Join the mailing list Free weekly briefing, no spam.
Cover Image: ESA seeks software ideas to bring smart satellites to life – Credits ESA
In Part 1, I argued that the bottleneck in new space is no longer the rocket (despite latest Falcon 9 updates) or the bus, but an operational paradigm that treats autonomy as a risk to be managed rather than a capability to be built. Better procurement would help. It would not be enough.
The structural problem is not what you fix with a policy document
The transition from hardware-centric to software-centric space architecture requires a different kind of organization. Not a different procurement framework: a different base.
Capital is part of the problem. European space ventures raised €1.4 billion in private investment in 2025, against a global total of €11.7 billion, almost €8 billion of which went to US companies. Europe’s figure fell 8% year on year while the global one grew 60%. The gap is not mainly at the early stage: there is reasonable seed and Series A activity. The critical gap is growth capital. In 2025, not a single growth-stage round in Europe was led by a European private investor, and the three largest private deals of the year were about €150 million, €150 million and €90 million. The investments required to build the engineering teams, the compute infrastructure, the in-orbit test campaigns and the operational stacks that turn a proof of concept into a deployable system are billion-class. Those investments are happening in the United States. They are beginning to happen in China, which in May 2025 launched the first 12 satellites of a planned 2,800-satellite distributed AI computing constellation, an orbital supercomputer in the making. Europe is starting to react (European space companies raised €1.6 billion in the first half of 2026 alone), but it is reacting from far behind.
Talent is the less-discussed but equally structural problem. European space engineering expertise is concentrated in hardware: systems engineering, AIT processes, test campaigns, radiation hardening programs. The skills data points the same way. In the 2023 Space Sector Skills Survey, which the European Commission itself cites, software and data were the most frequently reported gap, flagged by 72% of the organizations with skills gaps, well ahead of electronics design (43%) and systems engineering (39%). The engineers capable of building autonomous constellation management software, onboard inference pipelines and distributed edge architectures are being recruited by companies in adjacent sectors that often pay more and move faster, and 45% of the surveyed organizations report difficulty retaining staff.
The engineers who remain work within procurement frameworks that too often impose ECSS standards and CCSDS protocols even when they are not the answer, with focus on interface definitions and development methodologies designed to minimize risk in hardware integration. These frameworks are not irrational for their intended purpose. They are poorly adapted to iterative software development in a domain where the cost of a failed experiment in orbit is now low enough to be acceptable.
You cannot build a cognitive system purely on a waterfall.
The window, and what closes it
The commercial business case for space autonomy software is still being written. Constellation management platforms are emerging. Onboard inference is being demonstrated. The unit economics of autonomous EO retasking have not been fully proven at scale in competitive commercial markets.
This argument is correct. It is also largely beside the point.
One of the real drivers for the next phase of new space architecture is sovereign defense. Defense accounted for $73 billion of the $135 billion governments spent on space in 2024, and still about half of global public space budgets in 2025. Europe depends often on others for what that money buys: European states operate only 17% of NATO’s military satellites, and much of the access Europe relies on comes from the United States. Starlink’s role in Ukraine made this dependency visible in a way no strategy document could. A commercial constellation operated by a private American company had become a geopolitical instrument.
Every European defense ministry read that, and the budget implications are predictable: Germany has announced €35 billion for space security and defense through 2030, and France has added €4.2 billion for military space.
The question is not whether the demand exists. The question is who builds the capability that captures it. And the answer has a time component.
Software stacks, once deployed and integrated into operational doctrine, create lock-in. The actor who defines how autonomous tasking works, how constellation-shared situational awareness is structured, how sensor fusion pipelines handle conflicting inputs, how edge decisions are logged and audited in a way that satisfies military accountability requirements, that actor sets the standard. Subsequent entrants either integrate into that framework or rebuild at full cost, without the operational data that comes from years of deployment.
A small number of actors are already building these stacks, most of them American. These are demonstrations and young products, not yet doctrine. But they are the start of an accumulation that is hard to reverse: every hour of operation produces data, trust and procedures that later entrants will not have.
The asymmetric advantage is not a future risk. It is a present trajectory.
What this actually requires
The transition from hardware company to software company in the space sector is not a product roadmap decision. It is an identity decision.
It requires hiring for software architecture and machine learning rather than only for systems engineering and thermal analysis. It requires funding structures that support rapid iteration and in-orbit experimentation rather than milestone-driven development gates. It requires a tolerance for a specific failure mode, a satellite that takes an unexpected autonomous action, that the existing institutional culture is not designed to absorb and that regulatory frameworks have not yet learned to accommodate.
It requires, most fundamentally, accepting that the satellite is no longer the product. The satellite is the substrate. The product is how the satellite operates.
The companies that will define the infrastructure of new space are not the ones with the most nodes in orbit. They are the ones that understand what those nodes should do when nobody is watching.
The hardware problem is mostly solved. The paradigm problem is not. And the window to own the solution is narrower than the launch manifests suggest.
Frequently asked questions
What is the main argument of this article?
The bottleneck in new space is organizational. Europe’s gap in autonomous space systems comes from limited growth capital and an engineering base concentrated in hardware, within development frameworks built to minimize hardware integration risk. Better procurement would help, but it would not close the gap on its own.
How much private capital did European space companies raise in 2025?
European space ventures raised €1.4 billion in 2025, against a global total of €11.7 billion. The European figure fell 8% year on year while the global one grew 60%. The critical gap is growth capital: no growth-stage round in Europe was led by a European private investor.
Which skills does the European space sector lack the most?
Software and data. In the 2023 Space Sector Skills Survey, 72% of the organizations with skills gaps flagged software and data, well ahead of electronics design (43%) and systems engineering (39%). Also, 45% of the surveyed organizations report difficulty retaining staff.
Why does defense drive demand for space autonomy software?
Defense accounted for $73 billion of the $135 billion governments spent on space in 2024. European states operate only 17% of NATO’s military satellites, and Starlink’s role in Ukraine showed how a commercial constellation can become a geopolitical instrument. Germany has announced €35 billion for space security and defense through 2030, and France has added €4.2 billion for military space.
Why does timing matter for the software layer?
Software stacks create lock-in once they are integrated into operational doctrine. The actor that defines how autonomous tasking and sensor fusion work sets the standard. Later entrants must integrate into that framework or rebuild at full cost, without the operational data that years of deployment produce.
What does it take for a space hardware company to become a software company?
The article calls it an identity decision. In practice it means hiring for software architecture and machine learning, funding rapid iteration and in-orbit experimentation, tolerating the unexpected autonomous action of a satellite, and accepting that the satellite is the substrate while the product is how it operates.



