The Compute Goes Up, the Licence Stays Down
One number has been repeated by nearly every outlet that covered it, and it is wrong. On 22 July 2026 the United States Federal Communications Commission (FCC) voted, unanimously, to tear up the licensing framework that had governed small satellite constellations for a generation. The trade coverage that followed settled quickly on a single figure: a ten-million-US-dollar bond, mandatory, facing every new operator. This project's own files carried that reading too, until this week. It is not a flat barrier. It is a choice, and which choice an operator makes will decide whether its claim on scarce spectrum survives contact with a better-funded rival later.
The FCC's "Space Modernization for the 21st Century" Report and Order was adopted 22 July 2026 and released the following day. It eliminates the streamlined small-satellite licensing pathway outright, along with the two-hundred-satellite threshold that used to trigger it, and replaces the Commission's Part 25 satellite rules with a new Part 100. Trade press has called the result a licensing assembly line: standardised application requirements and predictable timelines in place of the case-by-case discretionary review the old framework required. Conditional grants now let an operator begin limited operations before final authorisation clears, and annual processing rounds begin with the Ka, Ku, V and Q spectrum bands. For an industry that has spent two decades filing applications one at a time and waiting for individual review, a fixed annual cycle with published criteria is a structural change in how a satellite company plans its own capital raising, not a footnote to it.
The old streamlined pathway existed because the FCC judged, years ago, that a constellation under two hundred satellites was small enough not to need the full case-by-case review a larger system received. That judgement no longer holds, and the Commission removed the threshold rather than raise it, which is itself a statement about how far constellation sizes have moved since the rule was written. Every applicant now enters the same rebuilt Part 100 structure. The two-pathway choice inside it is the mechanism through which that single structure still lets an operator's own financial position determine, in fact if not in law, how quickly and how securely it establishes its claim on spectrum.
The correction sits inside the bond mechanics themselves. Part 100 requires a ten-million-US-dollar surety bond, posted within thirty days of authorisation, only from operators of non-geostationary or multi-orbit satellite systems who voluntarily enter an annual processing round. That bond declines as deployment progresses and reaches zero once ninety per cent of a constellation is in orbit. It disappears entirely for geostationary operators, for non-geostationary systems that stay outside a processing round, and for a smaller category of variable-trajectory missions. Processing-round participants keep the same six-year and nine-year deployment milestones that applied under the old rules; everyone else falls under a newly adopted milestone framework aligned to International Telecommunication Union (ITU) practice instead. Enter the round and accept the bond in exchange for spectrum-sharing priority, or stay out and keep the cash, accepting a slower and less protected claim. That is the actual mechanism, and it rewards legal sophistication and capital planning rather than simply punishing every new entrant with a flat fee.
The same eight-day window produced the piece's commercial hook. On 24 August 2026 NVIDIA announced that SpaceXAI, the AI computing initiative supporting SpaceX's Grok workloads, will build its first-generation Starmind AI1 satellite on an optimised Vera Rubin NVL72 system. That is the same architecture that pairs seventy-two Rubin GPUs with thirty-six Vera CPUs in NVIDIA's terrestrial racks. Orbit changes the engineering. Radiation hardening replaces a shielded server room, a radiator replaces the facility water loop that cools the rack on the ground, and there is no technician who can walk over and reseat a card once the satellite is in position. NVIDIA describes prototype testing in early 2027, launch in the fourth quarter of 2027, and what the company calls significant scale-up in 2028; those are Elon Musk's and NVIDIA's own stated intentions, not a filed regulatory schedule, and should be read as such. One outlet's coverage of the announcement puts the satellite's compute payload at 120 kilowatts, peaking at 150 kilowatts. That figure matches the lower end of a power-rating dispute this series has carried since early August without resolving it. Nothing this week settles which figure is correct, and neither should be treated as final. NVIDIA's own materials also describe a long-term ambition of up to one million AI satellites operating as a distributed orbital supercomputer. That is a stated company vision, not a deployment plan, and it belongs in a sentence with the word "aspiration" attached to it.
The pirate radio parallel, one lever further
Space Mafia's organising analogy holds that 1960s pirate radio ships exploited a governance gap in international waters until regulation eventually caught up with them. Today's orbital compute operators are exploiting an equivalent gap left by the 1967 Outer Space Treaty, which has nothing to say about who may process data in orbit. This series has already walked that argument down two levels. It began with orbital jurisdiction itself. Three weeks ago it moved to the ground station, arguing that sovereignty in orbit is decided by whoever controls the antenna that talks to the satellite. This week's finding moves it one level further, to the licence that lets an antenna or a satellite operate at all.
That is the real story inside the FCC's rewrite. Jurisdiction over orbital spectrum was never simply occupied by whoever arrived first; it is granted, through an administrative process, and the shape of that process just changed substantially with almost no accurate public description of how. A rules-based, published, two-pathway regime is a genuine improvement over the discretionary review it replaces: every operator, regardless of size, now knows in advance what the trade-off looks like. But the choice itself favours whoever can correctly price a declining bond against a fixed milestone schedule, and that capacity is not evenly distributed among the operators the new regime now governs. Both readings sit inside the same set of facts, and this series is not going to resolve which one wins.
