Pippa Orbital, the DCSI mascot, in an astronaut helmet Data Center Stress Index Orbit · Data centers in space
Editorial illustration of Earth surrounded by a ring of small data center satellites in low Earth orbit, with the Moon at upper right showing a server-cube on its surface, and inter-satellite laser links between satellites. On Orbit · Click any satellite to read its dossier · Watchlist v1

Orbit, the first comprehensive public tracker of orbital data centers. Thirteen named projects, 154 verified requirements, the science behind compute that escapes the planet.

The opening

On March 6, 2025, a roughly two-kilogram solid-state data-storage payload touched down near the Moon’s south pole, the first commercial data-storage payload to reach the lunar surface. The lander it rode on, Intuitive Machines’ IM-2 Athena, tipped onto its side moments after touchdown and the mission ended early. The payload powered on and achieved partial validation (DTN comms + power-on) before power loss. Its operator, a Tampa company called Lonestar Data Holdings, billed the moment as proof that the regulatory regime governing your data could, in principle, be replaced by a treaty signed in 1967.

Eleven months later, the Federal Communications Commission accepted SpaceX’s modification request DA 26-113, to expand its Starlink license by up to one million orbital data-center satellites. There are roughly 11,000 active satellites in orbit today (Jonathan’s Space Report, May 2026); the application asks for about 90 times that. Initial public comments closed on March 5, 2026; reply comments closed March 23. On May 20, 2026, SpaceX filed its S-1 ahead of a reported ~$1.75T IPO — the same filing warns prospective investors that orbital data centers “involve significant technical complexity and unproven technologies, and may not achieve commercial viability.”

Twelve other companies have publicly announced data centers in space — from China’s state-backed 12-satellite constellation already validated in orbit, to Alphabet’s Project Suncatcher R&D announced November 2025, to Blue Origin’s 51,600-satellite Project Sunrise filed at the FCC on March 19, 2026. None operate at commercial scale yet. All of them face the same physics, and the same regulatory regime, built when data centers were rooms full of vacuum tubes and outer space was a place humans had only just reached.

This page is what’s actually up there. What it is, what it does, what it has to comply with, and what it costs the public to ignore. Click any satellite in the image above to read its dossier.

13 named projects 9 science constraints The chase, in 7 steps All 154 requirements
Thirteen named projects

From DARPA’s first hardware to SpaceX’s million-satellite request.

Five years. Twelve publicly announced projects. Filter by how real each one is. The earliest hardware goes back to 2021; the freshest active docket is SpaceX’s February 2026 FCC filing for one million orbital-compute satellites; the newest hyperscaler entrant is Alphabet’s Project Suncatcher, announced November 2025. Click any card to read its dossier.

Status 12 of 12
Thirteen named projects

Skim the watchlist horizontally.

Twelve cards, one row, ordered chronologically from DARPA’s first hardware (2021) to SpaceX’s active FCC docket (Feb 2026). Tap any card to open its dossier. The color on the left edge of each card signals status, operational (white) through hardware deployed, funded, and concept.

Swipe horizontally · Scroll · Tap any card
The science

Nine constraints every orbital data center has to satisfy

The physics is the gating input. Each tile below shows one constraint, click to flip. Use the Plain / Technical toggle in the top-right corner of each flipped tile to switch between everyday language and engineering precision.

The sustainability ledger

The case for orbital compute rests on three claims about being cleaner. All three are contested.

The marketing case for moving compute off-planet is also an environmental case. No water draw. No farmland conversion. No air-permit fights. No interconnection queue. No host community. Every pitch deck leads with some version of this. The peer-reviewed literature is more divided, and the parts of the literature that disagree are the parts the pitch decks tend to omit. Below: the three reckonings any honest orbital-data-center claim has to walk through.

1
Carbon · conditional

The carbon math is conditional on launchers that don't yet exist.

ESA's ASCEND feasibility study concluded that orbital data centers are climate-competitive only if launch decarbonizes by roughly an order of magnitude over the satellite's lifecycle. Two recent peer-reviewed analyses disagree about what that means in practice. A 2025 Nature Electronics paper finds orbital DCs can become carbon-neutral within 3 to 5 years of operation versus a terrestrial baseline. A 2025 working paper, Dirty Bits in Low-Earth Orbit, finds in-orbit systems incur embodied carbon emissions up to 10× higher than terrestrial equivalents, driven by launch and re-entry. Both can be true. The variable is launcher emissions and operational lifetime, and neither is fixed.

