Essay

Dyson swarms are closer than you think

We're approaching a technological tipping point that could allow fully automated mining. When this is unlocked, it creates a feedback loop that can help us completely automate resource extraction in space. If we commit to building it, we are on track to have the first complete ring of a Dyson swarm around Sol before 2100.

A brief explanation so we're all speaking the same language. A Dyson swarm is an orbiting collection of individual panels designed to extract energy directly from sunlight and make it usable throughout the solar system. A completed ring of this swarm is a band of these panels orbiting Sol, tilted about 10° from Earth's orbit and only 3° from Mercury's, so Mercury's mass drivers can launch panels straight into it. Twice a year Earth's line of sight to the Sun crosses the band, and the few panels in the way turn edge-on for a few hours, so the ring never dims Earth. For the numbers, we'll be talking about a swarm ring at a radius of 0.397 AU, right alongside Mercury's orbit, and a width of about 1 million kilometers — about 1/100th of a full shell.

Building such a swarm is no small feat — especially for one that keeps Earth's climate safe — but it would catch more than 100 billion times Earth's current energy production in sunlight and turn about 10 billion times that into electricity. A complete shell swarm would catch over 10 trillion times. From our perspective now, that's effectively unlimited energy.

Another word on the plan. The long-term goal here is to actually reduce energy expenditure on Earth's surface by using all the excess energy in space for industry. Keeping the energy off Earth means we avoid exacerbating any climate woes on our planet and can even begin reversing the trend. This is unavoidable, really: thermodynamics tells us that energy expended is heat generated. So that's one reason to buy in on the plan: save Earth from overheating by making it extremely easy to build in space.

The numbers here are stark. All of our machines together add about 19 TW of waste heat to Earth, about 0.04 W for every square meter of the planet. That's nothing next to the 2.7 W/m² of extra heat our greenhouse gases now hold in. But waste heat grows with energy use. Spend a hundred times today's energy on Earth's surface and the waste heat alone would outweigh everything our greenhouse gases are doing today, about 3 °C of warming. The swarm passes a thousand times Earth's energy production around 2080. Beaming that power home isn't an option. The industry has to move to where the energy is, and by then there will be plenty of material in orbit to build it with.

That's the punchline. With a thousand times Earth's energy available in orbit, almost no industry makes sense on Earth at all. It won't vanish overnight: factories on Earth will keep running until goods made in orbit undercut them and the plants reach the end of their lives. But the trend runs one way, and by 2090 about half of Earth's industry is gone, with the rest tapering into the next century. What's left on the surface is mostly transport and everyday life. And there's so much spare power that we can pull carbon back out of the air without caring how efficiently we do it. Carbon removal plants are the one industry that has to stay on Earth, because that's where the air is. Starting at about a gigatonne a year in 2069, once the swarm generates twice Earth's energy use, and doubling with the rest of the project, they pass 100 billion tonnes of CO₂ a year in the early 2080s. On the middle-of-the-road path the climate projections call SSP2-4.5, CO₂ would reach about 600 ppm by 2100. With the swarm it peaks near 560 ppm in the mid-2070s and is back to 350 ppm, the level of the late 1980s, by 2100. That's about 2.6 trillion tonnes pulled from the air, counting what we keep emitting on the way down and what the oceans give back, and it takes about two weeks of the swarm's 2079 output at today's direct air capture efficiency. Even at five times worse efficiency it's two months of output, and the waste heat is about a thirtieth of the warming it removes. We don't need to go all the way back to pre-industrial levels. Stopping at 350 undoes most of the warming without swinging the climate the other way as fast as we pushed it.

Completing such a megastructure would catapult humanity toward being a civilization which could occupy our entire solar system, with all the space and resources we could dream of for centuries. Even a small fraction of its completion would have an immediate impact toward putting several times Earth's energy production into available industry in space. Humanity would have the room and resources to reach a population of hundreds of billions within a few centuries' time, with an ultimate capacity of trillions of people living around Sol.

But a Dyson swarm is enormous and would require part of an entire planet's worth of material to construct, even on the ring we're discussing. Taking apart Mercury for the job isn't a new idea; it shows up in Armstrong and Sandberg's Eternity in six hours as the obvious first step for any civilization that wants the Sun's power. The panels themselves are mostly thin aluminum film, about 116 g per square meter including the laser that beams their power away, so the ring's 7.5×10¹⁴ one-square-kilometer panels come to about 9×10¹⁹ kg. Getting that much aluminum and silicon out of Mercury's rock takes roughly ten kilograms of ore for every kilogram of panel, about 10²¹ kg, or 0.3% of Mercury.

