The future of space is too big to fit inside a rocket

Existential Hope in Practice

In the Existential Hope in Practice essay series, our fellows write about concrete ideas and practical examples of how to build a great future.

---

The future of space is too big to fit inside a rocket

By Sidh Sikka

In brief:

• Humanity has learned how to reach space, but not yet how to build there at scale.

• Today’s space systems are still constrained by what can fit inside a rocket, limiting us to relatively small, fragile, mission-specific infrastructure.

• Kilometer-scale orbital construction would change that, enabling artificial-gravity habitats, orbital shipyards, massive telescopes, deep-space transport systems, and permanent industrial infrastructure beyond Earth.

• The key shift is from deployment to construction: instead of launching complete systems from Earth, autonomous robots should build large structures directly in orbit.

• This would make space safer, more useful, and more inhabitable, turning it from a hostile environment we briefly visit into a frontier where civilization can persist and grow.

• The required ingredients are starting to arrive: reusable launch, cheaper access to orbit, mass-produced spacecraft, better robotics, and increasingly capable autonomous systems.

• The result would not just be a larger space economy. It would be a civilization with more room, more resilience, more wonders, and a real frontier again.

---

We can’t build anything in space, at a scale larger than ~100 meters. And we can’t launch things larger than this into orbit either, we’re forced to confront something called the “Tyranny of the Rocket Equation”. This fundamental constraint of rocketry essentially says that the more mass you want to launch into orbit, the more mass of propellant you need to launch it, which adds more total mass to the system, requiring more propellant, and so on. But imagine for a moment that we’d already solved it.

What if we win?

Humanity becomes a frontier civilization again, a civilization with somewhere to grow.

The old frontiers of Earth were finite. Space is different. Space is not a continent. It is not a territory to be taken from other people (unless we find aliens, which would be very interesting). It is the final frontier in the literal sense: endless. A frontier of moons, asteroids, oceans, storms, rings, volcanoes, and worlds no human being has ever touched.

But frontiers require infrastructure. Ports, roads, ships, depots, workshops, shelters, maps. In space, those become orbital shipyards, artificial-gravity habitats, fuel depots, radiation shelters, repair stations, power systems, and construction swarms. Once those exist, exploration stops being a rare act of national theater and becomes a living frontier.

If we could build kilometer-scale structures in orbit, we could build space stations with artificial gravity. O’Neill cylinders and other rotating habitats were studied decades ago as a path toward large, habitable settlements in space, but they require structures far beyond what can be launched whole from Earth.

We could build motherships and orbital shipyards: large vessels assembled above Earth, serviced in orbit, and sent on regular cargo and crew routes to Mars, the asteroid belt, and the outer planets. Instead of treating every deep-space mission like a one-off expedition, we could begin building the logistics layer of a spacefaring civilization. The Solar System would become less like a set of distant targets and more like a connected world.

And we could build observatories that give humanity entirely new senses.

Kilometer-scale construction could also make new kinds of distributed observatories possible, including gravitational-wave instruments that probe parts of the universe ordinary light cannot reach. Today, the Cosmic Microwave Background is the oldest light we can see: the afterglow from the time when the universe first became transparent, about 380,000 years after the Big Bang. Before that, the universe was opaque to light. But gravitational waves are ripples in spacetime itself, and large space-based observatories may someday help us study traces of events from the universe’s earliest moments.

Other enormous observatories could shed light on the Cosmic Dark Ages, the mysterious period before the first stars and galaxies lit up the universe. Hidden in that era may be clues to some of the biggest open questions in physics, including the growing Crisis in Cosmology: the fact that different ways of measuring the expansion rate of the universe do not currently agree. In other words, in-space construction would not only let us build larger machines. It could help us answer the oldest human questions: Where did all of this come from? What is the universe made of? What is its ultimate fate? Are we alone?

The Solar System alone is full of wonders. Wouldn’t you want to see the ice volcanoes of Titan? The deep oceans of Europa? The diamond rain of Neptune? The volcanoes of Io? The rings of Saturn? The hexagonal storm at its pole?

Wonders can pull people out of the smallness of the present. They make the future feel large. They give civilization shared objects of awe, shared projects, shared reasons to look outward.

But none of that arrives on its own. Every part of it, the shipyards, the habitats, the observatories, depends on something we haven’t figured out yet: how to actually build at that scale in orbit. So what’s stopping us?

---

---

The Bottleneck

The International Space Station is one of the greatest construction projects in history, but it is still only 109 meters end-to-end. Its large components were delivered across 42 assembly flights, and astronauts performed spacewalks for construction, maintenance, and upgrades. That is an extraordinary achievement, but it is not a scalable recipe for building structures ten or a hundred times larger. Nor can these structures be launched whole. Falcon uses a 5.2-meter fairing, and even Starship is 9 meters in diameter. Those are enormous vehicles by today’s standards, but kilometer-scale habitats, shipyards, power systems, and observatories are not feasible payloads.

