The Land Grab Above the Weather: 105,000 Satellites Filed in a Single Week
[Navigation links not applicable for this news-driven series entry]
Two companies asked the same regulator, in the same news week, for permission to put a combined six-figure count of satellites into orbit. Orbital, a five-month-old startup from Los Angeles running on five million United States dollars of pre-seed funding, asked the United States Federal Communications Commission (the FCC) to authorise up to one hundred thousand satellites, forming a ten-gigawatt data centre that would run its computers in space rather than draw on any grid on the ground. Amazon Leo, filing through its Kuiper Systems subsidiary, asked separately for five thousand one hundred and five satellites that would speak directly to ordinary mobile phones. Together, that is roughly one hundred and five thousand satellites of new paper, reaching the press in a single week.
One correction belongs in the headline before we go further, because precision is the whole point of this series. The two filings did not land on the same day. Amazon Leo lodged its application on the twenty-fourth of July. Orbital lodged its own a month earlier, on the twenty-fourth of June, and only reached its main wave of coverage in the same week as Amazon. "Filed in a single week" describes when the world wrote about both, not when both were filed. That distinction matters, because the story here is not a coordinated event. It is five separate commercial actors, arriving independently at the same conclusion within a few months of each other: in a regulatory system with no capacity-weighted review, the rational move is to file early and file big.
What actually got filed
Orbital's application is the more extreme of the two. The company wants to place up to one hundred thousand satellites in sun-synchronous low Earth orbit (LEO), between five hundred and eight hundred and fifty kilometres up, each carrying roughly one-hundred-kilowatt solar arrays across a span of about one hundred metres. The satellites are built for artificial intelligence inference, not communication. Coverage framed the ten gigawatts of compute as comparable to the average demand of eight to ten million United States homes, delivered without terrestrial electricity, land or water. Data would move between satellites over optical links and down through third-party constellations; Orbital is asking the FCC only for backup telemetry spectrum. The company has funded this with five million dollars of pre-seed money and targets a single-processor demonstrator in 2027.
Amazon Leo's filing is a different kind of ambition. Its five thousand one hundred and five next-generation satellites are engineered for direct-to-device connectivity: voice, messaging, mobile broadband, emergency calling and machine-to-machine traffic, delivered to unmodified four-gigabyte and five-gigabyte smartphones with no special hardware. The system runs on spectrum from Amazon's pending acquisition of Globalstar, an eleven-and-a-half-billion-dollar deal announced in April 2026 and expected to close in 2027. This is a direct competitive move against the equivalent direct-to-cell service already in the market.
The number underneath the number
Neither filing means much on its own. What makes them matter is what they land on top of. Six weeks ago, this series set out the baseline: roughly one point one four million satellites already sit in FCC dockets from three earlier applicants, SpaceX at close to one million, Blue Origin's Project Sunrise at fifty-one thousand six hundred, and Starcloud at eighty-eight thousand. Against roughly ten thousand operational satellites of any kind currently in orbit, that is a ratio of about one hundred and fourteen to one. Paper to hardware. Applications to things that actually fly.
This week's two filings add another one hundred and five thousand satellites to that pile, and they come from two entrants with no operational relationship to the earlier three. One is a hyperscaler subsidiary. The other is a company that did not exist half a year ago. The pattern named six weeks ago has not stabilised; it has compounded. And the reason is structural, not speculative. This is the Energy Ceiling in plain view: compute is moving toward orbit because Earth-side land, water and grid capacity cannot keep pace with the demand that artificial intelligence is placing on them. When a five-month-old startup with five million dollars can file for a ten-gigawatt orbital data centre, the ceiling is no longer pulling only at hyperscalers. It is pulling at anyone who can complete a form.
The quiet story that tells you more
The louder story this week was the two filings. The quieter one was more revealing. Amazon Leo also reached a hard deadline it could not meet.
