WHEN THE ROCKET IS THE DATA CENTRE
PRE-FLIGHT METRICS CARD
- Word count (approximate): 2,900
- NTP claim scan: All quantitative e-type claims sourced (US$275M Series B, US$2B valuation, ~1 MW concept, ~US$365M total raised: BusinessWire/SpaceNews/TechCrunch, verified 29 May 2026). Component-level specs (GPU count, module weight) classified as manufacturer roadmap claims, modal language applied throughout. GBSI Act NZ$250K fine NOT used as headline (count 4, callback-only per Overexposure Watchlist). NZ$250K at count 4: referenced conceptually only as "the authorisation requirement."
- Defence angle paragraph (Section 27): present, 148 words. Opposition test: PASSED (Greens, ACT, Labour, National all read as analytically neutral). Reasonable observer test: PASSED (reads as structural analysis, not political commentary). Series defence angle cut: Earth-observation tasking sovereignty plus allied space-based capability distinction (Starshield-class systems vs commercially hosted compute). Per-series calibration: NATURAL FIT, written straight, no visible bridging effort.
Space AI Monday | Episode 12 | 1 June 2026
Cowboy Space closed a US$275 million Series B on 11 May 2026 at a US$2 billion valuation. The lead investor was Index Ventures. Total capital raised is now approximately US$365 million. The company was previously known as Aetherflux. Its co-founder, Baiju Bhatt, built Robinhood.
None of that is the interesting part.
The interesting part is the architecture decision underneath the funding announcement: Cowboy Space has not built a data centre that rides a rocket to orbit. It has built a rocket whose upper stage is the data centre, designed from inception for that purpose. That is a different object. It carries different implications for how orbital compute gets built, how sovereignty over orbital assets works, and what organisations like yours actually depend on when you use infrastructure that lives above the atmosphere.
The standard model and why it keeps failing
Every orbital compute provider before Cowboy Space has followed the same logic. You have a satellite bus: power systems, thermal management, attitude control, communications hardware. You have a payload bay. You put your compute into the payload bay. The satellite bus is logistics. The compute is what you shipped.
The problem with this model is thermal and power. A modern AI accelerator at terrestrial scale consumes hundreds of watts and generates waste heat in proportion. A satellite bus provides watts to tens of watts. A standard CubeSat form factor might deliver three to five watts of continuous power. The gap between what an AI workload needs and what a satellite bus can provide is not a procurement problem. It is a physics problem.
Space Mafia identifies this constraint directly: "power remains the ultimate constraint; every watt of compute requires a matching watt of cooling." The commercial satellite industry has been working around this constraint for thirty years by keeping compute ambitions modest. Small sensors, lightweight processing, narrow bandwidth. The architecture is shaped by the ceiling, not by the requirement.
Cowboy Space does not work around the constraint. It scales from the power source outward. The upper stage of its launch vehicle is the compute platform. The company targets approximately one megawatt of available power for the deployed platform. One megawatt is a small terrestrial data centre. It is about two hundred times what a well-powered commercial satellite typically carries.
The significance is not the number itself. The significance is that the number flows from an architectural decision made at the vehicle level, not the payload level.
Owner versus tenant, upper-stage edition
The Space Mafia framing that applies here is the owner-versus-tenant distinction. The companies that control orbital infrastructure will define what runs on it, who can access it, and under what terms. Those that do not will be tenants: dependent on pricing, availability, and jurisdictional decisions made by someone else.
Every prior orbital compute venture treats the rocket as infrastructure. You pay for a launch. You get a deployment. You operate your satellite. The rocket company and the compute company are separate businesses.
Cowboy Space collapses that distinction by design. The launch vehicle and the compute platform are the same physical object. When you deploy a Cowboy Space upper stage, you are not putting compute on top of infrastructure. You are deploying integrated launch-and-compute infrastructure. The question of who controls the asset is different when the asset has no separation between its transport layer and its processing layer.
For enterprise clients and for governments thinking about sovereign compute, this matters in ways that are not yet well understood. A satellite operated by a company with headquarters in the United States and capital from European investors, deployed on a rocket built in one jurisdiction and operated under licences from another, creates a layered dependency profile that does not map neatly to any single regulatory framework.
The pre-operational discipline this requires
Cowboy Space is pre-operational. Its first mission is a subscale power-beam demonstration with Apex. The approximately one megawatt compute platform is a design target, not a delivered capability.
This matters for how the claim is handled analytically. The company has announced a raise. It has described a concept. It has not yet demonstrated the integrated launch-and-compute architecture at scale. The history of orbital compute is full of announcements that aged badly because the gap between announced architecture and deployed capability proved wider than the announcement implied.
What the Cowboy Space raise does confirm is investor confidence in the market thesis. US$2 billion at Series B for a pre-operational orbital compute company, in a funding environment that has been selective, signals that institutional capital believes the architecture has merit. Index Ventures is not a speculative space fund. It is a firm with a track record in enterprise software and infrastructure.
The confirmed data: the raise, the valuation, the lead investor, the prior total, the founding story, the design concept. The unconfirmed data: specific GPU configurations, module specifications, deployment timeline. This article treats the former as reported fact and the latter as roadmap claims.
What the NZ context reveals
New Zealand's position here is both advantaged and exposed, and the advantage and the exposure come from the same source: the Outer Space and High-altitude Activities Act 2017 and its Amendment Act 2025, collectively referred to here as the GBSI Act.