Starmind AI1 belongs to a secondary thread running underneath the licensing story rather than competing with it. SpaceXAI extending SpaceX's own launch and orbital assets with a purpose-built NVIDIA compute payload is the same vertical-integration pattern this series named Corporate Kardashev when it first appeared. One entity is adding another layer, AI plus compute plus launch plus orbital infrastructure, to a stack it already controls end to end. The engineering detail belongs in service of that governance point, not as a spec sheet for its own sake. A single vendor relationship building a satellite is a commercial story. A licensing regime that decides who gets to defend spectrum rights against that vendor, or against anyone else, is the sovereignty story, and it is the one this article is actually about.
What actually happened in New Zealand this week
Rocket Lab's ninety-third Electron mission, "The Lightning God Defends," lifted from Launch Complex 1 at Māhia at 1.04 a.m. New Zealand Standard Time on 21 August 2026, deploying the ninth synthetic-aperture-radar satellite in Japan's iQPS constellation into low Earth orbit. It was the fourteenth Electron launch of 2026 and kept that constellation's mission-success record at one hundred per cent, with nine further dedicated launches still on the manifest through 2030. This is a genuine New Zealand hook, and it sits inside the same eight-day news window as the FCC order and the Starmind announcement. It is not derived from either. The iQPS payload is a Japanese Earth-observation satellite with no connection to Starmind, SpaceXAI or United States spectrum licensing, and this article is not going to invent one.
Fourteen launches by late August, at a perfect mission-success rate for a single dedicated customer's constellation, is the kind of cadence figure that used to be a headline on its own in this series. It is background here on purpose. The book's argument about launch economics was always that a reliable, repeatable launch cadence is the precondition for everything the licensing and hardware stories in this article actually turn on. A licence is only worth holding if you can put something into the orbit it covers, on a schedule you can plan around. Māhia's role in that argument does not change because the week's more consequential story happened in Washington rather than on the Māhia Peninsula.
New Zealand's own ground-based space infrastructure regime sits in the background of this story rather than inside it. The regime is created by the Outer Space and High-altitude Activities Amendment Act 2025, which came into force on 29 July 2025 and amends the underlying 2017 Act; it is commonly shortened to the GBSI regime, though it is not itself a stand-alone statute. It requires anyone operating ground-based telemetry, tracking, control or space-surveillance capability to hold authorisation and to run protective security and due-diligence systems, and its transitional automatic-authorisation window, which had covered existing operators, closed on 29 July 2026. The Ministry of Business, Innovation and Employment administers the regime; the Minister for Space is the statutory decision-maker. Nothing in this week's findings connects the FCC's rewritten regime to New Zealand's own authorisation requirements in any direct way. They are separate systems in separate jurisdictions, linked only by a general recognition provision addressed below.
Two further facts describe New Zealand's exposure to exactly the kind of infrastructure concentration this series tracks, supplied here as background rather than as this week's headline. Amazon Web Services' Asia Pacific (Auckland) Region has been operating since September 2025, three availability zones deep. Named government customers include the Department of Conservation, the Ministry of Health and the Ministry of Justice, against a reported investment above seven and a half billion New Zealand dollars. And New Zealand's international connectivity runs through five physical cables across three networks, including the original Southern Cross system's figure-eight design, which has already continued operating through at least two prior cable cuts without disruption. That is a more measured picture than a single-point-of-failure framing would suggest: real exposure, but architecturally mitigated exposure, and this article is not going to overstate it in either direction.
The question every dependent operator now has to answer
An organisation whose infrastructure, insurance or supply chain touches a satellite operator licensed through the FCC does not get to sit this decision out. Four questions travel well beyond this specific order. Does your dependency run through a processing-round participant, a non-participant, or through an operator whose choice you have not actually checked? If it runs through a processing-round participant, do you know where that operator sits against its six-year and nine-year milestones, given that a missed milestone forfeits the spectrum-sharing priority it paid a declining bond to keep? If it runs through a non-participant, do you understand that the slower ITU-aligned milestone framework trades speed for the absence of a bond, and what that trade means for how quickly that operator's capacity actually arrives? Section fifty-one of New Zealand's own Outer Space and High-altitude Activities Act 2017 lets the Minister take an existing FCC or Federal Aviation Administration authorisation into account when assessing a local application. Has anyone reread it closely enough to know how it would treat an operator's choice between these two pathways specifically? No source located this week answers that fourth question, and this article states that gap rather than guessing at it.
