Physical AI in Defence Satcom: Nokia's Claim Points in the Right Direction. The Engineering Evidence Doesn't Yet Follow.

On 1 October 2026, ICEYE and Nokia stood together in Helsinki to announce what they called a new era in sovereign satellite communications. The two companies committed to developing secure, sovereign, and high-performance broadband LEO satellite communications systems for governments and defence customers, covering missions from border security to emergency services globenewswire.com. The sovereign satcom market is moving from a niche to a core procurement category across European defence ministries, and this partnership lands squarely in that current.

The headline architecture is commercially coherent. ICEYE has already demonstrated it can hand over operational control of a radar satellite constellation to a national armed force in less than twelve months after contract signing. ICEYE delivered a four-satellite synthetic aperture radar reconnaissance constellation to the Polish Armed Forces under an approximately €200 million contract signed in May 2025 iceye.com. The system, operationally designated POLSARIS, was transferred to Poland's Geospatial Reconnaissance and Satellite Services Agency, known as ARGUS, on 15 May 2026 satellitetoday.com. The ICEYE-Nokia satcom offer applies the same ownership logic to communications infrastructure: purchasing governments would control the full stack, from the space segment to terminals satnews.com.

Then Nokia President and CEO Justin Hotard said something that stopped me. Hotard invoked "Physical AI" as the enabling technology for sovereign sense-decide-act operations across land, sea, air, and space globenewswire.com. Two words. Zero specifications. And this is exactly the credibility problem I want to take seriously, because it is increasingly the problem facing every European aerospace and defence founder pitching AI-enabled hardware to a procurement office.

The Legitimate Case for the Claim

Let me steelman Hotard's position before pushing back, because the underlying direction is real.

Nokia announced a leadership transition in February 2025, appointing Justin Hotard as the next President and Chief Executive Officer, with him starting in the role on 1 April 2025 sec.gov. He previously led the Data Center and AI Group at Intel sec.gov. His strategic orientation toward deterministic, AI-native network architectures is consistent and well-documented.

In November 2025, Nokia and Tesi, the Finnish state investment company, invested in NestAI, a Finnish defence AI startup. The combined value of the investments was €100 million, announced on 20 November 2025, in conjunction with a strategic partnership to advance AI-powered defence solutions nokia.com. NestAI works on physical AI, technologies that bring AI out of software and into real-world systems such as drones, autonomous vehicles, and command-and-control platforms, developing solutions for unmanned vehicles, autonomous operations, and C2 systems used in logistics, inspection, surveillance, security, and defence tesi.fi.

By July 2026, Nokia Defence and NestAI had announced the first operational capabilities from that collaboration rcrwireless.com. The first announced capability combines Nokia's deployable 5G networks with NestAI's NestOS operating system to support AI-enabled command and control and autonomous systems without relying on fixed infrastructure. The second integrates Nokia's radio-network planning models into NestOS mission-planning tools to allow operators to evaluate and adapt connectivity during multidomain missions nasdaq.com. This is not a press-release entity.

The ICEYE partnership extends that logic to LEO. The procurement template is already proven. Just one year passed between signing the approximately €200 million contract with Poland's Ministry of National Defence in May 2025 and the start of independent operations by ARGUS, the constellation's designated operator iceye.com. ARGUS was established by the Ministry of National Defence in 2024 and given the system its own operational name: POLSARIS, standing for Polish SAR Intelligence System aerotime.aero. That provenance matters. ICEYE's execution record is not rhetorical. And the connectivity problem Hotard is gesturing at is real: sense-decide-act loops genuinely do break down without resilient, sovereign communications.

Where the Claim Breaks Down

Here is the honest problem: the announcement provides no technical substance on what "Physical AI" actually does inside the system globenewswire.com. There is no description of which AI functions are embedded in the ground segment versus onboard compute. There is no mention of how the networking layer behaves under electronic warfare, jamming, or deliberate signal spoofing. There is no reference to any test programme, red-team exercise, or performance benchmark.

The "AI" in the announcement could refer to spectrum management heuristics, traffic prioritisation algorithms, or genuinely novel onboard inference. We cannot tell. And in 2026, that ambiguity is no longer acceptable to serious procurement offices.

Within the NATO AI Strategy, Allies endorsed six Principles of Responsible Use for AI in Defence: Lawfulness, Responsibility and Accountability, Explainability and Traceability, Reliability, Governability, and Bias Mitigation nato.int. The Explainability and Traceability principle is the one that bites hardest here. AI applications must be appropriately understandable and transparent, including through the use of review methodologies, sources, and procedures, with verification, assessment, and validation mechanisms at either a NATO or national level nato.int.

This is one of the most technically demanding principles for AI vendors because many high-performing AI approaches are inherently difficult to explain. NATO's expectation is not that all AI systems must use inherently interpretable approaches, but that vendors must be able to provide contextually appropriate explanations of system behaviour for the level of decision authority at which the system operates corvusintell.com.