Cites: ASCEND Phase 1 (2024) · Nature Electronics, 2025 · Dirty Bits in LEO (arXiv, 2025)
2
Atmosphere · new externality

Rockets deposit emissions where Earth has the worst options for processing them.

A surface emission spreads. A stratospheric emission lingers. Every launch deposits CO&sub2;, water vapor, and black carbon (soot) directly into the upper atmosphere, where radiative forcing per unit mass is disproportionate compared to the same mass at ground level. Re-entry burns deposit alumina particles at altitudes where they can persist for years. The cumulative effect on the ozone layer from a sustained mega-constellation launch cadence is not well bounded in peer-reviewed research, because nobody has flown a mega-constellation cadence long enough to measure it. The voluntary Space Sustainability Rating (EPFL eSpace, since 2022) is the industry's attempt to self-regulate before formal regulation arrives.

Cites: AGU Geophysical Research Letters on stratospheric soot · EPFL Space Sustainability Rating · ESA on launch emissions
3
Debris · the Kessler cascade

The orbit you fly is also the orbit you crash, and the orbit you leave behind.

In 1978 NASA orbital-debris scientist Donald J. Kessler hypothesized that at sufficient orbital density, collisions in LEO would generate debris faster than it deorbits, triggering a self-sustaining cascade. The threshold density is what the field calls the Kessler effect, or Kessler syndrome. For decades the concept lived in theoretical models. ESA's 2025 Space Environment Report now counts approximately 54,000 objects larger than 10 cm in Earth orbit and approximately 1.2 million between 1 and 10 cm. Kilometer-scale orbital data centers add cross-sectional area at exactly the altitudes already most crowded. The cascade is no longer theoretical; the question is how close to the threshold the industry is willing to push before regulation catches up.

Cites: Kessler & Cour-Palais, JGR 1978 · ESA Space Environment Report 2025 · IADC 25-year guideline · FCC 5-year rule (2024)
Deep dive Kessler syndrome · 1978 paper, 2025 empirical concern

Why every orbital data center pitch eventually has to argue about debris.

Donald J. Kessler was a NASA orbital-debris scientist. His 1978 paper, written with Burton Cour-Palais, modeled how a single collision in low Earth orbit produces fragments that can collide with other satellites, producing more fragments, in a positive-feedback cascade. Above a certain orbital density the cascade becomes self-sustaining and the affected altitude band becomes effectively unusable for generations.

This is not a thought experiment anymore. The 2009 Iridium-Cosmos collision, a single inadvertent strike between a U.S. communications satellite and a derelict Russian military satellite, produced roughly 2,000 trackable fragments from two objects. Every fragment is now an independent collision risk for every other satellite at that altitude.

The current state of low Earth orbit, per ESA's 2025 Space Environment Report:

54,000
Objects >10 cm in orbit (trackable, mostly cataloged)
1.2M
Objects 1–10 cm (mostly untrackable)
130M
Objects 1 mm to 1 cm (estimated)
~14 km/s
Peak relative impact velocity in LEO (typical ~10 km/s; ~14 km/s on head-on encounters)

The IADC's 25-year post-mission disposal guideline is the legacy rule; the FCC moved U.S. operators to a 5-year rule in 2024. Experts increasingly consider both inadequate at mega-constellation cadence. Active debris removal does not exist at commercial scale. Nobody has cleaned up anything yet.

