t cause convection (remember that from above), so the solution is more still with less circulation. And of course, no gravity means there’s nothing to flatten solid shapes as they form. One process that benefits from this is crystallization, that is taking a dissolved substance and carefully undissolving it to create a highly ordered geometric structure. I’ve talked before about the ISS growing high quality protein crystals in zero gravity and how these crystals have been a boon for drug research**. Drugs can also be made from microscopic protein crystals, creating dense bundles of the drug (meaning smaller doses) that slowly release in the body as the crystal dissolves. Drug crystals grown in microgravity can be larger and more evenly-sized, leading to a higher quality product. Another industrial crystal we use is semiconductors, materials with a changeable electrical conductivity that are a fundamental part of modern electronics. Not only could higher quality semiconductor crystals be grown in microgravity, but materials like gallium arsenide which are normally difficult to crystallize can be grown far more easily into higher quality semiconductors, both due to microgravity and due to hard vacuum. And since crystals grow slower in microgravity, manufacturing that wants to avoid crystallization can benefit too. There’s currently a module on the ISS experimenting with making ZBLAN fiber, an extremely effective kind of optic fiber that is normally very hard to make on Earth due to atoms bundling up near the bottom. Without a bottom, these fibers lose signal up to 100 times more slowly than standard silicon fiber optics.
ZBLAN fiber made in microgravity (left) compared to ZBLAN fiber made in Earth’s gravity (right). Remember, fiber optic cable works by bouncing light pulses through a reflective glass wire, so you want them to be as transparent as possible.
Of course, all of this depends on it being economical to send the manufacturing equipment and the raw materials into orbit. With the development of modern reusable rockets, launch costs have dropped significantly in the past decade, which is why companies have started looking into for-profit orbital manufacturing. But with the least expensive rockets still taking $1,500 per kilogram, we are nowhere near this being profitable. There’s a reason why one of the big pitches for products that can be manufactured in space is fiber optic cable; anyone who’s ever had to buy quality fiber optic cable will tell you it’s expensive. For space manufacturing, you want products that are expensive for their weight. There is also research into growing high-quality gallium arsenide on Earth from seed crystals grown in space. If this works, it could give the advantages of space-manufactured semiconductors with only a tiny portion of the product’s mass having to go to space. And the cost of rocket launches is still coming down. When it becomes low enough for these projects to become feasible is anyone’s guess, but this is probably the most realistic area of study I’ve talked about here today.
One thing that could make space manufacturing even more feasible is if we didn’t have to launch the raw materials up. This gets us to perhaps the most well-known concept in commercial space use; resource extraction. Mining the Moon or asteroids for usable materials. Space agencies have been researching in situ resource utilization for decades, looking for ways to use the materials found in space to lengthen missions. Currently, the most sought after resources are those that would be used for exploratory missions; water for drinking, growing food, or splitting into rocket propellent or oxygen for breathing. Building habitats out of local materials is also being looked into, from simply burying Earth-made modules in lunar dirt as a radiation shield to ways to make something akin to concrete from lunar regolith. Being able to do this means fewer materials have to be brought from Earth (thus saving on rocket launch costs) and lengthening the duration of missions as food and spare parts can be made on site. But there is some crossover with commercial considerations. One area of investigation is building solar panels from lunar resources, as a moon mission would require solar panels and the lunar regolith contains a lot of silicon and aluminum. If a moon colony can get good enough at building solar panels on the moon, our space-based solar power idea from earlier starts to look a bit more feasible. And there are elements that are more plentiful in space than on Earth, from fusion fuels like helium-3 (which I’ve talked about before) to rare earth elements, which are incredibly critical to modern electronics but are rare on Earth in mineable quantities. When Earth formed, these super heavy elements sank into the planet’s core and accumulated well outside our reach. Asteroids don’t have that problem. Now, we are still very far away from it being more economical to get these elements from space than from Earth-based mines, but research into how to build and mine in space could one day make it possible for our research bases to send stuff home for us.
Obstacles
All of this is ignoring the biggest elephant in the room; regulating space-based industry. From space data centers to orbital factories, all of these ideas involve putting a lot more stuff into Earth’s orbit. Currently, there are over 40,000 pieces of large space debris in orbit, meaning derelict satellites, empty rocket stages, or lost equipment. All of these are being continuously tracked from the ground, but those are just the pieces big enough to see. When large debris gets impacted by micrometeorites or by other pieces of space debris, they break apart into smaller debris that are harder to track. Current estimates are that there could be over 140 million bits of space debris in orbit. And these debris are all moving at least 25 times the speed of sound. We’ve already seen spacecraft and satellites being damaged by debris strikes, and satellites are spending more and more fuel just dodging debris. What’s worse, when satellites get severely damaged or destroyed, they create new space debris. The worst case scenario would be what’s called Kessler syndrome, where a cascade of debris destroys most satellite infrastructure and renders certain altitudes unusable for the long-term (this is the plot of the film Gravity, though the cascade is sped up for dramatic effect). We’re already at a point where tracking and avoiding everything is becoming difficult, and adding more satellites would make this even harder. There are proposals for how to clear space debris, but that has the same problem that pollution on Earth has; there’s no profit to be made in cleaning and maintaining a shared resource while there is profit to be made in damaging it.
