The trillion-dollar question: Will artificial intelligence spending pay off?

The trillion-dollar question: Will artificial intelligence spending pay off?

In this video insight, I wonder if the trillions being poured into artificial intelligence (AI) infrastructure can ever be justified. By 2028, industry hardware and data center spending could exceed $3 trillion, but customers would still need to spend well over $3 trillion a year — about 10 percent of U.S. gross domestic product (GDP) — to make the math work. Unless adoption and profits increase much faster than history suggests, this AI rocket ship could run out of fuel.

Copy:

Hello everyone, welcome back. Today we’re talking about something enormous – literal. The artificial intelligence (AI) industry is spending trillions on data centers, chips and servers… but here’s the trillion-dollar question: can customers actually spend enough to justify the investment? Are these hyperscalers or just hyperactive?

If customers aren’t willing to spend money on all the coming AI tools… this whole AI rocket ship could run out of fuel… According to McKinsey, 60 percent of AI data center spending goes on chips and hardware. These things wear out – they depreciate in about 5.5 years.

Morgan Stanley says the industry is on track to spend $3 trillion by 2028 and much of that is being postponed – meaning the big bills will hit later. So let’s do the math; If these assets do not generate economic profits after that 5.5 year period, the investments in semiconductor chips and hardware alone would need to generate net cash flows of more than $500 billion by 2028 to cover the capital costs of the equipment investment alone. That is the conclusion of the American investment house St James.

And there’s more: it gets even crazier. If data center operators want a 20 percent free cash flow margin — the kind that justifies their sky-high stock prices — they need $2.5 trillion in annual revenue.

Now, their customers – You know, the companies and people who use AI agents, Large Language Models (LLMs) and the like – they also want 20 percent margins. So how many do she should spend on AI services? The estimate is $3.1 trillion.

That is 10 percent of the entire US gross domestic product (GDP).

Let’s put that in perspective: US military spending? 3.5 percent of GDP. Total federal government spending in 2025? $7.3 trillion – that’s 23 percent of GDP. Total spending on Microsoft windows worldwide is $95 billion. So AI services should be bigger than the Pentagon… almost half of the entire US federal budget or 32 times the total global spending on windows.

St James also points out that Netflix brings in $39 billion from its 300 million subscribers.

The potential of AI is real. It’s transformative. But this math shows something crucial: the hype depends on generating revenue FAST – and on a scale we’ve never seen. And that’s not even counting energy costs, the possibility that a company might buy old chips and not get the full value before they’re replaced, or that a company might commit to a data center that can’t get cheap energy or… or… or… you know what I mean. The dream and hype meet commercial reality.

What if adoption slows down? If the return on your investment (ROI) is not visible in the boardrooms? Then all these investments become a multi-trillion dollar write-off. So Ask yourself: are we building the future… or a bubble?