Apollo Chief Economist: AI Capex Growth Is Twice the Pace of the 2000s Housing Boom
Taylor Wilson
Apollo Global's chief economist Torsten Sløk says AI data-center construction is adding to U.S. GDP at twice the peak rate of the mid-2000s housing boom — but warns that anything rising this fast can reverse just as hard if demand disappoints.
How fast is AI construction actually growing?
Sløk wrote in Apollo's blog "The Daily Spark" that AI data-center buildout is outpacing both the late-1990s telecom infrastructure boom and the pre-2008 residential housing expansion.
The AI cycle's construction pace is roughly twice the fastest rate of the housing boom.
This means → measured by how fast its GDP share is climbing, AI capex is expanding at the fastest pace the U.S. has seen in three decades.
Absolute size vs. speed of growth — which matters more?
AI infrastructure spending as a share of GDP still sits below the housing market's 2006 peak of about 6.2% — in absolute terms, it is not yet enormous.
But Sløk's focus is not the size of the slice; it is how fast the slice is growing. The upward slope of AI's GDP share already exceeds both prior cycles.
In plain terms = the pie is still relatively small, but it is inflating faster than anything before it — and speed itself is the risk signal.
How bad could a reversal be?
Sløk cites the historical precedent: housing's GDP contribution fell from 6.2% in early 2006 to 3.0% by late 2008 — that reversal is what made the recession so severe.
By contrast, telecom infrastructure was smaller; its bust triggered only a relatively mild contraction.
This means → the real macro risk is not "how much is being built" but "a cycle expanding at roughly 0.85 percentage points per year can contract at a similar rate if demand falls short."
Why did this analysis catch the Big Short investor's attention?
Michael Burry — the contrarian investor who famously predicted the 2008 subprime crisis — shared Sløk's analysis on social media.
Burry highlighted the argument that disappointing demand could trigger a reversal. He has repeatedly warned about AI bubble risk in recent years.
This reflects a notable signal: the investors most attuned to systemic risk are already drawing parallels between the AI capex cycle and the 2008 housing cycle.
What does this mean for investors?
Sløk's framework carries a two-sided message: AI capex both drives growth and stores up potential drag for when a reversal hits.
In plain terms = the same force lifting tech stocks will work in reverse if demand is disproved — and the drag could arrive just as fast as the boom.
This is not a prediction that AI will crash. It is a warning that speed amplifies consequences in both directions — what rises fast can fall fast.
Content is for reference only, not financial advice.