Two public debuts were scheduled for the same day this week, three days before this sentence reached you. Neither had happened when it was written. On 13 November 2026, a Chinese satellite operator called Adaspace is due to begin trading on the Hong Kong Stock Exchange. The listing puts a public market's price on an orbital-compute business model nobody has rated before. Hong Kong's South China Morning Post describes investor reception as wary, not sold. [4] On the same day, on the far side of the Pacific, Amazon Leo's first satellites are due to fly. The rocket configuration has not flown before either. The flight opens a thirty-eight-mission relationship between Amazon and United Launch Alliance. Whether either debut went to plan is not something this article can certify. It was filed ahead of the date it describes. What follows is what was true, filed and verifiable before that date, and nothing past it.
A third fact, several steps removed from either headline, belongs in the same paragraph. A satellite operator most New Zealanders have never dealt with directly already supplies more of the country's rural broadband than any other provider. The number most people would repeat about its reach is not the number that is actually true. That business is bandwidth from orbit. A second business model is rising on top of it: compute from orbit. A public market has not yet said what either one is worth.
What was actually true this week
Adaspace, founded in 2018, already operates twenty AI computing satellites: twelve launched into a single orbital plane in May 2025, eight more in October 2026. [1][3] It holds orders for 114 more, at a combined contract value of approximately 3.48 billion yuan, about US$522 million. [1][3] None of that is the news. The news is what the company sells. Adaspace's Star-Compute plan links its satellites so they share computing capacity, storage and software, and it sells access to that shared capacity rather than the imagery its satellites collect. [1] TechNode's own reporting gives a worked example: a traffic-analysis customer sends its model to a satellite already in orbit. The satellite runs the analysis, and only the result, not the raw imagery, comes back down. [1] Earth observation has sold pictures for decades. Adaspace is selling the thing a picture used to require someone else to do with it, afterwards, on the ground.
The Star-Compute plan's stated target is 2,800 AI computing satellites in total. TechNode alone reports a first-batch milestone of 100 by the end of 2028, a figure this article could not independently confirm against a second source. [1] The IPO itself raises US$127 million. Roughly 36 per cent of proceeds is earmarked for research and development and 24 per cent for a Shenzhen satellite design, development and testing centre, again per TechNode's reporting alone. [1] That raise sits separately from an April 2026 private funding round that valued the company at approximately 11.55 billion yuan, about US$1.7 billion. [3] The two figures measure different events, a private valuation and a public raise. They should not be divided into each other to produce a ratio neither number was built to support.
The business model, not the financing event
I named the moment commercial actors treat Earth's regulatory frameworks as something to escape the Kardashev Pivot, in my book Space Mafia. I called the version of it that runs on concentrated private capital, rather than state ambition, Corporate Kardashev. Adaspace's IPO does not fit that pattern cleanly. Going public is the opposite of escaping a regulatory regime. Adaspace is entering a stock exchange's disclosure and listing rules for the first time, submitting itself to exactly the scrutiny a jurisdiction-arbitrage play is built to avoid.
The financing event is not what fits the Kardashev reading. The business model underneath it is. Running inference beside the sensor moves the compute function into a space no single jurisdiction's rules were written to reach. The alternative is downlinking the image to a ground station, which some data-export or sovereignty regime was built to govern. That holds whether or not the company lists anywhere at all.
Hold both halves. A company can submit to a public market's disclosure regime and still build infrastructure that sits outside every regime built to govern where compute normally lives. Those are two different things happening in the same filing.
The same inversion is already under way somewhere less visible than a stock exchange listing. As I wrote in Space Mafia, SpaceX's third-generation Starlink satellites already carry inference and training capability of their own. Elon Musk told investors in December 2025 that Starlink is "compute at the edge of the atmosphere," quoted in Space Mafia. Adaspace did not invent the orbital-compute thesis. It is the first company asking a public market to put a number on it.