Under its original 2020 licence, Amazon was required to have half of its first-generation constellation, one thousand six hundred and eighteen of roughly three thousand two hundred and thirty-six satellites, in orbit by the thirtieth of July 2026. By that date it had launched only a few hundred. What happened next is the part worth studying. The FCC did not let the deadline lapse, and it did not enforce a penalty. It issued a conditional waiver, order DA 26-553, and it did so proactively: secondary coverage reported the waiver from as early as the ninth of June, ahead of the deadline itself.
The terms are real, not cosmetic. Satellites Amazon launches after the thirtieth of July lose the spectrum-priority ranking they earned in the original processing rounds. That priority is restored once Amazon reaches the fifty per cent milestone, or by March 2028 at the latest, brought forward to October 2027 if the company can show completed satellites and sufficient launch contracts. Amazon must also surrender its surety bond for missing the threshold. The final deployment deadline, the full constellation by 2029, is untouched.
Read that sequence again, because it is the governance argument in miniature. Confronted with a hard milestone that a well-resourced operator inside its own system could not meet, the regulator's only practical lever was to bend the deadline rather than enforce it. That is not a criticism of the FCC. It is a description of the tools available. And it tells you something about what happens when the same docket machinery receives a one-hundred-thousand-satellite application from a company with five million dollars in the bank. If the system cannot hold a deadline against an operator it already licensed, it has no obvious mechanism to weigh scale, credibility or capacity to execute against an applicant who has so far demonstrated only the ability to file.
This is the pirate radio parallel that runs through Space Mafia, sharpened into a procedural form. The grey space these operators exploit is not only physical, out beyond the reach of any terrestrial court. It is procedural: an application process that processes a speculative filing and a routine licence renewal with the same rubber stamp, because it was never built to tell them apart.
Why New Zealand is already inside this story
It would be easy to file all of this under "overseas news." It is not. This week's Amazon Leo filing comes from a company New Zealand already depends on at three separate layers.
Amazon Web Services opened its Asia Pacific (New Zealand) region in Auckland in 2025, an investment it has put at seven and a half billion New Zealand dollars and described as "sovereign by design," with three availability zones at launch. Amazon Leo, this week's filer, is the same corporate parent. And Amazon is a founding member of SpaceConnect, the trade body setting standards for the non-geostationary satellite industry. Cloud, connectivity, and a hand on the industry rulebook: one company, three layers, with a New Zealand footprint riding inside each of them.
That is worth naming precisely, because it is the difference between two of the book's patterns. Rocket Lab's acquisition of Iridium, which this series covered earlier, was collaborative vertical integration: separate companies, often across allied nations, consolidating into a shared capability. Amazon's arc is the other pattern, vertical integration under a single roof, connectivity and cloud and standard-setting accumulating inside one balance sheet.
One caveat applies regardless of where the data physically sits. Amazon Web Services operates under United States law, including the Clarifying Lawful Overseas Use of Data Act, the CLOUD Act, which lets United States federal law enforcement compel a United States provider to produce data no matter where it is stored. An Auckland data centre changes where the data lives. It does not change whose jurisdiction reaches it.
New Zealand has lived with concentrated physical-layer dependency before, and the closest precedent sits under the sea, not above the weather. The country has more than ten undersea cables, but almost all of its electronic communication with the world runs through five main links, carrying on the order of ninety-nine per cent of international traffic. Industry commentary holds that losing any single one would not be badly disruptive, thanks to spare capacity and cooperative rerouting; academic researchers cited by Radio New Zealand caution that a severe event affecting several cables at once could leave the country leaning on an indirect route through Sydney. In 2026 the government stood up a National Surveillance and Warning Capability at the Maritime Operations Centre in Wellington, described in coverage as a world-first, tracking vessels near those cables and raising an alert when one approaches closely enough to pose a risk. Dependency first, monitoring afterward. If direct-to-device connectivity and orbital compute become similarly load-bearing over the next decade, that is the pattern the country already knows.