On top of space pollution, there’s the more traditional kinds of pollution to worry about. Rocket launches produce a lot of carbon emissions. The most common liquid rocket fuels are cryogenic liquid hydrogen, liquid methane, and kerosene. Burning methane and kerosene obviously produces CO2, but burning hydrogen should only produce water vapor. That is until you consider that the most cost-effective way to produce liquid hydrogen currently involves running methane and high temperature steam over a catalyst, which produces hydrogen and CO2. You can produce hydrogen by running an electrical current through water, but that’s more expensive. Also, while water vapor normally isn’t bad, adding it to the upper atmosphere as you would with a rocket launch can affect cloud formation and weather in the areas immediately around the launch site. Additionally, solid rocket fuels*** are often made from chemicals that produce highly reactive products such as chlorine and hydrochloric acid. Adding these to the upper atmosphere can damage the ozone layer (which I’ve talked about before), potentially undoing all the good work of the Montreal Protocols. There are solid fuels being researched that won’t have this problem, but they are still under research. As of now, the pollution created by rockets pales in comparison to other sources of greenhouse gasses. But that would change if we started launching rockets as frequently as we would to build anything I’ve described here.
My Thoughts - A Reluctant Commentary
This is going to be the most overtly political I hope to ever be on this blog. I try to present only the facts here, or at least only present my opinion when it is based on a thorough understanding of the facts and is useful to do so. I also normally don’t debunk pseudoscience, conspiracy theories, or similar bad ideas on this blog, even when they’re put forth by powerful people, because I feel doing so gives these ideas too much legitimacy. Yet here’s a concept based on a faulty understanding of science that I felt a burning desire to talk about. Some of that is because I thought it would lead to an interesting topic to talk about (‘space is cold’ is such a common myth), but I also think there’s a degree of public misconception about the people pitching this idea and their motivations, not just about the science. I feel that not discussing these misconceptions would mean leaving important information out, both about the topic itself and why I felt it was important to talk about, but doing so necessarily means sharing some of my own worldview. So forgive me if this starts to feel too much like a rant, but I believe I speak for many of my age and station on these matters.
Since I was a child, I’ve been a science enthusiast. I love to know how the world works and how this knowledge can be used to make our lives better. As such, I’ve loved learning about space and have wanted to see a revival of interest and investment in it. Artemis II has stirred that excitement in me again, as we begin the steps to return to the Moon after a half century. So space data centers are quite thematic to me, because AI and the renewed interest in space are my two go-to examples of that childhood science enthusiasm colliding with my adult cynicism. Artificial intelligence has so many practical applications that are actually useful and interesting, from revolutionizing our understanding of protein folding to reverse engineering how the human brain works. But the only thing the companies that build AI seem interested in is large language models trained on questionably collected data with the motivation of replacing human workers. The promise of automation, especially in science fiction, has been that we would have machines do the backbreaking manual labor and unfulfilling drudge work for us, leaving humanity with more leisure time and only work that we find meaningful. I want that to happen, I think it could create a more equitable and more pleasant society. But the burgeoning AI age has shown that the CEOs controlling this technology want to use it to do away with the more meaningful work in the arts and sciences because those workers are the most expensive, and see the same work being done in less time as a reason to raise their expectations of workers instead of giving them more free time. And the continued growing investment into AI in spite of public disapproval serves to highlight how little the betterment of society factors into these CEO’s vision for this technology.
Likewise, I want to see a renewed space race. Space travel leads to the development of new technologies (here’s a list) and could one day lead to us becoming an interplanetary society. Plus, it’s just cool. But so much of the reasoning space companies have for wanting to go to space seem to revolve around maintaining control and dodging responsibility. Another problem with space data centers I wanted to wait to talk about is the issue of legal liability and law enforcement. A revolution in space commercialization means that much of the global economy would be occurring in international “waters”, which I suspect is considered a benefit to the companies that wish to do this work. I think it’s telling that when the companies proposing space data centers pitch the idea, the specific wording on the availability of space is usually some form of “no property tax or land-use regulation.” It’s a sentiment that leads one to worry that the people who want to make outer space colonies actually want to make company towns outside of legal jurisdiction. Additionally, there’s a concept in economics called the resource curse, which describes how nations whose economies primarily come from mineral wealth tend to have authoritarian governments that can remain in power for decades. Since the state needs very few of its citizens' involvement to make exorbitant profits, they can be easily cut out of the political process (think Saudi Arabia or the DRC). Now obviously, lunar and asteroid mining won’t be quite like terrestrial oil and mineral wealth since just getting into space requires extremely skilled labor. But since much of the actual mining itself would have to be automated, a larger cut of the profits would be going to the investors. The fewer people that are involved in bringing a resource to market, the less the resulting economic growth actually benefits the whole of society. There’s a common sentiment among futurists that asteroid mining will create the world’s first trillionaires****. We already have people wealthy enough to run presidential campaigns just to block candidates they don’t like or to buy communications platforms because they dislike what they say. I want to believe that these technologies will lead to a better world, but the people who control these technologies have proved to be untrustworthy. If you want proof that they aren’t fit to lead, a number of them earnestly proposed putting data centers in space.