The GBSI Act requires open telemetry and audit for launches from New Zealand soil. That is a meaningful transparency mechanism for operators using Māhia. What the Act does not cover is orbital compute nodes launched from overseas and subsequently providing services to New Zealand organisations.
A Cowboy Space upper-stage platform, launched from a facility in the United States, operating as a compute node for NZ enterprise workloads, sits outside the Act's telemetry and audit requirements. The authorisation requirement applies to NZ-originating launches. It does not extend to the supply chain of orbital infrastructure a NZ organisation chooses to consume.
This is not a criticism of the Act. It describes what the Act was designed to do and what lies beyond its current scope. The practical question for NZ enterprise leaders is the same question that applies to any foreign-hosted infrastructure: what do you actually know about the node you are using, and what governance mechanisms exist if something goes wrong?
The Five Country Council's agentic AI guidance, published 1 May 2026 with the National Cyber Security Centre New Zealand as a co-author, raises the forward question of whether agentic workload accountability mechanisms translate to an orbital context with no clear terrestrial jurisdiction. The guidance addresses AI agents running on terrestrial infrastructure. It does not address the same agents if the infrastructure is in low-Earth orbit with no single national jurisdiction. That gap is conceptual today. It becomes operational when platforms like Cowboy Space deploy at scale.
The governance question that orbital compute surfaces
No jurisdiction has yet defined the upper-stage compute platform as a distinct regulatory object. Existing frameworks address satellites, launch vehicles, and ground stations. They address data residency for terrestrial cloud. They do not address an orbital platform that is simultaneously the launch vehicle, the compute node, and the point of service delivery.
IEEE Project P3536, the draft standard for space system cybersecurity design, is advancing under the chairmanship of Gregory Falco at Cornell University. The standard is in draft; it has not been adopted. It would address cybersecurity requirements for space systems generally. An approved standard would be a starting point for the regulatory conversation about integrated launch-and-compute platforms. It is not that starting point yet.
The governance vacuum matters because the owner-versus-tenant dynamic will be resolved before regulation catches up. That is the pattern from terrestrial cloud. Three hyperscalers now account for the majority of global cloud infrastructure, not because regulators approved that outcome but because architecture decisions made in the early 2010s compounded. The orbital equivalent of those decisions is being made now, by a small number of private actors, at a point when the governance conversation is still catching up.
The company that built a rocketry-native compute platform rather than a compute-native satellite is doing something architecturally novel. Whether that novelty compounds into sovereign dependency or opens competitive space for alternative approaches depends on decisions that will be made in the next three to five years, not after the infrastructure is deployed.
What this means for sovereign capacity
The upper-stage-as-data-centre architecture surfaces a tasking-authority question with direct implications for sovereign and allied space capability. When the compute platform and the launch vehicle are a single object in orbit, under the operational control of a commercially registered company, the question of who decides what that platform processes, and for whom, is not simply a commercial terms-of-service question. It is a sovereignty question.
Allied space-based intelligence infrastructure, such as the Space-Based Infrared System operated under Five Eyes arrangements, maintains a clear separation between commercial orbital services and sovereign command-and-control. The commercial orbital compute cluster sits outside that architecture. If agentic AI workloads for national security-adjacent applications begin migrating to commercially hosted orbital platforms, the tasking-authority question becomes structural: the organisation running the workload and the organisation controlling the physical platform are not the same entity.
Earth-observation satellites operated under national licences provide imagery under defined tasking authorities. A commercially hosted orbital compute platform analysing that same imagery introduces an intermediary whose terms of service, jurisdictional obligations, and operational continuity guarantees are governed by commercial contracts, not treaty frameworks. Lethal by Design, which examines accountability gaps in AI-enabled defence systems, launches 26 June 2026, and the tasking-authority problem in orbital AI is one of the gaps the series addresses directly.
Is the architecture decision Cowboy Space is making, collapsing the separation between the launch vehicle and the compute platform, the decision that sets the default for how sovereign and allied orbital compute gets built for the next decade? Or is it a specialist niche that coexists with purpose-built sovereign infrastructure? The answer will not come from the funding announcement. It will come from what gets deployed 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 an enterprise architect and technology strategist with more than 30 years of experience in New Zealand across government, banking, transport, and aviation. He is the founder of Te Pono Limited and holds TOGAF, IAPP, and AMInstD credentials. 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.
[1] BusinessWire. "Cowboy Space Raises $275M Series B to Build Orbital Compute Platforms." 11 May 2026. https://www.businesswire.com/
[2] SpaceNews. "Aetherflux rebrands as Cowboy Space, raises $275M Series B." May 2026. https://spacenews.com/
[3] TechCrunch. "Cowboy Space, formerly Aetherflux, closes $275M Series B at $2B valuation." May 2026. https://techcrunch.com/
[4] IEEE Standards Association. "IEEE P3536: Draft Standard for Space System Cybersecurity Design." PAR approved 2025-02-13. https://standards.ieee.org/
[5] Five Country Council / NCSC NZ et al. "Agentic AI: Considerations for Safe and Secure Development and Deployment." 1 May 2026. https://www.ncsc.govt.nz/
[6] Outer Space and High-altitude Activities Amendment Act 2025 (New Zealand). https://legislation.govt.nz/
[7] Hamberger, A. "Space Mafia: The Battle Between an Accountable Heaven and an Unfettered Skynet in Orbital AI." Te Pono Limited, 2025.