There is a genuinely local instance of the in-orbit-processing idea, too, and it is small enough to hold in one hand. The University of Auckland's Te Pūnaha Ātea 2 mission, targeted for launch in early 2028, will carry six curated New Zealand payloads. One of them, from Lune Digital, is a maritime-domain-awareness instrument the university describes as smaller than a can of soft drink: it captures ocean imagery, uses onboard artificial intelligence to flag areas of interest, and downlinks only the relevant data rather than the full raw feed. That is the same architecture as Orbital's ten gigawatts, in miniature: process where the data is, move only what matters. The mission's funding and payloads are drawn from the university's own announcements, so treat the detail as the institution reports it rather than as independently confirmed.
What this means on Monday morning
By the time this article publishes, one piece of the New Zealand picture has already changed. The Ground-Based Space Infrastructure regime's transitional authorisation period, during which in-scope operators were automatically authorised, ended on the twenty-ninth of July 2026. From that date, an operator, new or existing, must have applied for and received formal authorisation to keep operating. The Ministry of Business, Innovation and Employment states that operators needed to apply at least four weeks ahead of the deadline to avoid a gap. The penalties for operating without authorisation reach fifty thousand New Zealand dollars and up to a year's imprisonment for an individual, and up to two hundred and fifty thousand dollars for an entity. Providing false information carries its own separate penalties. That is the mechanism, stated as it stands; it is now the enforced default, not an approaching date.
The reason this matters to organisations that would never call themselves space companies is the Hidden Space Operator problem the book names: a ground-station or infrastructure provider can find itself regulated as a space operator without ever having set out to be one. The same blind spot the FCC shows at scale, an inability to easily tell a credible operator from a speculative one, is the reason an authorisation regime exists at all. If your organisation touches ground-segment infrastructure, the question is no longer theoretical.
For boards and architects weighing multi-decade dependency on orbital connectivity or compute, this week hands you a sharper vendor question. Five or more distinct applicants are now filing for orbital infrastructure, and not one of them has passed a capacity-weighted regulatory filter, because none exists. The only qualifying step any of them has cleared so far is the ability to lodge an application. Choosing among them is an enterprise-architecture problem playing out at national-infrastructure scale, and the due diligence cannot be outsourced to the regulator, because the regulator is not doing it.
Heaven and Skynet, at the level of enforcement
Space Mafia frames every domain as a choice between an accountable path and an unaccountable one. Here the split runs through enforcement capacity. The hopeful reading is genuine: a real plurality of commercial actors, five or more distinct filers across compute and connectivity, means no single company owns the orbital land grab, and a five-month-old startup can, on paper, compete for the same regulatory attention as an established hyperscaler. That openness is worth something.
The harder reading sits right beside it. That same openness means the regulator processing all of these applications has, on this week's evidence, no reliable way to hold any single applicant, however well-resourced, to a hard deadline once it is missed. A governance system that cannot enforce its own milestones against an operator already inside it is a weak foundation on which to evaluate operators who have demonstrated nothing beyond a filing. The plurality is real. So is the vacuum. They are the same fact, seen from two sides.
The land-rush ratio has worsened, not settled. Roughly one hundred and five thousand newly filed satellites this week land on top of about one point one four million already in the docket, against some ten thousand that actually fly. Filing volume is option-staking on spectrum and orbital slots, not near-term capacity. The FCC's conditional waiver to Amazon Leo is the clearest evidence yet that the system cannot hold a hard line even against an operator it already licensed. For New Zealand organisations, the exposure is concrete: a company you already depend on at three layers is one of the filers, your Ground-Based Space Infrastructure obligations are now enforced, and the closest precedent for concentrated dependency is sitting on the seabed.
The open-source dimension is where a governance vacuum this size usually gets its first practical answer. The ground segment and flight software that will run these constellations do not have to be proprietary black boxes. Open standards from the Consultative Committee for Space Data Systems already define how spacecraft and ground stations speak to each other across operators and borders. Open mission-control and flight-software stacks, from NASA's core Flight System to community ground-segment tools such as Yamcs and OpenC3 COSMOS, let an operator, or a small nation, inspect and hold the software its dependency rests on rather than renting it sight unseen. When a regulator cannot weigh capacity at the filing stage, verifiable open standards and open ground-segment code become one of the few levers a downstream user actually controls. That is the on-ramp to the harder question: who gets to task, and to trust, the infrastructure overhead.