None of that ore gets left behind. Every kilogram mined is thrown to orbit, where the useful part is refined into panels and the rest is cast into glass, so Mercury's surface simply recedes, about 4 km for the first ring. The glass isn't waste either: it's a stockpile of shielding for habitats once there's energy to spare to build them.

A 4 km deep strip mine is deeper than anything we've built on Earth, but since Mercury's surface gravity is only 1/3 Earth's that depth isn't actually much different from open pit mines on Earth. From a distance, Mercury would change only from a scraggly cratered surface to something more like a smooth marble.

Still, you might be wondering: how could it be possible we could complete that in under 75 years from now?!

It's my goal here to walk you through the plan and timescale that could make this work. The milestones we need to achieve to keep us on track, and the challenges that could drag the project down.

The first and most important milestone: Earth needs to demonstrate fully autonomous industry. That means automating the full cycle: extracting minerals from the ground, refining them into usable materials, and finally constructing bulk frames from the material. With robotic frames run by language models, we have the key technology to make it possible. We need to iterate on it until it can run closed loop with minimal oversight.

I'm incredibly bullish on the time scale involved here and think since the keys are in place the main driver for the timeline is now ambition. With an ambitious schedule we could solve the core problems within a decade, by 2040. When this sequence is unlocked it drives the rest of the timeline, because after autonomous industry is ready, the extraction and construction process can exponentially expand itself in space. NASA worked through this idea in detail back in 1980, in a summer study of a self-replicating lunar factory published as Advanced Automation for Space Missions. What they lacked was the intelligence to run it. We're about to have that.

Our next stop is Luna, where extraction of water ice by automated mining can produce enough fuel to run space industry and make reaching distant bodies regularly possible. We need a regular fuel supply because of our target for the material source for the Dyson swarm's main panels. Luna also allows us to prove out the automated construction process in a harsh lunar environment and create the production facilities necessary to supply new endeavors across the solar system. I estimate a build on Luna to take the better part of the 2040s.

Luna is also where swarm production starts. Its highlands are 13–15% aluminum and launching off the Moon takes only 2.4 km/s, so lunar factories can start making panels by 2045. The first one doubles as a test article: it's commissioned near Earth, beaming about 50 MW back to Luna by 2047, before it sails in to the ring by 2050. Lunar panels are built lighter and beam only a quarter of their power so they can make the four and a half year trip in toward the Sun, and lunar output doubles about every 1.2 years.

Because all bodies outside Earth fall under the Outer Space Treaty, the build-up on Luna will be the time to begin building international agreement for the whole plan. A consortium of countries should control the final Dyson swarm so that no individual country monopolizes energy from the sun. Not only is this ethical, it's practical. The more countries on board with the plan of building a Dyson swarm, the fewer diplomatic blocks we can expect in its future. With every major power holding a stake in the project, its output can belong to all of humanity. There are political issues that would need to be resolved for this; US law (the 2011 Wolf Amendment) currently bars NASA from working with China on space.

There is also the question of resources. An organization or company spearheading the build and engineering will need access to enormous capital. If it is one of the first to demonstrate full industrial automation on Earth, some of that will already be funded. But even a strong revenue stream won't cover the tens of billions of dollars expense that will come every year from a project of this size, even in its early years.

So the consortium serves the second purpose of driving funding and production into the Dyson swarm project. It is a win/win for the contributing parties because every stage of the build is designed to begin producing returns for the investors within 10 years' time.

Our next target beyond Luna is Mercury, and this is where swarm production really scales. If we have launched for Mercury by 2050 with a consortium of countries supporting the mission we'll be on track.

Mercury might seem like a hostile place but it's actually about as manageable as the Moon once you're there. The voyage taking around two years makes it much less recoverable if something goes wrong. Nestled in the polar shadow with power from the sunlit edges, a robotic mining operation can be made self-sufficient with only small, hard-to-construct deliverables from Earth ("vitamins"). The vitamins will power the intelligence needed on Mercury to run the facility with local models. With the learnings from Luna and expansion in parallel with its bring-up, we build up a Mercury foothold over the 2050s, with most of the time spent in two deliveries' worth of travel time. By 2060 it's mining and refining its own aluminum and building its own factories and mass drivers.