Twelve thousand kilometres away, Amazon Leo is executing a thesis it has already financed. LV-01, its first flight on ULA's Vulcan rocket, is targeting 13 November 2026 at 17:54 UTC from Space Launch Complex 41 at Cape Canaveral. The flight carries forty satellites, the largest single Leo payload to date, on the first flight of a shortened Centaur V upper stage. [5][6] The mission opens a thirty-eight-mission contracted partnership between Amazon and ULA, under which Vulcan is intended to fly close to half of the first-generation constellation. [5] Amazon Leo stood at 396 satellites in orbit as at 9 November 2026. [6] One company is asking whether a business model works. The other is asking whether a rocket does.
Get the regulatory backdrop precisely right, because a drifted number is sitting in public view this same week. The US Federal Communications Commission requires 1,618 of Amazon Leo's first-generation satellites deployed by 30 July 2026 and all 3,236 deployed by 2029, under Order DA-26-553. [9] A lower pair of figures, 1,616 and 3,232, circulates widely in secondary coverage of the same order, two and four satellites short of the Commission's own numbers. The gap is not a rounding convention. It recurred again in reporting published this same week, independent of whatever first introduced it. A figure that drifts quietly in enough places stops reading as an error and starts reading as the record, which is exactly the kind of claim worth checking against the primary order rather than against whoever repeated it most recently.
What this means for New Zealand organisations
New Zealand's own stake in this shift runs through neither Adaspace nor Amazon Leo. It runs through Starlink, a constellation that already dominates a market segment most organisations assessing their own dependency would describe incorrectly. Starlink held 27 per cent of New Zealand's rural broadband connections in the year to 30 June 2025, up from 19 per cent the year before. It is now the country's largest rural broadband provider. [7][8] That is not the same claim as holding a fifth of the country's broadband overall. Starlink's share of New Zealand's total fixed broadband market, rural and urban combined, sat at approximately 2 per cent over the same period. That was built on roughly 85,000 subscribers nationally, 72,000 of them rural and 13,000 urban, up from about 58,000 subscribers the year before. [8] A dominant position in one under-served segment and a marginal position nationally are both true at once. A board briefing that quotes only the 19 or 27 per cent figure, without naming which market it describes, is quoting a true number about the wrong population.
New Zealand's Outer Space and High-altitude Activities Act, the GBSI Act in this series' own shorthand, is not the instrument in play in either of this week's headline stories. It governs ground infrastructure, launch and payload authorisation through the New Zealand Space Agency. Neither a Hong Kong listing nor a Cape Canaveral launch engages it. An organisation checking its own GBSI Act obligations this week learns nothing new from either headline.
What does change, for a board or an architecture team weighing dependence on a foreign-controlled satellite operator, is the shape of the question worth asking. That question arrives once the operator starts selling compute as well as connectivity. A constellation that already carries a rural market by a wide margin, on bandwidth alone, inherits a second kind of hold over its customers. That hold begins the moment it also starts running inference beside the sensor, rather than only relaying a signal through it. The dependency changes the moment the same operator decides to sell a second product over the same infrastructure. Nothing about a broadband contract obliges the operator to announce that decision in advance.
Four questions worth asking before either public market answers its own, not after:
- Do you know your organisation's actual dependency on a single satellite operator, scoped to the right market, not the one most often repeated.
- If that operator starts selling compute time instead of only bandwidth, does your exposure change in kind or only in scale.
- Who, outside the operator itself, can independently check a claim about what an orbital compute node actually processed.
- Has your board reviewed this dependency since the operator's own business model last changed, or only since it was first adopted.
What to watch
Adaspace's own listing will answer the first open question by Thursday: whether a public market prices an unproven orbital-compute thesis as a going concern or as a speculative bet with a satellite attached. Amazon Leo's LV-01 will answer a narrower one: whether Vulcan's return to flight, and the thirty-eight-mission relationship riding on it, clears its first hurdle on schedule. Neither answer was available when this article was filed. Both will be public knowledge well before most readers reach this sentence. Check the outcome before repeating either as settled.
A business model is about to be priced by a public market for the first time. The market's verdict is not public yet. Check it against the exchange's own record before you repeat whatever headline runs first.