What to watch next
Two things about this week's findings remain genuinely open. First, the Starmind AI1 power-rating dispute, 250 kilowatts peak against 150 kilowatts peak depending on the source, has narrowed but not resolved. This week's coverage of the 24 August announcement lines up with the lower figure, from a source not previously logged against either number. That is all that can honestly be said about it. Second, no New Zealand-specific measurement of satellite-navigation jamming or spoofing exposure could be located this week, despite an active search; global evidence of rising interference exists, but this series will not manufacture a New Zealand figure from a global trend. Both gaps stay open until better evidence closes them.
The open-source dimension of this story is worth naming before the sovereignty question closes. SatNOGS is the Libre Space Foundation's crowdsourced, openly licensed network of more than five hundred ground stations. It gives universities and small national programmes the same practical access to satellite telemetry that a hyperscale operator builds in-house, without needing to win a spectrum-licensing round of its own. It downlinks from public and amateur satellites, not the commercial constellations the FCC's Part 100 regime governs. But it demonstrates the alternative the licensing debate tends to obscure: ground-segment capability that depends on participation, not capital or a regulatory contest. As spectrum rights consolidate around whoever can price a declining bond correctly, open, community-run infrastructure remains the one layer of the stack that still asks nothing but that. That contrast is what makes this week's sovereignty question the licence rather than the satellite.
The FCC's rewritten Part 100 licensing regime is, at root, a sovereignty instrument. It decides which satellite operators, anywhere in the world, can hold spectrum rights that a court would actually defend against a later, better-funded rival. New Zealand-linked ventures that depend on United States-regulated spectrum face the same choice as every other applicant: enter an annual processing round and accept a declining bond with fixed six-year and nine-year milestones in exchange for priority, or stay outside it under the International Telecommunication Union's slower, un-bonded milestone framework. Section 51 of the Outer Space and High-altitude Activities Act 2017 already lets New Zealand's Minister recognise a foreign authorisation such as this one when assessing a local application, so a choice made offshore can carry weight at home. This is regulatory mechanics, not a verdict on it.
If your organisation depends on a satellite operator you have never asked which licensing pathway it chose, what would it take to find out, and what would you do with the answer once you had it?
This is one of seven weekly series in The Hamberger Report. Subscribe on LinkedIn and the next one arrives in your feed.
• • •
The views expressed in this article are entirely my own, informed by more than 30 years of professional experience in architecture, security, and technology leadership in New Zealand. I write as director of Te Pono Limited; the views are personal and do not represent the position of any client, any government agency, or the New Zealand government. My commentary on legislation and policy is analytical, drawing on publicly available sources and my professional expertise in architecture, security, and AI governance, and it is politically neutral.
• • •
Andreas Hamberger is a New Zealand leader in Architecture & Security and Associate Member of the Institute of Directors. Space Mafia examines the sovereignty implications of orbital compute infrastructure.
• • •
This article was produced with AI assistance under my direction. Research, drafting and images pass through a pipeline I built and govern: automated gates for source verification, forbidden language and political neutrality, and my own review before anything is published. The tools include Claude, Gemini and Openart. The frameworks, arguments and editorial judgements are mine and are the same discipline I apply to the AI systems I audit for clients. AI accelerated the work; the thinking, and the responsibility for it, are mine.
• • •
[1] Federal Communications Commission. "Space Modernization for the 21st Century, Report and Order." FCC-26-47A1. 23 July 2026. https://docs.fcc.gov/public/attachments/FCC-26-47A1.pdf
[2] Greenberg Traurig. "FCC Space Modernization Series (Part 1): The New Bond and Milestone Framework." August 2026. https://www.gtlaw.com/en/insights/2026/8/fcc-space-modernization-series-part-1-the-new-bond-and-milestone-framework
[3] National Law Review. "FCC Space Modernization Series (Part 2): The New Licensing Assembly Line." 2026. https://natlawreview.com/article/fcc-space-modernization-series-part-2-new-licensing-assembly-line
[4] NVIDIA Newsroom. "SpaceXAI Adopts NVIDIA Vera CPU to Accelerate Agentic AI at Massive Scale." 24 August 2026. https://nvidianews.nvidia.com/news/spacexai-adopts-nvidia-vera-cpu-to-accelerate-agentic-ai-at-massive-scale
[5] Tom's Hardware. "SpaceXAI Will Deploy Standalone Nvidia Vera CPUs for Grok's Agentic Workloads." 24 August 2026. https://www.tomshardware.com/tech-industry/big-tech/spacexai-will-deploy-standalone-nvidia-vera-cpus-for-groks-agentic-workloads
[6] Rocket Lab. "Mission Success: Rocket Lab Launches 93rd Electron Mission." 20 to 21 August 2026. https://investors.rocketlabcorp.com/news-releases/news-release-details/mission-success-rocket-lab-launches-93rd-electron-mission
[7] Ministry of Business, Innovation and Employment. New Zealand Space Agency regulatory regime page. https://www.mbie.govt.nz/science-and-technology/space/our-regulatory-regime
[8] RNZ. "Government Exploring Monitoring of Undersea Cables." 2026. https://www.rnz.co.nz/news/national/578821/government-exploring-monitoring-of-undersea-cables