[mike_hearn]

On Earth it has to be fossil fuels. With solar you can't do it. I haven't done the calculations myself so, full disclosure, AI work ahead. But I asked GPT 5.6 how big a solar and battery farm would have to be to power a one gigawatt datacenter such that it could ride out a few cloudy days in a row (this is far below the reliability bar).

Its answer is you'd need an "absurdly large battery" and a solar farm of over 100 square kilometers, quite possibly up to over 200 square kilometers. And it would need to be built in the desert.

Also, normal utility scale LiON battery sites are only meant to provide stabilization services for hours, not days.

The problem is you have to overspec the solar farm quite a lot because it must not only be able to power a one gigawatt facility, but also generate more than that so it can recharge the battery packs during the day too.

So this just doesn't work. You think datacenter permitting is bad when it only requires some on-site gas turbines? Imagine the difficulty of getting a 200 square kilometer solar farm over the finishing line.

No, it's quite possible Elon is going to win this one. Regulation is expensive! He can just fire a constant stream of satellites into space. The constellation needed to provide a gigawatt of inferencing would be about the same size as today's Starlink system, well within the bounds of what he's proven can be done.

The threat to space based datacenters isn't really the physics of it. It's the risk of a bubble pop that temporarily craters demand and/or an overbuild of terrestrial capacity. Outside of the software industry we don't see huge new blocks of inferencing demand, and OpenAI at least doesn't seem capacity constrained at all right now.

[bronson]

Why restrict yourself to 100% solar? Solar augmented by grid is pretty great.

Also, you very conspicuously avoided talking about the difficulty/impact of a 1GW datacenter in space. One might suspect your motives aren't as impartial as you seem to think.

[mike_hearn]

The whole problem is difficulty getting grid connections on Earth. If you can just hook up to the grid then why not just use it for everything.

I don't have any motives, what are you talking about? I don't own SpaceX stock if that's what you're thinking.

What's the difficulty of a 1GW datacenter in space? The number of satellites it requires seems to be within the realm of what SpaceX has already achieved. If the calculations are wrong, by all means say so, I haven't checked them.

[rented_mule]

> What's the difficulty of a 1GW datacenter in space?

Simple back of the envelope calculations?

Let's assume your numbers are correct and that panels in space can generate 10x the power per square meter per 24 hour period (due to efficiency, lack of night, and lack of atmosphere/weather). That means 10 KM^2 are needed for one such data center. The largest we've built in space is ~3,000 M^2 on the ISS. So one of these DCs in space will require an array ~3,000 times the size of the largest we've built before. Or, it's going to require ~3,000 satellites, each with an array the size of the largest we've ever deployed.

The largest Starlink satellites currently deployed generate less than 30 KW each. We'd need more than 30,000 of those to generate this much power. Starlink has launched ~12,500 to date, with ~10,000 still functioning. So 3x the size of the functioning fleet that's taken 8+ years to deploy. Currently, we are deploying well under 5,000 per year. Let's assume we can come up with enough spare capacity to launch 5,000 per year. That's 6 years to deploy the first 1GW DC. By the time we get to 1GW, the average GPU is 3 years old. So we need to continue to deploy 5,000 satellites per year to maintain 1GW of GPUs that are, on average, 3 years behind current generation GPU designs.

This is all before we get to the economics of launching, the cost of the satellites themselves, heat dissipation, radiation hardening, hardware failure rates (~20% of deployed Starlink satellites are no longer functioning), etc.

[mike_hearn]

The whole thing does assume Starship works, so at that point satellites could be launched much faster.

I assume Starmind would use Tesla designed chips, which removes Nvidia's margin and so the cost of the chips is not necessarily so high.

Rad hardening isn't important at low altitudes, and inferencing can smoothly recover from glitched calculations/lockups.

I don't know if it makes economic sense or ever will, but technologically it does seem possible. And if you can't build on Earth for political reasons, then space can win by default.

[rented_mule]

In terms of what we decide to build, impossible vs. impractical is not a particularly interesting distinction (that distinction can give rise to fun science fiction). Practicality is at the core of good engineering, and possibility is just the first of many steps in evaluating practicality.

If we're going with a distributed approach, there are other ideas that are more practical. For example, send 1 KW boxes to a million people (< 1% of American homes, but there's no reason to limit this to the US). Very roughly, these are beefed-up gaming consoles - you could even start with existing gaming consoles. That box plugs into electricity and ethernet. Pay each of those people $1,000 a month to leave their box running 24/7. For people with a lot of solar at home, it could be a nice income stream. At $1B per year for payouts, it's far cheaper than doing it in orbit. And it doesn't rely on promised but as-yet undelivered technologies. But data centers are probably more practical.

[mike_hearn]

I've thought of that but it wouldn't be easier.

1. Running modern models requires a whole rack of extremely loud equipment that can easily consume 130kw. That's not a load you can or should run at home.

2. It would require a very strong end-to-end implementation of confidential computing to keep the prompts encrypted. If there is any flaw in that infrastructure at all, like someone discovers a way to physically tamper with the hardware to read the GPU memory, then the entire network is a writeoff and would have to be replaced. That's a massive risk. Confidential computing can work in cases where hardware replacements aren't an issue or where the vendor has a long history of successfully defending the platform. But Nvidia is the only GPU vendor with these capabilities and their CC is untrusted.

Putting the workloads in space means noise and data privacy aren't an issue, assuming you trust SpaceX to not peek at your prompts on their way to the cluster.

[AnthonyMouse]

> The whole problem is difficulty getting grid connections on Earth. If you can just hook up to the grid then why not just use it for everything.

Because you already need inverters and batteries for backup power and then using solar + batteries for ordinary operations and the grid for backup is cheaper than the other way around.

Which is also why you may not even need the grid connection. Suppose you have diesel generators to use in case you have some kind of a fault in the solar system, the same as was traditionally done in case of a failure of a grid connection, and then use those for the one week every two years when solar generation is too low.

[bronson]

> What's the difficulty of a 1GW datacenter in space?

Why don't you ask your AI, the way you did with your earth-based scenario?

[abirch]

How do you cool the data center in space cost effectively? Unfortunately in space there isn't a lot of material (air) to transfer the heat.

[mike_hearn]

You can just use ordinary radiators. Do the calculations, cooling isn't the problem it's made out to be.

[abirch]

It's the size of the radiators and the expense of getting them up into orbit (along with maintenance)

Most of my assumptions come from discussions like: https://news.ycombinator.com/item?id=46876105

[sokka_h2otribe]

(1) these are not ordinary radiators because ordinary radiators aren't gold plated and don't unfold dramatically from compact shipping sizes (2) It's absolutely loads of radiators

[bryanlarsen]

100 square kilometers sounds like a big area. And it is, in continental Europe or in an urban area. In Montana, it's enough space for a few hundred cows. In desert regions, it's worth even less. You can buy 100 square kilometres of unserviced desert land for ~$10M.

[mgerdts]

Even better, get rid of the mandate for ethanol and cover a portion of the ex corn fields with solar panels with native plants under and around them. A small portion of the land that is near existing transmission lines is needed to replace combustion vehicles with electric vehicles. The bulk of the rest of this land could be available for other uses, including data centers.

This gives: sustained income for landowners, CO2 reduction, power for new uses, and partially restored ecosystems that have been greatly harmed by chemical laden monocultures.