SpaceX and Google negotiating deal to launch data centers into space
Though few details have been confirmed, according to the Wall Street Journal SpaceX and Google are in advanced negotiations to launch data centers into space.
We don’t know if these data centers will be part of a SpaceX/Google partnership, or whether Google is merely negotiating a SpaceX launch deal to place its own data centers in orbit. Nor do we know if this deal will use SpaceX’s Falcon rockets, or is aimed at using Starship when operational. Neither would surprise me. Nor would it be surprising if both occur.
The story is in linked to SpaceX’s impending initial public stock offering (IPO), expected to the biggest in history.
On Christmas Eve 1968 three Americans became the first humans to visit another world. What they did to celebrate was unexpected and profound, and will be remembered throughout all human history. Genesis: the Story of Apollo 8, Robert Zimmerman's classic history of humanity's first journey to another world, tells that story, and it is now available as both an ebook and an audiobook, both with a foreword by Valerie Anders and a new introduction by Robert Zimmerman.
The print edition can be purchased at Amazon or from any other book seller. If you want an autographed copy the price is $60 for the hardback and $45 for the paperback, plus $8 shipping for each. Go here for purchasing details. The ebook is available everywhere for $5.99 (before discount) at amazon, or direct from my ebook publisher, ebookit. If you buy it from ebookit you don't support the big tech companies and the author gets a bigger cut much sooner.
The audiobook is also available at all these vendors, and is also free with a 30-day trial membership to Audible.
"Not simply about one mission, [Genesis] is also the history of America's quest for the moon... Zimmerman has done a masterful job of tying disparate events together into a solid account of one of America's greatest human triumphs."--San Antonio Express-News
Makes sense. Google/Alphabet is a major AI player and faces the same siting, power and regulatory limitations as every other company attempting major terrestrial AI data center projects. SpaceX has a way to dodge around all of those limitations and can certainly deploy capacity capable of handling the needs of others as well as its own. These talks tell the market that AI data centers in orbit can be profitable almost as soon as they are launched, just like Starlink but with an even quicker payoff. That will support a fat asking price for shares when the IPO occurs.
As Robert has pointed out, the Superheavy Booster was operational for single launches. No reusability yet, but if you want to launch a lot of tonnage, it would work. I realize the next launch is not the previous Super Heavy booster. I wonder if companies would ever ask / purchase from SpaceX a single use launch? Google certainly can afford it, and the orbiting data centers could be established.. Just wondering.
Will these orbital data centers have their own ground link systems or will they use Starlink?
I would think it would be cheaper to use starlinks.
Ronaldus Magnus,
Do you mean merely expending a vehicle rather than reuse, or a fresh booster that is only used once? If the former, B1060 is an example of a first stage that flew 20 times, with its last launch being a payload of Galileo satellites. If the latter, B1030 launched EchoStar 23 to geostationary orbit and was expended, B1034 launched Inmarsat-5 F4 to GEO and was also expended, and there have been others. With Block 5, however, it’s been the rare flight to see no attempt at landing a new booster at all-B1054 launching a GPS satellite is one example. I do not believe SpaceX has expended any F9s on their first flight since late 2024, but someone can correct me if I’m wrong. So there is precedent for what you ask, but I suspect it won’t happen very often as both orbital refueling and tug use expand. Data centers in the hundreds of megawatts or gigawatts will require many launches, so whether it will make sense to expend a booster or not is dependent on cost to reuse a stage versus building a new one, and what payload Starship can manage reused versus expended.
Ronaldus Magnus,
You seem misinformed anent Super Heavy reuse. Super Heavies have been caught three times and reused twice. Booster 14 launched Flight 7 and also Flight 9. Booster 15 launched Flight 8 and also Flight 11. Both were expended on their second flights because they were V2 configurations and would not be compatible with the upcoming V3 Starships.
You also appear to be laboring under the misapprehension that an orbiting AI data center will be a single satellite deployed on a single launch. Not so. An individual AI data center satellite will be a significant nugget of computational horsepower, but not compared to an entire major terrestrial data center with miles of server racks inside. AI data center sats will be deployed in a constellation. It is the constellation that will grow to be as powerful, and then more powerful, than any terrestrial data center. The individual sats will connect to their orbital neighbors via laser links, as is the case with Starlink birds, to form a single, connected, computational complex. SpaceX has applied to launch as many as a million such AI data center sats. Doing so will require thousands of Starship launches. Reuse of both the Boosters and the Ships will be essential to accomplishing this task economically.
Dick Eagleson,
I think that Ronaldus Magnus‘s assumption about large data center satellites comes from early reports of SpaceX’s expectation of a million-satellite constellation that uses a combined trillion watts, making for megawatt satellites.
Comparing that with the geostationary communication satellites that I worked on:
5 kW of power, 30 m^2 of solar arrays, 10 m^2 of radiative heat dissipation surface, 10 m^3 of satellite volume (including propellant tanks), 2 tonnes (including propellants). A megawatt satellite would be a good 200 times larger: 6,000 m^2 of solar arrays, 2,000 m^2 of radiative surface, 2,000(?) m^3 of satellite volume (may not need large propellant tanks for station keeping), 400 tonnes (less, if there is less propellant). I can see that the expectation would be a satellite that is a 10+ m cube with a square radiator almost 50 m on a side and a square set of solar arrays that are about 75 m on a side. That is pretty big. At least in our imagination. This would require somewhat more than the current standard Starship.
I’m pretty sure that the reality will be somewhat different. Smaller satellites, with each needing less than the calculated megawatt, and maybe more total satellites or less total computing power in the constellation.
Edward,
Here is a link to a SpaceX artist’s concept of what the company calls its “Mini” AI data center satellite. As you can see, it has a “wingspan” of about 600 feet and a “wing chord” of roughly 20 feet for roughly 12,000 sq. ft. of solar array area. Neither the body of the sat nor the radiator is nearly so big, proportionally, as you have suggested.
The total mass of such a bird would depend upon exactly what the nature of the solar array technology employed would be. The “wings” and radiators could be folded or rolled up and the sats stacked for deployment within a Starship payload bay as Starlink V3 sats will be, though the load-out for AI data center sats of this size will be rather lower in number. It might also require a slightly modified version of the “PEZ dispenser” Starships we have seen to this point with a taller dispenser slot door.
I’m unclear on whether this “Mini” design is to constitute the entirety of the Earth-orbiting AI data center constellation or whether a larger version is to take over at some point – perhaps to be deployed from a hammerhead Starship upper stage. I’m reasonably certain that the plan for the AI data center sats to be built on the Moon and deployed with the notional mass driver is for them to be at least ten times the size of this “Mini” design when fully unfolded.
Edward,
SpaceX has said the initial satellites would use 100kW, so I think that future satellites will either be much bigger, or SpaceX will never actually deploy a million of them. Probably both.
Gentlemen,
My “suggested” satellite was intended to be an assumed satellite due to the imagination that derived from previous statements as compared to current reality. What would it take for a million satellites to collect and use a combined terawatt of power?
As I said, “I’m pretty sure that the reality will be somewhat different.”
However, thank you for the artist’s rendering and comparison to the launch vehicle. It is a nice size for the “mini” version, but we do have to wonder what the “maxi” version would look like and just how large it could be.
The size of the solar arrays described by Dick is consistent with a 100kW satellite, as described by Nate.
Edward,
You’re right. As for larger satellites, perhaps something like Starcloud’s renderings on their website.