The Cycle of Hardware

Welcome Back to XcessAI

Recent market moves in software stocks have been framed as fear.

Fear of new AI tools.
Fear of disruption.
Fear that entire categories of software may no longer be worth owning.

That framing is understandable, but incomplete.

What we’re seeing is not a sudden loss of confidence in software. It’s the re-emergence of a familiar cycle. One that appears every time a new layer of abstraction becomes too powerful, too cheap, and too abundant.

When intelligence commoditises, value migrates.
And it almost always migrates downward.

Software’s old advantage: extrapolation

For more than a decade, software enjoyed a unique position in capital markets.

Recurring revenue.
Low marginal costs.
Predictable growth.
Long-duration cash flows.

Software wasn’t just profitable, it was extrapolatable almost indefinitely. Investors could project usage, pricing, and margins far into the future with relatively little uncertainty. That visibility justified high multiples.

The key asset wasn’t the product itself.
It was confidence in the future.

AI disrupts that confidence.

Not because software stops working, but because it becomes harder to know what it will be worth.

What AI actually does to software economics

AI doesn’t eliminate software. It compresses it.

Features become easier to replicate.
Barriers to entry materially decrease.
Differentiation shifts from interfaces to outcomes.
Tasks spill across application boundaries.
Switching costs weaken.

As intelligence becomes more general-purpose, software increasingly looks interchangeable. Not useless, but less defensible.

This is the critical distinction:
AI doesn’t destroy software margins.
It destroys software scarcity.

And scarcity is what supported premium valuations.

When intelligence becomes abundant, constraints matter again

As software loses its uniqueness, value flows to what remains constrained.

Compute.
Chips.
Energy.
Data centres.
Capital intensity.
Supply chains.

These things cannot be spun up overnight. They require planning, financing, execution, and physical capacity. They are bounded by reality.

Abundance pushes value toward scarcity.

That is why hardware, infrastructure, and energy reassert themselves every time a new computing paradigm matures.

This isn’t new, it’s cyclical

We’ve seen this pattern before.

In the mainframe era, hardware dominated.
The PC era layered software on top, until it commoditised.
The internet multiplied software again, until margins compressed.
The cloud shifted value back to infrastructure, then abstracted it away.

AI accelerates the same dynamic.

Each wave expands the software surface area.
Each wave eventually erodes differentiation.
Each wave pulls value back toward the physical and capital-intensive layers underneath.

The cycle doesn’t end.
It rotates.

Why hardware outperformance is rational

Markets aren’t irrational for favouring hardware and infrastructure today.

These businesses offer:

• clearer demand visibility

• committed long-term customers

• tangible capacity constraints

• revenue tied to physical build-out

Software, by contrast, faces a harder question:

Not how much can it grow, but what is it actually defending?

That uncertainty compresses multiples before fundamentals deteriorate, not after.

Markets hate uncertainty, CFOs price it every day.

Why CFOs feel this shift before investors do

Long before valuation multiples adjust, this cycle shows up inside companies.

In capital-intensive businesses, CFOs are already seeing the same pattern:

• Software spend growing faster than output

• AI tools reducing headcount leverage but increasing infrastructure cost

• Marginal productivity shifting away from licences and toward compute, energy, and integration

In many large organisations, software now represents 15–25% of operating cost, yet delivers diminishing incremental efficiency once core workflows are digitised. By contrast, AI-driven compute, energy, and data infrastructure spending is rising at 30–50%+ annually in early deployments, not because it is discretionary, but because it is structurally required.

This creates a quiet inversion.

Software still looks asset-light on an income statement.
But the economic centre of gravity is moving back toward fixed assets, long-term contracts, and capital planning.

CFOs don’t experience this as a narrative shift.
They experience it as:

• higher capital intensity

• longer payback periods

• tighter sequencing decisions

• greater sensitivity to utilisation and scale

In other words, the return of constraint.

Markets eventually reprice this reality.
CFOs live with it first.

What investors are really struggling with

For years, software was treated as “own and forget.”

AI breaks that mental model.

Investors are no longer comfortable extrapolating margins, pricing power, or competitive position. When the future becomes harder to model, duration loses its premium.

This isn’t panic.
It’s repricing.

What still survives on the software side

None of this means software disappears.

But the winners change.

Software that:

• controls workflows rather than features

• is embedded in physical systems

• sits close to regulation, safety, or liability

• is paired with assets rather than abstractions

These businesses retain leverage because they are harder to displace.

Purely cosmetic or convenience-layer software does not.

Naming the phase

This is not the end of software.

It is the return of constraint.

AI doesn’t kill the software era.
It ends the illusion that intelligence alone deserves a premium.

When intelligence becomes cheap, value flows to what cannot be copied.

And that has always been hardware, infrastructure, energy, and execution.

Until next time,
Stay adaptive. Stay strategic.
And keep exploring the frontier of AI.

Fabio Lopes
XcessAI

💡Next week: I’m breaking down one of the most misunderstood AI shifts happening right now. Stay tuned. Subscribe above.

Read our previous episodes online!