Not everyone agrees Mercury is the right place to start. A 2022 review in Physica Scripta (Smith, "Review and viability of a Dyson Swarm as a form of Dyson Sphere") passed over Mercury for Mars for exactly these reasons: about 12.6 km/s to get there from Earth and a hot surface to land on. But we only have to get a seed factory there once. Everything after that is built and launched from Mercury itself, and from there the ring is only about 4.9 km/s away.

Once we've established a self-sufficient industry on Mercury it's time to build the first panel for the swarm. There will likely be some trial and error getting the design right here, and no panel lasts forever anyway. The thin film itself should hold up for decades: the radiation damage that wears down solar cells heals itself in amorphous silicon at the temperature these panels run at. The lasers that beam the power away wear out first, so the panels are built with laser modules that robots swap out every eight to ten years, and each panel retires after about twenty years. Our first panel goes up by 2060.

The first working swarm panel will be a huge milestone because energy delivery back to Mercury will bootstrap the rest of the industry to build the next panel, and the next, until the Dyson swarm is large enough that its excess energy can begin to transmit back to near Earth orbits.

This energy is likely to exceed the rate of expansion of usage on Earth. After the panels are started, the number of them active will rise to full rate with the bottleneck being Earth industry for vitamins. From here on out, if production keeps scaling the next milestone comes in 2067, when the swarm matches all the energy Earth uses today, nearly all of it from lunar panels thanks to Luna's head start. Mercury's faster doubling overtakes the Moon around 2088, and from then on Mercury builds the rest of the ring. And then our final milestone in 2100 for the goal is the completion of the ring.

Those 40 years to complete the ring require production rates on Mercury to double 50 times at 0.8 year intervals. This exponential being driven autonomously is the only thing that makes the ambitious timeline possible. Earth's supplies of vitamins also need to be phased out by coarse semiconductor manufacturing providing radiation tolerant parts for the later panels as their production rate increases past a handful per year.

This also makes Mercury the model for how much industry a planet can carry. Every joule its mines and launchers spend ends up as heat the planet has to radiate away, so the pace of the whole project is capped by Mercury's heat budget. The plan raises that cap on purpose: a reflective sheet laid over its day side, radiator fields glowing at 1200 K, and lines of mass drivers that fire through the night and bank their heat for the radiators to shed by day. Earth has a heat budget too, just a far tighter one, set by what people and ecosystems can live with. Working out that accounting on Mercury first gives us the rules for every other world we industrialize.

Once the ring is finished, about 5% of it retires every year. Those panels sail themselves back to the refineries in orbit, which recover about 90% of their aluminum and silicon, so keeping the ring running lowers Mercury's surface by about 20 m a year instead of nearly 200.

And the first ring isn't the end. With Mercury's factories already at full speed, eleven more rings at other radii and tilts can follow within about three to six years, ten times the power again. Past that, the swarm becomes a full shell of panels on independent orbits, and however dense it gets, the panels on Earth's line of sight to the Sun turn edge-on as it passes, so it never shadows Earth. The leftover rock from Mercury, cast into glass in orbit, is enough shielding for habitats with over ten thousand times Earth's land area.

Excess power can be used to accelerate Earth even further by placing compute directly in the swarm, allowing energy production and planning to be centralized into a "Matrioshka" brain for directing the power usage near Earth efficiently.

So who controls all the power of the sun after the swarm is built? The compute on the swarm only orchestrates it: to direct power, ensure it's pointed at the proper receivers, and balance the output. So who gets to request power? Anyone with a valid receiver, to the limits of what the receiver is capable of. That means that control of the power is granted to those that are allowed to produce valid receivers, and who can revoke them.

Who will get to do that? If the consortium maintains its relationship they would be the ones who decide, delegating the administration to whomever they felt could do the task best for their interests. For safety reasons — the directed output of a misaimed power transmission could cause serious damage — these signatories also need to retain the power to cancel a receiver when it is misplaced or misused.

The ability to decide who gets power resting in so few hands may unnerve you, and I wish the engineering and political landscape could produce a better plan. Right now it's the best we can do.

None of this needs a breakthrough we don't already see coming. The economics work, and the engineering timelines are tight but achievable. What it needs is the ambition to start. If we begin by 2030, there's a ring around the Sun by 2100, an Earth with its industry moved off-world and its air back to 350 ppm, and the whole solar system open to us.

Sources