When Does Big Tech Stop Being a Capital-Light Business?

For much of the past two decades, some of the world’s largest technology companies have enjoyed an extraordinarily attractive economic model.

Build the platform, software or network and then add users at relatively little incremental cost. Revenues could grow significantly faster than the physical capital required to support them. Margins expanded, cash accumulated and shareholders were rewarded.

It is one reason investors have historically been willing to attach premium valuations to technology companies.

Artificial intelligence could be changing that equation.

Not because AI is failing. Quite possibly because it is succeeding.

The price of staying at the frontier

The numbers involved in building AI infrastructure have become extraordinary.

Microsoft, Alphabet, Amazon and Meta are investing hundreds of billions of dollars in data centres, processors, networking, power and associated infrastructure. Morgan Stanley now estimates hyperscaler capital expenditure of around $800 billion in 2026 and $1.16 trillion in 2027.

This is no longer simply research and development spending. It is physical infrastructure on an enormous scale.

And that raises an investment question which is somewhat different from whether AI will succeed.

If technology companies become structurally more capital intensive, should investors value them in the same way they did when their businesses were predominantly capital light?

Recent results illustrate the dilemma remarkably well.

Microsoft’s latest quarter provided compelling evidence that AI demand is real. Azure revenues increased 43%, Microsoft Cloud revenues grew 27% and commercial remaining performance obligations reached $678 billion.

Yet across the industry, the investment required to generate that growth is consuming increasing amounts of cash. Reuters analysis of consensus forecasts suggests the five major US hyperscalers could collectively spend more on capital expenditure than they generate in free cash flow by 2027.

Both things can therefore be true.

AI can be transformational.

And investors can still demand greater evidence of an adequate return on the capital being invested in it.

The AI trade is becoming three different trades

This distinction may become increasingly important because we tend to describe “AI” as though it were one investment.

It isn’t.

The first phase principally rewarded the enablers: semiconductor manufacturers, memory, networking and the businesses supplying the infrastructure.

The next phase places greater scrutiny on the hyperscalers. Microsoft, Alphabet, Amazon and Meta now need to demonstrate that enormous expenditure creates sufficient additional revenues, profits and cash flows.

But there is potentially a third group.

The adopters.

Banks, healthcare companies, industrial businesses, insurers and retailers may be able to use AI to automate processes, improve productivity and increase margins without themselves having to finance the enormous infrastructure required to develop the technology.

That creates an intriguing possibility.

The next great beneficiaries of AI might not necessarily be the companies spending billions of dollars building it.

They could be the companies using somebody else’s billions to improve their own economics.

Why this has mattered for Qualis Growth

This evolving distinction has influenced how we have positioned the MGTS Qualis Growth Fund.

During 2026, we deliberately reduced our dependence on the narrow AI and technology trade.

Most significantly, we moved our Nasdaq 100 exposure into the broader S&P 500. We have also increased exposure to smaller companies and to businesses displaying greater emphasis on quality, cash generation and shareholder returns.

Importantly, this was not a decision to abandon AI.

Through the S&P 500 and our other US allocations, Qualis Growth retains exposure to Microsoft, Nvidia, Amazon, Alphabet, Meta and many of the companies at the centre of the AI revolution.

What has changed is what we require from the portfolio.

We no longer want its success to depend upon the assumption that the companies which won the first phase of AI must automatically win the next one.

A broader US allocation gives us exposure not only to those established technology leaders, but also to financials, industrials, healthcare companies and other businesses which could become beneficiaries of AI-driven productivity.

From expenditure to return

None of this means the AI investment cycle is approaching its end.

Nvidia’s most recent results certainly don’t suggest it. Data-centre revenue increased 92% year-on-year and demand for computing infrastructure remains exceptional.

But markets rarely wait for an investment theme to end before asking harder questions.

The question in the first stage of AI was largely:

Who will benefit from all this spending?

The next one may be:

Who will earn the best return from it?

That is a subtly different question, but potentially a much more important one for the next stage of this extraordinary investment cycle.

For us, it is also the reason why diversification away from the narrowest expression of the AI trade should not be confused with becoming bearish on AI.

We remain enthusiastic about its potential.

We are simply becoming more demanding about who ultimately gets paid.

This article is for information only and does not constitute investment advice or a personal recommendation. Capital is at risk, and the value of investments can fall as well as rise.