In Old Norse, smiðr didn't mean someone who hits metal in a shed. It meant the person who understood material, process, and function as one integrated act. A shipbuilder was a smiðr. A weaponmaker was a smiðr. The person who looked at birch, iron, and water and saw a vessel that could cross the North Atlantic. That was a smiðr.
No separation between design and make. No gap between knowing and doing. The smiðr held three things in one head: the conviction that something needed to exist, the capital understanding of what it would cost and what it was worth, and the capability to actually make the thing work in the real world.
That integration is the point. Not a manager who coordinates specialists. Not a consultant who writes frameworks. A single mind that holds conviction, capital, and capability together, because separating them is what breaks the work.
The split
Industrialisation brought scale. Scale brought specialisation. Specialisation separated those three things into different buildings, different budgets, different people who increasingly couldn't speak each other's language.
Strategy: conviction without capability or capital
Finance: capital without conviction or capability
Engineering: capability without conviction or capital
Each function optimised locally. None owned the integration. The role of the smiðr, the person who couples all three, quietly vanished from the org chart. We stopped even having a word for it. "Manufacturing engineer" is the closest modern English gets, and most boardrooms treat that as a back-office function.
The lineage that survived
One region kept more of the smiðr DNA intact than anywhere else. Not by accident. By structure. The Nordic industrial tradition is built on families and institutions that understood something the anglophone world forgot: you don't separate the person who knows how to make something from the person who decides whether to fund it.
The Wallenberg family didn't invent the capability. They built the capital architecture around a society that already had it in the soil.
Sandvik. Steel that understood its own grain. Material science wedded to process from day one.
Atlas Copco. Compressed air and vacuum. Infrastructure for everyone else's factory.
SKF. The self-aligning ball bearing. One component that the entire mechanical world depends on.
Volvo. "It rolls." Built around a safety conviction that preceded the engineering, then became inseparable from it.
Saab. Aircraft engineers who decided to build cars. The conviction that if you can make a fighter jet, you can make anything.
Every one of these: conviction about what matters, capital structured for the long term, capability that understood material and process as one thing. The smiðr, institutionalised.
The Great Asymmetry
Now billions in capital are flooding toward "physical AI." Robots, digital twins, smart factories. The conviction is there. The money is there. But the rooms where these decisions are being made have a structural absence.
The people buying AI for industry don't know what they're buying. The people selling it don't know what it needs to connect to. And the people who actually understand the physical systems, the manufacturing engineers, aren't in the room.
One side of the room: VCs, sovereign funds, corporate M&A, government grants
The other side: process engineers, toolmakers, manufacturing systems, shop floor knowledge
In between: a void filled with consultants who've never held a tolerance
That void is where value gets destroyed. Not because the AI doesn't work. Not because the factory is incompetent. Because nobody in the conversation can read both sides of the interface and say, honestly, where you are versus where you think you are.
The infrastructure pattern
Look at the people who built transformative companies in the last three decades. None of them sold the thing. All of them became the layer that everyone else's thing depends on.
Bezos didn't sell books. He built the logistics and compute infrastructure everyone else sells through. Books were proof of concept. AWS was the product.
Musk didn't sell cars. He vertically integrated around the bottleneck everyone else outsourced. The Gigafactory is the competitive advantage, not the vehicle.
Huang didn't sell chips. He built CUDA, the ecosystem lock-in that means every AI workload on earth passes through his tollbooth. The GPU is the hardware. CUDA is the moat.
The pattern: find the bottleneck that gets worse as the market grows. Build the infrastructure at that bottleneck. Own the coupling layer.
The bottleneck that only gets worse
As AI floods into physical industry, the constraint isn't the AI. It isn't the machines. It's the coupling between them. The translation layer where someone has to say: this model connects to this process in this way, and here's what breaks if you get it wrong.
Nobody owns that interface. Everybody needs it. And demand for it scales with every dollar poured into physical AI.
This week, Nordic Compass launched. Jacob Wallenberg, alongside the chairs of KONE, Aker, Novo Nordisk Foundation, and Nordea, publicly declaring that Nordic industrial capability needs a new integration layer across capital, energy, technology, and defence. That's the demand signal. The compass. What they don't yet have is the survey.
The people who can provide that survey, who can stand in the void between capital and capability and give a calibrated, honest signal, are the rarest people in the system. They are the modern smiðr. And the question facing industrial AI is not whether we need them.