The Orbital Debt

SpaceX's plan is not a thought experiment anymore — it has a name, a spec sheet, and a filing with the US FCC. This letter treats it the way this publication treats every infrastructure claim: take the vision seriously, then take the constraint just as seriously, and see what's left once both numbers are on the table.

The Vision: Power and Compute in the Void

The core of the plan is AI1, described by SpaceX as a football-field-sized orbital data center. The published specification calls for 120–150 kW of solar generation, an onboard compute module, and a liquid cooling loop that radiates heat directly into the vacuum of space — the single largest cost terrestrial data centers carry that an orbital one simply does not pay. SpaceX's own timeline targets 2028 for the first deployments, scaling toward a stated ambition of a full terawatt of orbital compute. If even a fraction of that materialises, it is a genuine new category of infrastructure, not an incremental upgrade to an existing one.

The Debris Problem, Priced Honestly

The plan collides — the word is not chosen loosely — with an orbital environment that is already under measurable strain. Kessler Syndrome, the chain reaction in which collisions generate debris that causes further collisions until an orbit becomes unusable, is not a hypothetical tail risk in the modelling anymore. A recent study places the time to a catastrophic collision, if avoidance manoeuvres failed, at roughly 2.8 days — a tenfold deterioration from the 121-day estimate made as recently as 2018. That is the trend line SpaceX's own plan proposes to accelerate.

The company's existing fleet is the evidence. Starlink, at roughly 9,300 satellites, already performs a collision-avoidance manoeuvre on average every 1.8 minutes. The AI1 plan, at scale, proposes up to one million additional satellites — roughly a hundred times the number of active satellites currently in orbit, full stop, across every operator combined.

The disposal plan compounds the concern rather than resolving it. SpaceX's FCC filing proposes retiring most of these satellites not by burning them up in the atmosphere, the standard end-of-life method, but by pushing them into graveyard or heliocentric disposal orbits around the Sun. Hugh Lewis, a space debris researcher at the University of Birmingham, has called the resulting accumulation of derelict hardware "unsustainable," and flagged that the associated collision risk would reach dangerous levels quickly enough to threaten orbits well beyond the ones AI1 itself would occupy.

What Actually Exists to Fix This

The debris-mitigation technology this plan needs is real, funded, and still early — worth stating plainly rather than either dismissing it or overstating its readiness. RESTORE (REusable Spacecraft Teams for on-Orbit debris REmoval) is the most developed Active Debris Removal concept: a formation of spacecraft flying together, carrying a net, capturing debris and directing it into a controlled atmospheric re-entry, designed to be reusable across multiple targets rather than single-use — the detail that determines whether ADR can plausibly scale to the volumes this plan would generate. A parallel approach uses robotic arms fitted with gecko-inspired adhesive grippers, built specifically to grasp tumbling, non-cooperative debris of irregular shape.

The more structurally interesting solution reframes debris as feedstock rather than liability. The European-funded DEXTER project is developing robotic tools to harvest and dismantle retired spacecraft in orbit, with scrap aluminium reprocessed into propellant for plasma thrusters — refuelling other spacecraft rather than adding to the junk count. This is the version of the solution that scales with the problem instead of racing it, but it remains a research programme, not deployed infrastructure, in mid-2026.

What This Costs the $1Q Thesis, Specifically

This publication's $1Q framework counts AI and automation as one of the five compounding engines behind a $105 trillion global economy reaching a quadrillion dollars by 2040. Orbital compute, if it works, is a genuine extension of that engine — it relaxes two real physical constraints, power and cooling, that terrestrial AI infrastructure is already straining against. The honest complication is that it does so by drawing directly against a second, largely unpriced constraint: the finite, shared, and currently degrading resource of usable orbital space. A Kessler cascade in a commercially critical orbit would not just strand AI1 — it would raise the cost and risk of every satellite-dependent system underneath the $1Q thesis's own connectivity and infrastructure assumptions, from GPS-timed financial transactions to the satellite capacity this publication has separately tracked as emerging-market infrastructure (Letter 106).

The timeline is the specific number worth watching. SpaceX's own target is 2028 for first deployment. RESTORE and comparable ADR systems are not yet operating at the cadence or cost that would let debris removal keep pace with a launch schedule of that speed, and DEXTER's recycling approach is further still from operational scale. A two-to-four-year gap between "compute goes up" and "cleanup technology matures" is the window this letter is actually pricing — not whether orbital data centers work in principle, which the engineering case supports, but whether the debris side of the ledger clears before the compute side commits.