A few weeks on from the Munich Security Conference, something many of the Resilience Media community no doubt attended, I keep returning to three things. Beyond immediate headlines, which can age quickly in a space like this, I want to understand the underlying patterns that I believe will still be true in five years.
So I went in with a manufacturing and industrial lens. That’s my background, and it shapes how I hear conversations about defense and resilience. What struck me most is how consistently the hard constraints came back to the same problems I’ve watched trip up hardware companies for years: supply chain depth, production scale, and whether the infrastructure underneath actually holds. Here’s what’s sitting with me:
Hidden infrastructure, strategic consequences
A general said something early in the conference that framed everything that followed: non-experts talk strategy, experts talk logistics.
Ukraine made this concrete. The limiting factor wasn’t doctrine or political will. It was production rates – whether shells, drones, and components could be manufactured and replenished faster than they were consumed. When the answer is no, deterrence erodes regardless of what’s been agreed on paper.
The conversations in Munich about European defense spending and procurement reform were largely focused on platforms – which systems to build, which programs to fund. That’s the visible layer. What got less attention was the mission-systems layer underneath: energetics, guidance subsystems, sensors, and communications components. This is the layer that determines surge capacity. It’s fragmented across Europe, undercapitalized, and rarely makes it into the headlines. It’s also the layer that replenishment actually depends on.
Treating the supply chain as a strategic asset – not just a cost center – is where the real work is. Companies like Onodrim are consolidating this fragmented landscape, aggregating mission-critical suppliers into something that can actually respond at scale. That kind of industrial infrastructure is as consequential as the platforms above it, and builds far more slowly.
A test flight is not a production line
This is the one I felt most directly, because I’ve lived a version of it.
After our renewable energy software company was acquired, I spent time scaling deployments across utilities. The product worked. The challenge was everything else – integration complexity, operational change, the gap between a pilot that succeeds and a rollout that actually sticks. Many of the failures we saw in that period weren’t technical. They were about the distance between “proven in the lab” and “running reliably at scale.”
Defense is the same problem, with less room to iterate and so much higher consequences when things slip. And so in Munich, the question I kept asking was: what does the path from that test flight to thousands of units in inventory actually look like? Who are the suppliers? What does the production line look like at unit 5,000?
Ahead of MSC, Hypersonica hypersonic missile prototype completed its first test flight in nine months. That’s a real achievement – very few teams in Europe have the physics, systems, and execution depth to get there so fast.
I wrote about this gap – the second valley of death between prototype and production scale – in my last piece for Resilience Media. What gave me some encouragement in Munich was that more conversations had moved past “the technology works” and into delivery timelines and supplier depth. That’s progress. A defense company that can’t answer production rate questions isn’t really a defense company yet. It has a prototype and a roadmap.
AI doesn’t layer on top – it has to be built in
The framing that keeps getting used – add an AI layer to an existing system, integrate intelligence into a legacy platform – is challenging. What I heard in Munich, and what I think is right, is that AI is now infrastructure. Not a product feature, not an upgrade path. Infrastructure, in the same category as power and supply chain.
The practical reason is decision speed. Modern defense systems live or die on the latency between sensing and acting. If the intelligence layer is bolted on after the hardware architecture is set, you’re fighting physics. The systems being built by peer competitors aren’t designed that way – they’re AI-native from the ground up, which means the gap isn’t just technological, it’s structural.
For builders, the practical implication is that retrofitting is hard. The companies that designed with and embedded AI in from the start – where the intelligence layer shapes the hardware architecture, not the other way around – have a structural advantage that grows over time. Helsing is the clearest European example: built around sensor fusion and decision speed as core functions, not as add-ons to a hardware product.
The piece that came up less in Munich but matters just as much is energy. AI infrastructure at defense scale requires reliable, competitively priced power. Europe’s energy constraints are a direct ceiling on what its industrial base can produce. You can’t manufacture satellites, advanced munitions, and autonomous systems at volume without solving the power problem underneath. Energy flexibility isn’t separate from defense capability – it’s part of the same stack.
The rearview mirror view of Munich: the alignment on urgency was real. Whether the institutional response – procurement reforms, European-wide investment frameworks, production scaling – moves at the pace the situation demands is the question that will define the next few years.
You can read more on this in General Catalyst’s recently published thesis: From Orbit to Factory Floor: Europe’s Resilience Stack.
Robin is an investor at General Catalyst, where he leads the firm’s resilience thesis in Europe with a focus on seed-stage investing across manufacturing, energy, and defence. A former founder, he built and sold a renewable energy software company to Kraken Technologies and later led global implementations for utilities, giving him deep, hands-on experience scaling critical systems. The views and opinions expressed in this article are solely those of the author and do not necessarily reflect the views of the author’s employer or any affiliated entities. This article is provided for informational purposes only and should not be construed as investment advice, a recommendation, or an offer or solicitation to buy or sell any securities or financial instruments. Any references to market performance or investment strategies are general in nature and may not be suitable for all investors. Readers should consult their own financial, legal, or tax advisers before making any investment decisions.