That question is, in the end, a sovereignty one. Direct-to-device satellites that reach an unmodified handset without touching a domestic carrier, and orbital compute that processes intelligence before it ever reaches the ground, are dual-use by construction. The same maritime-domain-awareness architecture flagging vessels for a university payload is what allied military constellations, United States Space Force Starshield among them, are built to deliver at scale. For a country whose exclusive economic zone (EEZ) is among the largest in the world, the capacity to task earth observation over its own waters, and to know whose upstream analysis it depends on, is not abstract. Allied intelligence sharing, from space-based missile warning to maritime surveillance, rests on who owns the sensor, the compute and the downlink. When those layers concentrate inside a handful of commercial operators no regulator can weigh, sovereign tasking becomes a question of dependency, not capability.
When the infrastructure your country depends on is filed faster than anyone can evaluate it, what is the question your organisation should be asking before it signs a multi-decade contract with whoever files first?
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. They do not represent the views of my employer, 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. I follow the Public Service Commissioner's Code of Conduct for the Public Sector and social media guidance.
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.
I use AI tools, including Sudowrite, Claude, Perplexity AI, DeepSeek AI, ChatGPT, Grok, Copilot, Openart and Gemini, as deliberate production tools, not ghostwriters. This is consistent with my position: AI amplifies human judgement; it does not replace it. The frameworks, arguments, and editorial decisions in this series are original work. AI accelerated the process. The thinking is mine.
References
[1] SpaceNews. "Orbital files plans for 100,000 orbital data centers." 2026. https://spacenews.com/orbital-files-plans-for-100000-orbital-data-centers/
[2] Data Center Dynamics. "Orbital files for 100,000 space data satellites for 10GW compute constellation." 2026. https://www.datacenterdynamics.com/en/news/orbital-files-for-100000-space-data-satellites-for-10gw-compute-constellation/
[3] Broadband Breakfast. "Orbital seeks FCC approval for 100,000-satellite data center constellation." 2026. https://broadbandbreakfast.com/orbital-seeks-fcc-approval-for-100-000-satellite-data-center-constellation/
[4] CNBC. "Amazon files for direct-to-device satellite network." 27 July 2026. https://www.cnbc.com/2026/07/27/amazon-satellite-internet-network.html
[5] SpaceNews. "Amazon files application for direct-to-device satellite constellation." 2026. https://spacenews.com/amazon-files-application-for-direct-to-device-satellite-constellation/
[6] SatNews. "Amazon files FCC application for 5,105-satellite constellation to enable direct-to-device mobile services." 27 July 2026. https://satnews.com/2026/07/27/amazon-files-fcc-application-for-5105-satellite-constellation-to-enable-direct-to-device-mobile-services/
[7] Federal Communications Commission. "Order DA 26-553 (Kuiper Systems conditional waiver)." 2026. https://docs.fcc.gov/public/attachments/DA-26-553A1.pdf
[8] IndexBox. "FCC grants Amazon Kuiper conditional waiver on satellite deployment deadline." 2026. https://www.indexbox.io/blog/fcc-grants-amazon-kuiper-conditional-waiver-on-satellite-deployment-deadline/
[9] Ministry of Business, Innovation and Employment. "Ground-based space infrastructure: apply for a licence or permit." 2026. https://www.mbie.govt.nz/science-and-technology/space/apply-for-a-licence-or-permit/ground-based-space-infrastructure
[10] Radio New Zealand. "Government taking 10 initiatives to safeguard undersea internet and power cables." 2026. https://www.rnz.co.nz/news/political/594779/government-taking-10-initiatives-to-safeguard-undersea-internet-and-power-cables
[11] University of Auckland. "TPA-2 prepares six payloads for orbit." 2 July 2026. https://www.auckland.ac.nz/en/news/2026/07/02/TPA-2-prepares-six-payloads-orbit.html
[12] IT Brief New Zealand. "AWS Auckland region opens doors for New Zealand cloud adoption." 2025. https://itbrief.co.nz/story/aws-auckland-region-opens-doors-for-new-zealand-cloud-adoption