Digital Is Physical Now - Infrastructure Is the Strategy
Still Thinking Series: Part 3
Building on insights from the McKinsey Technology Trends Outlook 2025 (July 2025), this post explores why AI’s future depends not just on algorithms, but on energy, supply chains, and physical systems we can’t ignore.
We like to think of the digital world as weightless.
Code. Cloud. Creativity. Speed.
But the more powerful our tools become, the more physical they get.
AI may feel abstract—but it’s built on hardware, housed in data centres, powered by grids, and constrained by geography.
We’ve spent years designing scalable software. Now we need to design scalable infrastructure.
Because digital ambitions don’t float above the real world.
They sit inside it.
The Cloud Has a Footprint
According to McKinsey, demand for data centre capacity is projected to more than triple by 2030, driven by AI training workloads, robotics, and immersive digital environments.
That’s not just a cloud problem.
That’s a land use problem. A power grid problem. A materials problem.
Even “lightweight” products now rely on:
Custom chips and application-specific semiconductors (ASICs)
Stable power supply for GPU clusters
High-speed networking, often in remote or underdeveloped regions
A global supply chain at risk from delay, scarcity, and geopolitics
The bottleneck isn’t intelligence. It’s infrastructure.
Every Digital Product Touches Ground
This brings us back to something I wrote a while ago in Digital Products and the Places They Don’t Reach. We often design with an unconscious assumption:
That the user is connected, powered, and cloud-ready.
But the real world is spikier than that.
Some users are on satellite internet.
Some live off-grid.
Some governments demand sovereign infrastructure.
Some countries can’t meet the energy demands of advanced AI at all.
When we forget these physical and political constraints, we don’t just build badly—we build unjustly. We exclude. We overreach. We design for a world that doesn’t exist.
Strategy Is Now a Systems Problem
In this new era, strategy must account for infrastructure:
Can your AI product run at the edge, not just the centre?
Is your cloud vendor subject to regulatory or geopolitical pressure?
What happens if compute becomes expensive or energy becomes scarce?
McKinsey notes that massive, centralized compute clusters still dominate—but distributed workloads and edge environments are rising. So is national interest in sovereign clouds, localized chip fabrication, and energy resilience.
These aren’t backend details.
They’re front-end design constraints.
And they’re strategic.
The Dirt Road Problem
A phrase I often come back to:
“From whiteboard to dirt road.”
It’s the gap between the world we design for, and the world we deploy into.
In an era where AI agents write our emails and route our supply chains, it’s easy to forget: those emails still require power. Those supply chains still move through borders, ports, and policies.
When the infrastructure fails, the intelligence stalls.
Closing Thought
We used to ask: Can it scale?
Now we need to ask: Can it survive?
In energy terms. In locality terms. In climate terms. In sovereign terms.
Because digital is physical now.
And the future doesn’t just run on code.
It runs on copper, coal, silicon, sunlight—and the strength of the systems we build beneath it.