Star Cloud (formerly LumenOrbit) - Data Centers in Space

[catdlr]

https://x.com/LumenOrbit/status/1832157234207621465

Lumen Orbit is building data centers in space to use 24/7 solar energy and passive cooling.

White Paper (link and pdf attached):  https://lumenorbit.github.io/wp.pdf

https://www.ycombinator.com/launches/LmD-lumen-orbit-data-centers-in-space

[trimeta]

Ah yes, passive radiative cooling, widely regarded as being far more efficient than the conductive or (especially) convective cooling you're forced to deal with for Earthbound data centers. Why have a complex fan blow across your heatsink fins, when you can simply make those fins an order of magnitude bigger and probably include internal heat distribution channels to make use of the larger surface area?

"Doesn't need a fan!" is basically the same as the famous advertising slogan for Ginsu knives, "never needs sharpening!"

[jstrotha0975]

Can somebody explain the pros of putting a data center in orbit because I'm having a hard time finding any.

[launchwatcher]

Can somebody explain the pros of putting a data center in orbit because I'm having a hard time finding any.

I'm somewhat familiar with the challenges around datacenters at the tens-of-megawatt scale.   Power availability drives the location of large datacenters on earth.

Their justification seems to start, and end, with "free power" - as almost all of the savings are power related ($140m in power bills, $20m in backup generator capital cost per 40MW). 

At least they're honest that getting the cooling right remains the biggest unsolved problem:

This component represents the most significant technical challenge required to realize hyperscale space data centers.

Their intended use case (batch jobs training AI models) can at least cope with the network latency and locality issues in orbit better than most - the white paper even shows docking ports for one or more "data shuttle" spacecraft showing that they are targeting latency-insensitive workloads.

I remain highly skeptical but this is at least a not completely ridiculous use case.

[markbike528cbx]

Can somebody explain the pros of putting a data center in orbit because I'm having a hard time finding any.

Se the claims https://www.lumenorbit.com/

Per https://www.geekwire.com/2024/lumen-orbit-stealth-2-4m-data-centers-space/

Lumen’s business plan calls for deploying about 300 satellites in very low Earth orbit, at an altitude of about 315 kilometers (195 miles). 

That seems a little too low to me.
If the can keep the plane of the solar panels exactly in the orbital plane, then a minimum of 4m^2 (4km x1mm) of frontal drag is present . Any deviation and the area would go up as the sin of the angle x 16km^2

[HMXHMX]

The White Paper doesn't seem to be available; anyone have a link or a PDF?

[catdlr]

The White Paper doesn't seem to be available; anyone have a link or a PDF?

Here's both: (pdf attached)

 https://lumenorbit.github.io/wp.pdf

[launchwatcher]

The White Paper doesn't seem to be available; anyone have a link or a PDF?

It's titled "Why we should train AI in space" and lists Ezra Feilden, Adi Oltean, and Philip Johnston as authors.

I found it on their home page at https://www.lumenorbit.com ; scroll down until you see the "View our White Paper" button.

[Vultur]

One argument for this is to reduce the environmental impact of AI training data centers by taking them off the existing power grid

[catdlr]

One argument for this is to reduce the environmental impact of AI training data centers by taking them off the existing power grid

Same for Bit Coin data mining.

[Vultur]

Yeah. I am not totally convinced of the specific application -- while there is ridiculous money being thrown at AI training right now, that might not last -- but the idea of moving high energy tech applications to space is a good one, I think.

[Asteroza]

One argument for this is to reduce the environmental impact of AI training data centers by taking them off the existing power grid

More specifically, if a corporation subscribes to ESG or SDG's green goals, then getting both power and heat off planet would be an end goal. An activist investor could barrage AI companies to obligate them to preferentially purchase orbital compute capacity over terrestrial.

But, the more proximate issue is datacenter licensing/building (power availability in terms of transmission capacity and local grid capacity, thermal exhaust limits in the form of cooling water temperature).

Contrast to someone building a power and fiber mooring turret offshore beyond the 12 mile limit from a powerplant and parking a turboelectric containership full of containerized datacenter modules. That's a proximate solution if terrestrial datacenter capacity gets tight (and locates compute close to population centers). A natural evolution of this is an offshore OTEC power facility with a colocated pressure stabilized platform to host the containerized datacenter modules (which is functionally close to Seasteading...).


Regardless, the satellites themselves would probably cluster in terminator riding low SSO polar orbits to ease power and cooling issues I would think, so if this takes off then there is an interesting megaconstellation issue due to extremely high density of satellites in a specific polar orbit as there is no real plane distribution to help reduce conjunctions. The alternative is higher orbits to improve spacing but that cuts into upmass though.

[Asteroza]

While they seem to show solar PV, in this specific application, a solar thermal design might be more suitable?

Many SPS designs try to distribute the thermal issues by using distributed parts almost joined to the solar PV backside in the aim of locally rejecting heat by having low thermal load density.

If you are concentrating compute into tightly packed modules though, the high thermal load density would pressure you to more substantial thermal management.

