← All insights

Theme 3 - Conviction 8 out of 10

The biggest capex boom in history

Every AI query consumes power, chips and infrastructure. Rather than guess which AI model wins, we prefer the businesses that supply the build-out.

Jim Mills, Merit Financial Services  ·  The Merit-Morgans Partnership  ·  September 2026  ·  3 min read

Global AI-related capital expenditure in US dollars: 131 billion in 2023, 233 billion in 2024, estimated 414 billion in 2025, 688 billion in 2026, 1,026 billion in 2027, 1,486 billion in 2028, 2,086 billion in 2029 and 2,706 billion in 2030.
Global AI-related capital expenditure, US$ billion. 2025 onwards are estimates. Source: Citi Research, September 2026.

Something counterintuitive has happened with artificial intelligence. As the cost of running AI has fallen, demand for it has exploded. Tasks that were too expensive a year ago are now worth doing, so businesses are using far more of it. That demand has to be served by physical infrastructure, and it is being built at a pace the world has not seen before.

Citi estimates AI-related capital expenditure rising from US$131 billion in 2023 to US$2,706 billion by 2030. The largest data centre now under construction is expected to be several times the size of the largest operating today.

Where the money goes

Of every 100 dollars of AI data centre spending: semiconductors 50, power 20, networking 15, cooling 7.5, facilities 7.5.
Where each US$100 of AI data centre spending flows. Source: BNP Paribas Equity Research, September 2026.

Half of every dollar goes on semiconductors, split between the AI processors and the memory they draw on. The rest goes on power, networking, cooling and construction. That is why our preferred exposures are memory, electrical engineering and infrastructure construction: the picks and shovels.

How client portfolios capture it

  • Global value funds that own memory makers growing profits quickly while still trading on modest valuations.
  • Emerging markets, particularly Korea and Taiwan, where much of the AI supply chain is built.
  • Australian engineering and mining services businesses that build data centres, transmission and mines.

What we are watching

Positioning in the largest US technology companies is crowded. A slowdown in AI spending, weaker returns on that spending or regulation could cause sharp short-term moves. Spreading AI exposure across infrastructure, power and value businesses, rather than relying on a handful of mega-cap names, is how the portfolios are built for that.

How is your portfolio positioned?

Merit clients can talk through these themes at their next review. Call 1300 827 439 or email jim@meritfp.com.au.

More from the September update

This information is general advice only. It has been prepared without taking into account your objectives, financial situation or needs, so before acting on it you should consider whether it is appropriate for you and speak with your adviser. Past performance is not a reliable indicator of future performance. Forecasts are estimates by third parties and may not be achieved. Views are those of the Merit-Morgans Partnership Investment Committee and are current at the date of publication. Merit Financial Services Pty Ltd is a Corporate Authorised Representative (No. 416822) of Paragem Pty Ltd, AFSL 297276. Unit 41/280 New Line Road, Dural NSW 2158. Research, dealing and administration are provided by Morgans Financial Limited, ABN 49 010 669 726, AFSL 235410. Financial Services Guide.