I abhor the notion of being a luddite. Science and technology has had such an immense net positive impact on our species. Easier access to space and the projects I’ve described here could ultimately lead to humanity having access to far greater living space, resources, and energy, which could lead to far greater population with greater education and quality of life for all people, which leads to the kinds of societies that are most nice to be a part of. I don’t want to deny humanity something good just because of what downsides might come from them. But I feel as though I live in an age where anything with the ability to empower humanity is instead being used to empower a select few in their goal to disempower everyone else. I’m sure this has always been true, but I hate seeing a field I love being used for inequitable ends. Even the cause for optimism I see is a cynical one, that being that all these endeavors are likely to ultimately fail. We’re already seeing the collapse of the AI bubble, as the companies that run AIs are still not turning profits and the tokens to use these AIs are proving to be no cheaper than hiring a human workforce. Like many people, I suspect their pitching space data centers as a way to keep investors interested in the technology because it’s still dependent on investor money. My hope is that the bubble bursting will create room for actually good uses of the technology. Likewise, I personally doubt we’re on the brink of a coming revolution in space commercialization, at least in the upcoming decade. The costs of rocket launches are still too high to be profitable, as reusable rockets can’t really fix the problem that you need 4,500 tonnes of fuel to get said rocket into orbit. And there are other futurists who question the “world’s first trillionaire” sentiment, thinking it’s more likely that asteroid mining would mean flooding the rare earth metals market. While I do believe space manufacturing and resources will come eventually, it will likely not be done by men who demand an immediate return on investment. Again, the people trying to use this technology to make the world worse are short-sighted enough to think putting data centers in space is a good idea. The Artemis missions are a collaboration between multiple national space agencies and companies, which means costs are split between multiple entities, but also the self-serving interests of millions of people cancel each other out, leaving only the single common desire to explore our universe. I hope to one day live in a world where my first question for any new technology won’t be “how is this going to be abused.” I have hope that such a world is possible because I know the economic and political systems that try to abuse these technologies are far more transient than the innate human desire to explore and experience this universe of ours.
For More Details
*In case this is helpful, heat pumps are mechanisms that move heat energy from one place to another. You see, when a liquid or gas is compressed, its temperature rises as the same amount of thermal energy is packed more tightly. Likewise, decompressing a liquid or gas causes its temperature to drop. Heat pumps use a compressor to compress a working fluid, causing it to heat up. Since the fluid is now hotter than the surrounding environment, heat defuses out. Once the fluid is back down to room temperature, it is pumped to another environment and is decompressed so it cools down. Now colder than the surrounding environment, it absorbs heat to return to room temperature. This is how a refrigerator works (the cold fluid absorbs heat from inside the insulated fridge and releases it into the surrounding air from the radiator on the back) and is how spacecraft thermal controls work (the cold fluid absorbs heat from the spacecraft interior and releases it into the radiator). Keep in mind that heat pumps require powerful, constantly running electric motors, so they produce a bunch of heat themselves that then needs to be expelled. This potential vicious cycle is part of why heat management for spacecraft is so difficult.
**In short, protein crystallography is a process to determine the structure of proteins. A pure solution of a protein is created and these proteins are made to join together into a highly ordered structure, namely a crystal. Because crystals are so ordered in their structure, the shape of the crystal can tell you about the shape of the molecules making up the crystal. By bombarding these crystals with x-rays and measuring how they refract, computers can build up a model of the crystal’s structure and by extent the protein’s shape.
***Solid rocket fuels are usually used for solid rocket boosters (those two white pylons on the side of the space shuttle or SLS). Solid fuels are extremely powerful, but they have the drawback of being impossible to turn off once they’re ignited (think really big firecracker). This combination makes them useful for the initial launch. Liquid fuels are preferable for the remainder of the flight, where you want to be able to control the amount of thrust.
****While I wrote this, Elon Musk was very briefly the world’s first trillionaire, due to a SpaceX IPO offering largely based around funding space-based data centers. He then lost this status after the value of SpaceX stock was reassessed. This drama aside, I believe my point about asteroid mining potentially worsening wealth inequality still stands.
Comments