Compute vs concrete: For investors, not a binary choice

IN THE last decade, markets fell in love with companies that travelled light. No factories. No cables. No concrete. Just code, scale and margins that seemed to expand by themselves.

That world view is starting to crack.

The reason can be summed up in a single word that is quietly returning to the global vocabulary: autarky.

At its simplest, autarky means self-sufficiency – the desire by nations and corporations to control the systems they depend on, rather than assume that global supply chains will always hold. However, in practice, this is less about isolationism and more about deliberate recalibration. And we see this reshaping how capital is deployed.

In an autarkic world, the race is no longer only about building the smartest software, or “compute”. It is also about securing the physical systems, or “concrete”, that software cannot exist without.

A false choice

Investors should resist framing the next phase as a binary choice between “compute” and “concrete”. There will be winners on both sides. Software remains the orchestration layer of the modern economy. But the more capable that software becomes, the more it leans on hard foundations underneath it.

Data centres do not run on air. They run on electricity, cooling systems, substations, transformers, cables and backup power. Intelligence is fast becoming digital, but capability remains brutally physical.

Artificial intelligence was first sold as a story about models, chips and applications. Increasingly, the real constraint is elsewhere: throughput, latency, energy availability and reliability. This is a real issue today, as I now occasionally find my paid premium AI subscription degraded to a more subpar standard (slower and less thorough).

As software evolves from responsive tools into persistent, agent-like systems, the bottleneck shifts from what code can do to whether the physical system can sustain what the code wants to do. You cannot prompt your way around power shortages, grid congestion or transformer bottlenecks.

Follow the power

Demand for power is no longer concentrated in a single fashionable corner of the market; it is spreading across an entire vertical stack.

Software needs more compute. Computing requires data centres. Data centres need electricity. Robotics need charging, connectivity and real-time inference. Defence systems are becoming more electrified, autonomous and sensor-dense. The ever-increasing number of electric vehicles need charging.

Different sectors may tell different stories, but many end up pulling on the same thread: more power, more resilience and more control over networks.

That is why we see that “electrons” – power and electricity – have become one of the most useful lenses through which to view the next investment cycle.

The quiet bottlenecks

The most interesting opportunities often sit one layer further out.

These second-derivative beneficiaries are not the utilities or the data centres themselves, but the companies supplying the parts that make electrification reliable, scalable and investable. These would include transformer components, switchgear, cooling systems, and the materials that keep all of these running.

For investors, the test is practical. Look for businesses with qualification moats, pricing power and products on a critical path. If a grid upgrade, a data centre build or an automation roll-out cannot proceed without them, they deserve attention.

These may not be the loudest stories in the market, but they are often where bottlenecks become margins.

Autarky sharpens this dynamic. Governments and companies are now far less comfortable outsourcing strategic dependencies to long, brittle global chains.

We acknowledge that the global supply chain remains very integrated, and so full self-reliance is well-nigh impossible. However, selective control over semiconductors, energy systems, critical inputs and infrastructure integrity appears to be the order of the day.

In that environment, hard assets gain a second form of value. They are not just economically productive; they become strategically important. Owners of reliable power assets, qualified components or constrained grid infrastructure often gain pricing power born of indispensability.

Hard-to-price tail risks

None of the above makes software yesterday’s story; it’s quite the opposite.

One reason that “concrete” is back in focus is that “compute” is becoming more powerful. As frontier application AI systems such as Mythos from Anthropic or those from rival labs become more capable, they also introduce a tail risk that markets are only beginning to grasp.

In the wrong hands or without adequate controls, highly autonomous systems could stress energy networks, cyber defences and critical infrastructure faster than institutions can respond.

That makes resilience, safeguards and power security more valuable, not less.

The other tail risk is that of our intelligence. As the “cost of intelligence” approaches zero, it is inevitable that more people will subcontract their critical thinking to AI models. Will our children be raised as slaves to the chatbot?

Benefiting from a handshake

For investors, the takeaway is straightforward: Do not build portfolios as though the future belongs only to asset-light digital champions or only to heavy industrial incumbents. That is a false choice.

The more useful question is where digital ambition collides with real-world constraint. Wherever software adoption forces a physical upgrade – in electricity, networking, cooling, automation or secure capacity – value can accrue on both sides of the interface.

If the last cycle taught investors to respect the scale of software, the next may teach them to respect the limits of the physical world.

Code can optimise and automate. But it still needs a grid, a wire, a transformer and a reliable flow of electrons. The strongest portfolios will not choose between “compute” and “concrete”. They will own what benefits from the handshake between these domains.

The writer is head of investment strategy, UOB Private Bank