There is an open-source dimension worth naming here. Adaspace and Amazon Leo both run proprietary command and control over their own constellations. Neither inversion, compute sold instead of imagery, bandwidth sold at a growing rural share, depends on anyone outside the company checking what a satellite is actually doing. Open ground-segment tooling answers that question from outside. OpenC3 COSMOS, a community-maintained, inspectable command-and-control system, lets an operator send commands and read telemetry through code a third party can audit. The Two-Line Element standard, published through the US Space Force's own Space-Track.org, lets anyone independently verify where an object actually is without taking an operator's word for it. As compute ownership consolidates into a handful of operators, the layer that lets an outside party check what is happening stays open in a way the compute riding on top of it does not.
The same inversion that interests Adaspace's investors interests defence planners for a different reason. The Pentagon's Defense Innovation Unit has run a pilot under the NEW HORIZON Act, inviting commercial orbital-data-centre operators to bid for government workloads. The pilot treats processing power in orbit as a procurement category rather than a telecommunications afterthought. Adaspace is a Chinese commercial operator with close ties to state industrial policy, reportable here on its own published filings rather than on any interpretation of intent. Amazon Leo's launch vehicle family also flies the United States' national security missions. Neither fact resolves this week's open questions. Both are a reminder that a business model priced by a stock exchange and a capability evaluated by a procurement office are reading the same infrastructure for different reasons. A small country weighing dependency on one foreign-controlled constellation inherits that same double reading, regardless of which flag the operator flies.
Has your organisation ever mapped which of its dependencies would look different the day the provider changed what it sells?
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] TechNode. "Adaspace Put AI on Satellites. Now It Wants to Sell Computing Time." 10 November 2026. https://technode.com/2026/11/10/adaspace-satellite-ai-computing-hong-kong-ipo/
[2] Caproasia. "China AI Satellite Company ADA Space Hong Kong IPO to Raise $127 Million with Expected IPO Listing on 13th November 2026, Founded in 2018." 7 November 2026. https://www.caproasia.com/2026/11/07/china-ai-satellite-company-ada-space-hong-kong-ipo-to-raise-127-million-with-expected-ipo-listing-on-13th-november-2026-founded-in-2018/
[3] BigGo Finance. "ADA Space, Valued at 11.55 Billion Yuan, Clears HKEX Hearing with 114 AI Satellite Orders in Hand in Sprint to Hong Kong Listing." 30 October 2026. https://finance.biggo.com/news/1052e323-c5da-4e04-9565-2a14b91c55da
[4] South China Morning Post. "Chinese Space-AI Tech Hopeful Adaspace Faces Wary Investors in Hong Kong IPO." https://www.scmp.com/business/china-business/article/3373830/chinese-space-ai-tech-hopeful-adaspace-faces-wary-investors-hong-kong-ipo
[5] Amazon. "Amazon Leo LV-01: 40 Satellites in the First Vulcan Launch." https://www.aboutamazon.com/news/amazon-leo/amazon-leo-vulcan-satellites-launch
[6] Orbital Radar. "Orbital Briefing, 9 November 2026." https://orbitalradar.com/blog/orbital-briefing-2026-11-09
[7] Commerce Commission New Zealand. "Rural Connectivity: A Market Reset in Motion." 2026. https://www.comcom.govt.nz/news-and-media/news-and-events/2026/rural-connectivity-a-market-reset-in-motion/
[8] NZ Herald, Tech Insider. "How Elon Musk's Starlink Earned $100m From New Zealand in 2025." https://www.nzherald.co.nz/business/companies/telecommunications/starlink-revenue-surges-as-sparks-market-share-slips-in-commerce-commission-report-tech-insider/premium/USJQEPRQWBFEPAWCEXPP7A3RYI/
[9] Federal Communications Commission, Space Bureau. "SB Acts on Kuiper's Request to Extend or Waive Interim Milestone." Order DA 26-553. 5 June 2026. https://www.fcc.gov/document/sb-acts-kuipers-request-extend-or-waive-interim-milestone

