Apollo's Torsten Slok Says AI Capex Is Growing Nearly Twice As Fast As The Housing Boom — Warns It Can 'Unwind At A Similar Pace'

In a note on Thursday, Slok said consensus forecasts imply hyperscaler capital expenditures will reach about 3% of U.S. Gross Domestic Product annually between 2027 and 2029, up from 0.3% in 2019 and 1.4% in 2025.

• Slok said that the AI buildout is accelerating at roughly 0.85 percentage points of GDP a year, versus 0.5 points during the housing boom and 0.15 points during the telecom cycle.
• He noted hyperscaler capex is set to exceed the late-1990s telecom peak as a share of GDP but remain well below the 6.6% peak of the housing boom, while warning that its rapid pace also raises the downside risk.
• Slok said that the telecom bust contributed to the mildest post-war recession, while the housing downturn made the Global Financial Crisis far more severe.

The artificial intelligence (AI) infrastructure spending boom is building at a pace unmatched by previous investment cycles, according to Apollo Chief Economist Torsten Slok, who warned that the rapid expansion could become a macroeconomic risk if AI demand eventually disappoints.

In a note on Thursday, Slok said consensus forecasts imply hyperscaler capital expenditures will reach about 3% of U.S. Gross Domestic Product (GDP) annually between 2027 and 2029, up from 0.3% in 2019 and 1.4% in 2025.

He argued that while the buildout remains smaller than the housing investment boom in absolute size, it is building at a much faster pace and could unwind just as quickly.

AI Capex Boom Outpaces Previous Investment Cycles

Slok said hyperscaler capital expenditures are expected to rise from 0.6% of GDP in 2023 to 3.1% in 2027, representing a 2.5 percentage-point increase.

By comparison, the late-1990s telecom buildout increased by 0.4 percentage points, while residential investment climbed 2.2 percentage points during the housing expansion that preceded the global financial crisis.

He added that the AI buildout is accelerating at roughly 0.85 percentage points of GDP per year, compared with about 0.5 percentage points during the fastest phase of the housing boom and roughly 0.15 percentage points during the telecom investment cycle.

“The AI cycle is building at close to twice the pace of the housing boom at its fastest,” Slok wrote.

Slok Warns That Reversal Is The Risk, Not The Buildout

While hyperscaler capital spending is projected to exceed the peak of the late-1990s telecom and fiber buildout as a share of GDP, Slok noted it remains well below the 6.6% peak reached by residential investment in 2005.

However, he argued that the speed of the expansion also increases the potential downside.

“A cycle that builds at 0.85 percentage points a year can unwind at a similar pace, and that, rather than the buildout itself, is the macro risk if AI demand disappoints,” he said.

Past Investment Booms Offer A Warning

Slok said previous investment cycles offer a useful benchmark for assessing today's AI infrastructure buildout. He noted that the collapse of the telecom and fiber spending boom in the early 2000s contributed to what he described as the mildest post-war recession.

By contrast, the reversal in residential investment proved far more damaging. Residential investment fell from 6.2% of GDP in early 2006 to 3% by the end of 2008, a decline Slok said helped make the Global Financial Crisis recession significantly more severe.

According to a Bloomberg report, the four big hyperscalers in the U.S., Alphabet Inc. (GOOG, GOOGL), Microsoft Corp. (MSFT), Amazon.com Inc. (AMZN), and Meta Platforms Inc. (META) have committed nearly $2.4 trillion in spending over the next few years, pointing to a continued surge in the AI infrastructure boom.

The S&P 500 ETF (SPY) is up 21% over the past 12 months, while the Invesco QQQ Trust (QQQ) is up 26%.

The Global X Artificial Intelligence & Technology ETF (AIQ) is up 39% during this period, while the iShares U.S. Technology ETF (IYW) is up 38%.

For updates and corrections, email newsroom[at]stocktwits[dot]com.<