Capex: AI Boom evident

Written by Damien Klassen | February 27, 2026

I've been having a second look at the private capital expenditure (capex) forecasts for Australia released yesterday. Capex is a key driver of economic cycles in Australia and getting an accurate read is critical to understanding the Australian economy.

What I really want to know is how big the IT / datacentre boom is in Australia. In the US it looks to have added 1%+ of GDP, in Australia it looks more like 0.25%, maybe stretching to 0.3% next year. Better to have it than not, but not enough to move the needle.

The other sector to watch is utilities which continue to be a significant source of spending:

I need to preface all of this with the observation that for some sectors, forecasts are misleading. You need to be careful about which conclusions are safe to draw. Some sectors are good at forecasting capex, and some are terrible.

Overall, capex is gently trending up as a % of GDP:

All of that is coming from the non-mining part of the economy:

Another disclaimer that while many subcomponents of capex have a decent forecasting record, adding the good ones with the bad ones to create an aggregate number doesn't fix the underlying problems. Non-mining capex aggregates are chronically badly forecast:

Capex forecasts are a classic case study in forecasting. The Australian Bureau of Statistics asks companies (a) how much did you spend in the last quarter (b) how much are you going to spend next quarter/six months and (c) how much are you going to spend next year.

The answer to question (c) for some sectors is almost invariably seriously underestimated, which affects the overall numbers. The median first forecast for total capex is usually nearly 15% too low because of some of the subsectors. But, strip this out, and look at the subsectors and more interesting data emerges:

One key sector we would love to get more statistics on is the construction sector - keeping in mind this is not total construction activity, just capex by construction companies themselves, not their clients. Unfortunately, this sector has such poor forecasts it renders the data all but unusable:

My best estimate of what is really going on involves converting the estimates to z-scores (long story, but basically looking at how each forecast compares to prior corresponding forecasts and how wrong the forecasts have been) which suggest construction capex will be 1.6 standard deviations below normal. But it is hard to have too much faith in construction forecasts.

Net effect

In the US, data centre construction looks to be adding around 1% to US GDP growth. i.e. it has increased by 1% of GDP. In Australia, the increase is more like 0.25%.

It is not going to move the needle much for the Australian economy. But it is worth continuing to watch, in an area where there is not a lot of good news.

View full post