If you had a solar thermal SPS design as a basis though, then by definition you have a stupidly large radiator available. I think some of those designs were baselining steam or ammonia cycles, so using datacenter heat as a feedwater heater is an interesting premise.

[catdlr]

Scott puts his comments on this plan

Sep 9, 2024
Lumen Orbit raised 2.4 million dollars from investors based on the notion of putting data centers in space, computers that could be used to train AI, or maybe mine scam. Their argument is that at low enough launch costs the ability to get 24/7 solar power makes power much cheaper to the point that it's worth the cost of the launch:

[AndrewM]

Some updates since September.

TechCrunch reported they had a $11M seed round and are valued at ~$40M.

https://techcrunch.com/2024/10/24/lumen-orbit-closed-one-of-the-biggest-rounds-from-y-combinators-last-cohort/ [Oct 24]

Lumen Orbit has closed an oversubscribed, eight-figure seed round of more than $10 million, a source familiar with the details told TechCrunch. That would make it one of the hottest deals, if not the hottest deal, of the most recent Y Combinator batch.

The company went through YC’s 2024 summer batch and garnered a significant amount of attention from VCs, multiple VCs told TechCrunch. This interest led to an extremely competitive deal process for the startup’s seed round.

While Lumen has a lofty mission, the company seems to be making notable progress already. It was founded earlier this year and is planning to launch its demonstrator satellite in 2025 in partnership with Nvidia’s Inception program.

https://techcrunch.com/2024/12/11/200-vcs-wanted-to-get-into-lumen-orbits-11m-seed-round/ [Dec 11]

The Redmond, Washington-based company closed on an $11 million seed round at a $40 million valuation, confirming prior TechCrunch reporting that the company had raised a competitive double-digit round as one of the buzziest startups out of Y Combinator’s Summer 2024 batch. The deal was led by NFX — NFX general partner Morgan Beller will join the company’s board — with participation from VCs, including Fuse.VC, Soma Capital, and scout funds from Andreessen Horowitz and Sequoia, among others.

Lumen Orbit co-founder and CEO Philip Johnston told TechCrunch that due to the high investor demand — more than 200 hundred VCs reached out to the startup — the company has since opened up another SAFE round on top of it, at a higher valuation, to let more investors in.

Spacecraft update:
https://twitter.com/starcloud_inc1/status/1860380150975856878 [Nov 23]

Final stretch on the assembly of our Lumen-1 satellite ready for launch next year 💪🏃🏼‍♂️‍➡️🚀 @LumenOrbit, @ezrafeilden, @adi__oltean, @johnstonphil

(They even let me near it! 😅)

They renamed from Lumen Orbit to Star Cloud.

https://twitter.com/starcloud_inc1/status/1888176705195401677 [Feb 8]

Lumen Orbit is now Star Cloud!

[catdlr]

Philip Johnston
@PhilipJohnst0n
Do you or someone you know want to join the most 𝗰𝗿𝗮𝗰𝗸𝗲𝗱 𝗲𝗻𝗴𝗶𝗻𝗲𝗲𝗿𝗶𝗻𝗴 𝘁𝗲𝗮𝗺 𝗼𝗻 𝘁𝗵𝗲 𝗽𝗹𝗮𝗻𝗲𝘁 (http://Starcloud.com/team)? We’re looking for the following:
- Electrical Engineer
- Power Electronics Engineer
- Spacecraft Software Engineer

Experience in the data center or space industry required. Green card holders only. In-person in Redmond WA only. Hardcore people on a mission only 🚀
Apply here http://Starcloud.com/careers

My DMs are open if you fit the criteria above.
@Starcloud_Inc1

https://x.com/PhilipJohnst0n/status/1962567954962559254

[catdlr]

It changes yet again. Collaboration:

Mission Space | Space Weather@mission_space
We’re proud to announce our strategic partnership with @Starcloud_Inc1! Together, we are shaping the future of orbital data centers—resilient, high-performance, and ready for the challenges of space!

Press Release:  Starcloud and Mission Space Forge Strategic Alliance to Integrate Orbital Datacenters with Next-Gen Space Weather Data


https://twitter.com/mission_space/status/1968667198538313924

[catdlr]

Philip Johnston@PhilipJohnst0n
·
Earlier today, Jeff Bezos said we will build “𝗚𝗶𝗮𝗻𝘁 𝗴𝗶𝗴𝗮𝘄𝗮𝘁𝘁 𝗱𝗮𝘁𝗮 𝗰𝗲𝗻𝘁𝗲𝗿𝘀 𝗶𝗻 𝘀𝗽𝗮𝗰𝗲”. It’s become obvious now. Need to get moving! 🚀
@Starcloud_Inc1, @JeffBezos

https://twitter.com/Starcloud_Inc1/status/1974075658470109581

[Tywin]

A prototype soon to flight:

https://twitter.com/PhilipJohnst0n/status/1974095427579584571

[catdlr]

Jeff Bezos Predicts We'll Have Gigawatt Data Centers in Space in 20 Years

PC Mag
Michael Kan
Senior Reporter