The ICEYE-Nokia press release gives procurement offices nothing auditable to work with globenewswire.com. "Physical AI makes that possible" is a positioning claim, not a technical assertion. The risk is real: if a defence ministry signs a contract based on AI-inflected marketing and subsequently faces adversarial electronic warfare conditions, the gap between the claim and the capability will become visible at the worst possible moment.

The Context That Sharpens the Stakes

This is not an academic concern. The wider defence technology environment is moving fast enough that the credibility gap between AI branding and mission-validated AI performance is widening by the month.

On 23 September 2026, ESA initiated the selection phase for in-orbit demonstration missions supporting the European Union's planned military Intelligence, Surveillance, and Reconnaissance satellite network satnews.com. Rather than relying on fragmented national satellites or commercial Earth observation suppliers, the programme will deploy a coordinated LEO constellation capable of continuous optical, SAR, and signals intelligence collection. ESA has already secured €167 million for the work and is seeking to increase that to €350 million satnews.com. Participating European governments will formalise financial commitments ahead of the 30 November 2026 subscription deadline spaceintelreport.com. The sovereign satcom and ISR market ICEYE and Nokia are entering is becoming crowded with technically ambitious, well-funded programmes. In that environment, differentiation based on unsubstantiated AI language is a fragile commercial strategy.

On the institutional side, the Nokia-NestAI capabilities announced in July 2026 are the first products from a €100 million investment made in November 2025, reaching announced operational capabilities in roughly eight months rcrwireless.com. Eight months from investment to announced capabilities is fast by defence-industry standards, though announced capabilities and fielded capabilities are different things. That distinction matters precisely because the gap between announcement and operational proof is where marketing language becomes a liability.

Where I Actually Stand

I support applied AI in aerospace and defence, without reservation. Adaptive link management under jamming, onboard inference for priority routing, AI-driven anomaly detection in ground segment security: these are real engineering problems where AI delivers auditable, mission-relevant value. The sensor-to-decision pathway that Hotard sketches is a legitimate design objective for sovereign satcom.

But I draw a hard line between that and unverifiable positioning language deployed in a press release aimed at government procurement offices. The engineering capability is plausibly there. NestAI's co-founder Peter Sarlin has emphasised the company's goal to advance technological independence in Europe, drawing on a team that includes former employees from Intel, Kongsberg, Palantir, and Saab nokia.com. ICEYE built and launched four synthetic aperture radar satellites, including an initial three-satellite baseline delivered within ten months of contract signing, which ICEYE describes as the fastest operational deployment of a satellite programme on record iceye.com.

But if the capability is there, Hotard should be willing to specify it, even in general terms. Which functions run inference onboard? Against what electronic warfare threat model was the resilience architecture tested? What is the latency budget for a sense-decide-act cycle across the space-to-ground link under contested conditions?

These are the questions a serious defence procurement office will ask. They should have partial answers available in a product announcement aimed at defence ministries. The "Physical AI" label, on its own, adds zero information. And when labels substitute for evidence, they damage the credibility of every founder in this sector who is doing the harder work of building AI that can be tested, validated, and explained to a programme manager under scrutiny.

For Founders

If you are building AI-enabled hardware or software for defence satcom or space ISR in Europe, the ICEYE-Nokia announcement tells you something important: the primes and large integrators are now using AI language as table stakes. That means the language alone no longer differentiates you.

What does differentiate you is the audit trail. The NATO AI Strategy's six principles have generated concrete expectations that translate directly into procurement criteria for AI-enabled defence software nato.int. Your tender response will be evaluated against them. Here is what that means in practice for an early-stage team:

• Document your model provenance now, before the contract exists. Training dataset lineage, version histories, and architecture decision logs are not bureaucratic overhead. They are your compliance baseline.

• Build adversarial test records into your development cycle. A red-team result, even a modest one against a defined threat model, is worth more in a technical data package than any amount of capability language. Procurement officers are beginning to ask for it by name sipri.org.

• Separate what is AI from what is not. Procurement offices cannot evaluate claims they cannot locate. If your system uses AI for spectrum management but conventional methods for routing, say so precisely. Clarity beats ambition in a source selection lieber.westpoint.edu.

• Position against the audit gap, not the capability gap. Nokia can claim "Physical AI." You can prove where, how, and against what standard yours works. That asymmetry is your competitive moat.

The proof points in this sector are real. The approximately €200 million POLSARIS contract was signed in May 2025, with independent operations beginning one year later under ARGUS iceye.com. The Nokia-NestAI partnership was announced on 20 November 2025 with combined investments totalling €100 million, and reached its first announced operational capabilities by July 2026 2 sources. The gap between those proof points and the "Physical AI" label in the satcom announcement remains unbridged, and that gap is where your opportunity sits.

The founders who win the next wave of European defence satcom contracts will not be the ones with the most compelling AI narrative. They will be the ones whose AI can be tested, traced, and explained to a programme manager under scrutiny. With European governments formalising financial commitments ahead of the 30 November 2026 subscription deadline for the EU's military ISR constellation spaceintelreport.com, the procurement window is open now.

The ICEYE-Nokia partnership is a meaningful development for European sovereign satcom. Whether the AI inside it is meaningful, we will only know when someone publishes the test results.