Do You Still Need Patents in Space?

The hook: the biggest space deal, the thinnest patent book

SpaceX is reportedly about to run the largest IPO in history. So we pulled the patent portfolios across the space economy. The headline is counterintuitive.

Company	Extended patent families
Maxar Technologies	3,905 (mostly inherited via DigitalGlobe, SSL)
Blue Origin	304
Rocket Lab	226 (SolAero acquisition-heavy)
SpaceX	87
Relativity Space	44
Sierra Space	39
Planet Labs	31
Astroscale	25
Astranis	3
Starcloud (orbital data centers)	1

SpaceX holds about 87 families, fewer than Blue Origin and Rocket Lab, a rounding error next to legacy Maxar. For the most valuable launch and space company on the planet, that is strikingly thin.

The texture matters more than the number. Most of those 87 families cluster around Starlink user hardware: phased-array antennas, radomes, user terminals, uplink synchronization. The consumer layer SpaceX sells to millions gets patented. The launch and Raptor-engine know-how does not.

Why so few? A deliberate split

A patent is a published recipe. You trade disclosure for a time-limited right to exclude. For launch and propulsion, where the edge lives in hard-won process know-how that is difficult to reverse engineer, SpaceX keeps the secret and skips the filing. For mass-market consumer hardware that ships to every customer and is easy to copy, it patents, because design-arounds are cheap and the commodity edge needs a fence.

That is the real lesson, sharper than "SpaceX does not believe in IP." SpaceX chose trade secrets for the moat and selective patents for the commodity edge. The 87-family number, concentrated in Starlink, is the evidence.

The question the IPO does not answer: enforcement in orbit

Say you do file. Once your invention is operating in orbit, who can enforce a patent on it?

US law reaches into space. Under 35 U.S.C. 105 ("Inventions in outer space," added 1990), an invention made, used, or sold on a space object under the jurisdiction or control of the United States is treated as made, used, or sold within the United States. That is the trigger that activates infringement liability under 35 U.S.C. 271.

But Section 105 has two carve-outs. It does not reach objects covered by a specific international agreement (the ISS allocates IP through its own intergovernmental agreement), and it does not reach objects carried on the registry of a foreign state under the Registration Convention, unless the US and that state specifically agree otherwise.

Under Outer Space Treaty Article VIII (1967), the state on whose registry a space object is carried retains jurisdiction and control over it while in orbit. Registry functions like a ship’s flag. Read together with Section 105, the picture is striking: where you register a satellite can determine whose patent law governs the invention running aboard it.

The detection problem

Even with a patent and the right jurisdiction, orbit defeats the assumptions courts rely on.

• No inspection access. You cannot board an orbiting object to confirm a patented method is being practiced. If it never returns, the evidence never becomes reachable.
• The making-versus-using split. Manufacture and sale usually happen on the ground, observable and clearly within a territorial forum. The infringing use happens in orbit, where jurisdiction depends on registry and observation is impractical. A pure method-in-orbit claim is the hardest to enforce.
• Flag of convenience. Because jurisdiction follows the registry, an operator can register in a state with weak or absent patent coverage to sidestep claims, the maritime analogy commentators keep returning to.

Worth stating plainly: the jurisdiction-in-orbit question is largely untested in court. There is no confirmed major in-orbit patent infringement ruling. Most of the analysis is theoretical, and the "registry equals governing patent law" rule is the prevailing interpretation by analogy, not a binding holding.

The strategic response: anchor on Earth

The practitioner consensus across the space-IP commentary is consistent.

1. Anchor claims on the ground. Draft to the apparatus and the ground-side manufacturing method, not only the in-orbit method. That captures infringement where jurisdiction and detection are realistic.
2. File broadly across the hubs that matter: country of manufacture, country of sale, the likely state of registry, and the launch state. Use the PCT, the European Patent, and the Unitary Patent for efficient multi-jurisdiction coverage.
3. Mind the registry. The US is currently the only nation that has explicitly extended its patent law to space objects under its jurisdiction. Most other space-faring states ratified the treaties without an equivalent domestic statute, which is what creates the flag-of-convenience exposure.

Why this gets sharper: AI infrastructure is moving to orbit

Compute is heading off-planet. Starcloud (Nvidia-backed) reported training an AI model in space in December 2025 and has floated a constellation of orbital data centers. Lonestar, Axiom, NTT, and Sophia Space are circling the same idea.

The patent whitespace is wide open. We found only about 72 families worldwide on space-based data centers and on-orbit compute, with no owner holding more than 14 (Mitsubishi Electric 14, then CASC, Caltech, and Starcloud’s single foundational filing). That is exactly the kind of open category where early patents can still define the space.

It also concentrates the enforcement problem. AI training and inference are textbook "use in orbit" activities: the hardest to detect, the easiest to flag-of-convenience. Where a data-center satellite is registered could decide which patent regime governs the compute happening aboard it.

The bottom line

The lesson from the biggest space deal of the year is not that patents stopped mattering. It is that where and how you file became a strategy question.

• For founders: in young categories like orbital compute, early ground-anchored filings can still stake a claim. In mature hardware, copy SpaceX’s discipline and decide asset by asset what to file and what to keep secret.
• For investors: a roadshow valuation is not an IP position. Ask what a space company actually owns, where it is filed, and what it deliberately keeps as a trade secret.
• For acquirers: portfolio size alone misleads. Maxar’s 3,905 families are largely inherited; SpaceX’s 87 are a deliberate, concentrated bet. Read the strategy, not the count.

This is strategic framing, not legal advice. For jurisdiction and filing decisions, work with qualified patent counsel.

Supporting Patent Landscape Charts

Sources

• Patent portfolio counts: Minesoft Origin, extended-family searches, retrieved 2026-06-02. Per-company queries in data/space_economy_portfolio_counts.json.
• 35 U.S.C. 105: https://www.law.cornell.edu/uscode/text/35/105
• Outer Space Treaty, Article VIII: https://www.unoosa.org/oosa/en/ourwork/spacelaw/treaties/outerspacetreaty.html
• Registration Convention: https://www.unoosa.org/oosa/en/ourwork/spacelaw/treaties/registration-convention.html
• ABA Landslide, "The Space Maze" (2024): https://www.americanbar.org/groups/intellectual_property_law/resources/landslide/2024-spring/navigating-international-law-protecting-patents-low-earth-orbit-beyond/
• Aird & Berlis, "Patent Law and Space-Bound Inventions" (2023): https://www.airdberlis.com/insights/publications/publication/patent-law-and-space-bound-inventions-the-conundrum-and-strategies-for-combating-it
• Starcloud orbital AI training, CNBC (Dec 2025): https://www.cnbc.com/2025/12/10/nvidia-backed-starcloud-trains-first-ai-model-in-space-orbital-data-centers.html