That is the bottleneck. Kilometer-scale construction cannot be done as one giant object launched from Earth. It has to be distributed across many smaller modules, many robotic agents, and many assembly steps. But once construction becomes distributed, the problem stops being only structural. It becomes a problem of autonomy, coordination, sensing, control, and reliability at scale. NASA has recognized this as an active field rather than a solved capability. Its 2025 ISAM State of Play organizes in-space servicing, assembly, and manufacturing into multiple capability areas, including robotic manipulation, rendezvous, docking, mating, relocation, inspection, and maintenance. Many of these capabilities have been demonstrated in pieces, but not yet integrated into a validated architecture for large-scale autonomous construction.

There are three reasons this bottleneck hasn’t been solved yet.

The first is structural. Kilometer-scale systems cannot be launched as single objects, and they cannot simply be “deployed” like larger versions of today’s satellites. They have to be assembled from many smaller pieces. That means any serious architecture for large-scale space construction is necessarily distributed: many modules, many robots, many assembly steps, and many points where the system has to keep working even if individual parts fail.

The second is technical. Once construction becomes distributed, the problem stops being just mechanical and becomes computational. Centralized planning breaks down as the number of agents, parts, and interactions grows. Human operators cannot micromanage every motion from Earth, and a single master planner cannot reliably coordinate thousands of physical interactions in real time. The path forward is decentralized autonomy: robots that can sense locally, make local plans, allocate tasks, correct errors, and coordinate with partial information.

But autonomy alone is not enough. Space construction is physical. Robots have to transport, align, dock, attach, and repair real objects under uncertainty. Small errors in timing, force, or orientation can propagate through flexible structures and destabilize the system. Sensing and communication may degrade. Delays, occlusion, vibration, thermal shifts, and actuation error are also factors. That means the robots need robust low-latency control, tolerant docking interfaces, and autonomy stacks that work in an uncertain environment.

The third is institutional. This problem does not fit neatly inside existing research structures. Academic labs can develop pieces of the solution, but usually lack the resources to validate full orbital construction systems. Companies are often pulled toward nearer-term markets. Agencies can advance servicing, docking, robotics, and manufacturing capabilities, but tend to do so incrementally. No existing institution is naturally built to solve the whole problem end to end.

But now is a better time than ever before to tackle this problem.

Why now?

Launch is no longer the same bottleneck it was twenty years ago. Reusable rockets have driven down the cost of access to orbit, and just as importantly, they have increased cadence. Space development can now look more like an iterative engineering cycle, which matters because large-scale construction will not be solved in one perfect mission. It will require repeated attempts, failures, upgrades, and demonstrations.

Space hardware is also becoming industrialized. Satellite constellations have turned spacecraft production from bespoke aerospace engineering into something closer to high-throughput manufacturing. Kilometer-scale construction will not be performed by one exquisite spacecraft. It will require many agents, many modules, and many replaceable parts.

At the same time, improvements in compute, robotics, multi-agent control, and AI systems make it plausible to coordinate construction without a human operator directing every step. The old model of carefully scripted human-in-the-loop assembly cannot scale to thousands of parts and hundreds of robots. The new model has to be distributed, adaptive, and increasingly autonomous.

A realistic timeline for this may have three stages. Near-term, over the next 0–5 years, the goal is a rudimentary in-space construction service: small teams of robots assembling, moving, and docking simple structures in orbit. Mid-term, over the next 5–20 years, those teams grow into larger construction swarms, building orbital yards, staging platforms, and the first persistent structures hundreds of meters across. Long-term, over the next 20–40 years, we can shift to in-situ: instead of lifting every beam and panel from Earth, we start mining metals from asteroids and using space itself as the construction site, enabling kilometer-scale systems and continuous orbital industry.

What could go wrong?

The first potential problem is militarization. The same capabilities that let robots inspect, dock with, repair, or assemble spacecraft could also be used to interfere with them. A construction swarm is also, in principle, a manipulation swarm. This is inevitable for all dual-use technologies, and the first satellite-to-satellite manipulation weapons have already been tested. The second is debris. If we put large numbers of robotic agents and modular structures into orbit, failures matter. A broken component is a collision risk. A bad architecture could make the orbital environment more crowded and more fragile. This is NOT inevitable, and responsible design can entirely mitigate this issue.

There are also real engineering risks. Large systems can suffer from vibration through flexible structures, thermal stress, control instabilities, cascading faults, and unexpected behavior from many agents acting at once. This is another reason for rapid, iterative development cycles. If there’s a large sim-to-real gap, then we need to start trying to build on-orbit, fail, gather lessons and try again.

So there are real risks, but they can mostly be designed for, governed, tested, and reduced. The worst version of this future is one where the capability emerges accidentally, privately, and without standards. The best version is one where we treat orbital construction as civilization-scale infrastructure from the beginning.

That choice defines what this new frontier becomes. Handled well, none of these risks have to stand between us and what’s on the other side: a civilization with room to grow, wonders to chase, and a horizon that never closes. If we learn to build in space, we build a larger future.

---

---

About the author

Sidh Sikka is a Ph.D. student at Purdue University working on space robotics and autonomous in-space construction. He is one of the founders of Manifold, and a relentless technological optimist who hopes to see a future where humans live among the stars.

---