What this means for orbital data centers specifically:

  • A kilometer-scale array presents millions of square meters of impact-vulnerable surface, at exactly the altitudes most crowded with debris and active constellations.
  • A single penetration of a working-fluid loop can take a multi-megawatt pod offline. At gigawatt scale over a 10-year lifetime, this is not an edge case; it is a design problem.
  • The 2024 FCC 5-year disposal rule shortens the period during which a defunct orbital DC contributes to debris density, but does not address fragments produced mid-mission by impacts.
  • The voluntary Space Sustainability Rating (EPFL eSpace) is currently the industry's strongest cleanup signal. Insurance markets are starting to require it. Regulators have not.
  • A correlated debris cascade and a correlated solar weather event — see Constraint 9 — are the two systemic risks the orbital compute industry has not yet learned to price.
The honest summary The question is not "are orbital data centers cleaner than terrestrial ones?" The question is "under what launcher emissions trajectory, operational lifetime, debris-management regime, and atmospheric chemistry assumption are they cleaner?" On the present-day launch fleet, with the current debris environment, and without any commercial-scale orbital cleanup infrastructure, that case cannot be made today. Anyone who claims otherwise is selling the technology, not analyzing it.
Compliance pathway

A U.S. operator launching a 100-satellite LEO compute constellation

35 verified touchpoints — the on-page sample, drawn from the full 154-requirement tracker — from T-36 months to end-of-life: federal regulators, treaty obligations, insurance practice, standards bodies. Click any numbered waypoint along the trajectory below to jump to that phase's requirements. Solid borders in the full view are required; dashed borders are voluntary but de-facto enforced by insurers or trade partners.

Bronze-etched scientific illustration of a single satellite lifecycle. A dashed trajectory arcs from a launch site on the lower left, climbs through the atmosphere, levels into low Earth orbit across the upper canvas, then descends in re-entry on the right. Seven numbered waypoints along the arc mark the lifecycle phases: capital formation, spectrum coordination, imagery and export controls, pre-launch, treaty obligations, continuous operational, and end-of-life. Earth's curve dominates the lower third with sage-green atmospheric haze; stars and a faint Moon in the upper sky.
Click a waypoint  ·  or click the image to open the full pathway
The full tracker
All 154 requirements · filterable · cite-or-cut

International treaties, U.S. federal regulators, standards bodies, insurance practice, sustainability ratings, cybersecurity baselines, all in one searchable table. Built from the May 2026 audit, refreshed quarterly.

Open the tracker
Methodology

How this watchlist is maintained

This is a watchlist, not a stress index. There aren’t enough operating orbital data centers to score the way DCSI scores U.S. counties, and the regulatory dimensions are different. What we do publish: a curated dossier per project with capital raised, partners, launch vehicle, jurisdiction of incorporation, and the policy questions each one raises. Date-stamped. Cite-or-cut. Errors logged in the public AI Accountability ledger.

The discipline
Five rules every entry follows

1. Public source for every claim. Corporate filing, press release, SEC, NASA contract, FCC docket, scholarly paper, or named journalism. No anonymous tips or single-source assertions.

2. Date-stamped review. Every dossier has last_reviewed. Quarterly re-audit appends rows; status changes get a one-row edit + commit.

3. Cite-or-cut. If we can’t source a fact to a public document, the fact gets cut. UNVERIFIED items get a flag, not a publish.

4. Public error log. Errors are written down here, in public, before they are fixed. The log is the receipt.

5. Invitation to correct. If you know of a project we missed, a number we got wrong, or a regulator we didn’t list, submit a correction. Logged like everything else.

AI Accountability · Caught by this audit
Five hallucinations the agents corrected from my prompts

Anna’s research prompts to the audit agents contained five citations that turned out to be wrong. The agents flagged them rather than perpetuating. Documented here for the same reason climate scientists publish model uncertainty: the only way to trust a number is to know how it was made.

  • ERR-073Prompt cited CCSDS 911.5-B-2 for Bundle Protocol/DTN. Correct doc is CCSDS 734.2-B-1; 911-series is monitor & control, unrelated.
  • ERR-074Prompt cited NIST SP 1800-29 for spacecraft cybersecurity. Document doesn’t exist; relevant PNT-resilience guide is NIST IR 8323 Rev.1.
  • ERR-075Prompt cited an “Open Compute Project Space initiative.” Doesn’t exist as of May 2026; Axiom uses Red Hat Device Edge but not under an OCP working group.
  • ERR-076Prompt cited an “Uptime Institute Orbital Tier” classification. Doesn’t exist; Tier I–IV are terrestrial only. Operators can certify ground segment only.
  • ERR-077Prompt assumed Climate Neutral Data Centre Pact applies to orbital DCs. Confirmed terrestrial-EU only; orbital eligibility undefined.