First steps toward a Dyson Swarm

[John Clark]

Data Centers in Space Could Launch a New Space Economy

John K Clark    See what's on my new list at  Extropolis

[Keith Lofstrom]

On Fri, Oct 31, 2025 at 11:27:53AM -0400, John Clark wrote:
> *Data Centers in Space Could Launch a New Space Economy*
> <https://www.youtube.com/watch?v=iLNrYwx0th0>

I published papers about space data centers before 2009,
see http://server-sky.com

I'm a semi-retired chip designer. I certainly don't know
everything, but I have a pretty good handle on What Ain't.

I live near the Intel D1X chip fab (the local IEEE solid
state circuits chapter meets in their design center) and
have friends who are wafer-slingers in nearby TSMC fabs.
Those buildings aren't as tall as the Vertical Assembly
Building at KSC, but they are more expensive, containing
tens of billions of dollars of expensive equipment from
more than a thousand other companies around the world.

Some of those machines are VERY gee-sensitive; they are
brought from Europe and Japan in modified 747s which are
essentially piloted by accelerometers. I cannot imagine
how similar machines could be transported to space, even
with a low-gee non-rocket system like a launch loop ...
which I've written about since 1981,
see http://launchloop.com

(and yes, if a software jockey reading this can help me
upgrade my server to Debian Bookworm, connectivity to
Wireguard, and help manage certs, those sites will be
https. Around here, anyone who can compile hello.c has
a mid-six-figure job, while clueless five-figure drones
waste more time and resources than they can create.
Been there, done that, wasted the T-shirts)

----

Station-keeping for very large data center constellations
requires gossamer structures and lightsail-like station-
keeping thrust. Big box systems require propellant.
Propellant reaction mass doesn't vanish. It is captured
in Earth or solar orbit, and eventually crosses the orbits
of the systems emitting it. Kessler syndrome will be the
main barrier we face, and can be far worse at atomic scale.

Hence, light-sails, with reflected solar photons leaving
the solar system at light speed. Even a heavy light sail
can station-keep if the orbit is semi-stable and there
are few significant perturbers.

An Earth-orbiting data center swarm must contend with
hundreds of thousands of live, dead, or fragmented
artificial space objects, gravitational perturbations from
Sun and Moon, and the van Allen belts. Sadly, there is no
garbage service Out There, and debris makes more debris.

Unless you park on a metastable gravitational hill, like
the Earth-Sun L1 point (ESL1). Stationkeeping thrust
near ESL1 is less than a meter per second per year,
"easily" done with light-sailing gossamer systems.

There are currently a handful of solar observation probes
in semiperiodic (~6 month) Lissajous orbits around ESL1.
Their orbits encompass a Sun-facing area more than ten
thousand Earths in extent.

ESL1 is bombarded with full cosmic ray flux and occasional
solar coronal mass ejections; semiconductor chips must be
rad hard, fault tolerant, and replaceable. Fortunately,
deep-submicron transistors are VERY VERY small targets,
nanoparticles per year per transistor, and the with proper
redundant/parallel design and tail-end Moore's Law scaling,
the chips will be obsolete (and replaced with better
milligram naked semiconductor chips) before most are
cripplingly degraded.

Note, some radiation-resistant techniques are intentionally
NOT used by chip manufacturers to avoid ITAR export
restrictions. I can neither confirm nor deny ... or nearby
companies will be shut down and friends will lose jobs.

Anyway ... imagine "one thousand Earths" of sunlight flux,
with a 2.7 Kelvin universe perpendicular to the Earth-Sun
line to radiate waste heat into. I can imagine compute
systems in the Sun shadow penumbra with chip temperatures
below 0C, perhaps as low as -70C. Rule-of-thumb for chip
wearout is doubling per 10C, so these ESL1 devices might
have thermal lifetimes a million times longer than the
chips in a data center supercomputer. And like those
chips, their lifetime will be obsolescence limited, not
thermal wearout limited.

The main thing that lower temperatures permit is lower bit
energies and smaller voltage differences between "1" and
"0" bits. Thermal dissipation is proportional to voltage
squared and clock rate. Space chips can be far more
efficient than Earth chips. They must be; life is short
and there's much to compute before replacement and recycle.


What I cannot imagine is launching the vast heavy fragile
fabs that make the chips. The thinned repair chips weigh
milligrams per kilobuck - launching them from Earth will
cost rounding error compared to the cost of fabrication.

Don't judge the cost of fine art by the cost of paint
and canvas and brushes.

So - imagine ten thousand Earths of sunlight with an
Earth-illuminating donut hole in the middle. At 10%
average array fill (thus 1% self-shadowing), that is
about 990 Earths of illumination or 1.7e20 watts.

( 1 Earth ≅ 1360 MW/km² * π * (6370km)² ≅ 173e15W )

A human brain is 12 watts, 8 billion of us is 1e11W.
Our farms and machines consume far more ...

... and our emitted CO₂ to date traps FAR more heat energy
than our farms and machines use and emit. However, that's
a kettle of controversy and partisan virtue-signalling
that I would rather not split the elist with.

-----

We have a solar system to conquer, let's not worry about
slogans and group affiliations. That's mudball thinking.
AI-L1 may someday be our senior partners. Hail to the
NEW boss ... same as the OLD boss? Perhaps a LOT better
than some contemporary bosses I could (but WON'T) name.

Keith L.

-----

P.S. Regards "John Clark"s:
my UC Berkeley superconductivity physics professor
John Clarke earned the 2025 Nobel Physics Prize with
two of his grad students. If I had stayed for my PhD,
I might be on my way to Stockholm now. More likely,
I would have busted John's niobium plasma sputtering
system, and none of us would have gotten the prize.
Cynics could say I haven't done much after Berkeley.

--
Keith Lofstrom kei...@keithl.com