How do we get 1 TW a year of orbital compute by 2035?

- 50k Starship V4 launches a year, each carrying 200 t (20 MW of compute). That’s 137 a day, one every 2 minutes in two windows of roughly 5–7am and 5–7pm local. As Earth spins, the launch site passes under that target orbit twice a day, at local dawn and local dusk. Globally we’re at about 1 launch a day right now.

- The always-sunlit orbit needs launches heading north or south over open water. On US coasts only south works, and packing every usable spot with Starship pads gets us to just 25–30 launches a day. We need 5× that.

- So we will go out to sea, 300 km off central California. Inside the US EEZ, clear of Cape Mendocino, between the air routes to Hawaii and Asia. Open water both ways gives us both daily windows.

- We will place 35 floating pads in one 68 km row, so no rocket flies over another pad. The water is 4 km deep, so thrusters hold the pads in place.

- Need a huge daily supply: 190 kt of LNG, 680 kt of liquid oxygen, 27 kt of satellites. - About 7 LNG tankers and 10 liquid-oxygen tankers in constant rotation. - Roughly a container ship a day of satellites from LA, so 2–3 ships in rotation. - About 20 shuttles refilling pads between windows.

- The gas for all that is 2.5 times what all of California uses. The California Coastal Commission has really not been playing ball with such things so far.

- So we make it in Mexico. Baja’s Costa Azul terminal already turns US pipeline gas into LNG, so we expand that site with much bigger LNG production, a liquid-oxygen plant, new pipeline capacity from Texas to Baja and 20 GW of power. This would be one of the biggest natural gas sites on Earth.

Unexplored territory: - Terafab (we need at lesat 0.5M wafer starts per day, mostly memory. This is 10x more than TSMC makes today. maybe doable by 2035) - Starship V4 (scheduled for 2027) catching booster and ship at sea (Falcon boosters already do sea landings with legs) - large liquid-oxygen carriers (essentially an LNG carrier with oxidizer safety) - starship-scale floating pads (Sea Launch did this two decades ago. But much smaller rockets)

3 points | by chaisan 12 hours ago

2 comments

  • ggm 11 hours ago
    Nobody wants this. This is south Sea bubble mania, I cannot think of a single real world problem crying out for TW compute in spaaaaace.

    Seriously. Not since operation ploughshare has there been such a solid larding of "we could build it" and for why?

    At least Reagan's star wars tried to fling money at a problem people in hypothesis thought existed. But, this is just flinging money at problems which don't exist because we don't need the compute. We don't need it on the ground or in space. We need storage, and we need data. Hell, we need aws and cloudflare and gcp. But terrawatts of LEO compute? Sorry.. what's this for again?

    • jsLavaGoat 11 hours ago
      Kind of a strange example considering that our missile defenses are so good that our allies don't want us fighting when there aren't enough interceptors. What's the message, that in 40 years there will be only half as much compute in space as we thought?

      For what it's worth, I think people are deluded about the amount of things we can do in space. We are several unforeseen breakthroughs away from most of what people think is possible.

      • eesmith 10 hours ago
        Wrong kind of missile.

        SDI concerned ballistic nuclear missiles -- ICBMs. We have no good defenses against them. The modern incarnation of SDI is Golden Dome.

        The systems you are thinking of are for short- and medium-range missiles.

  • bradknowles 9 hours ago
    You don’t. It’s a dog whistle.

    Elmo can’t get what he wants on the ground, so he has to own all of space and put all of his toys up there. Whether they will work up there or not doesn’t matter, he’s just got to own all that space and make sure that no one else can launch anything.

    Then he hits Kessler Syndrome, but he doesn’t care, because he has successfully prevented anyone else from doing anything useful in space.

    The only question is how long is he going to be allowed to continue down this path before he gets permanently stopped?

    If that doesn’t happen soon enough, then we get Kessler Syndrome anyway, and then nothing matters.