AI overtakes US housing market as hyperscalers fuel boom

In Q2 2026, spending on data centres and information processing hardware climbed to US$752 billion. This figure surpassed residential housing investment for the first time, which reached US$748 billion during the same period.

The milestone represents a shift in the US economy: technology has become a large enough entity to outperform housing as a focus for spenders.

Adam Shapiro, vice president at the Federal Reserve Bank of San Francisco, commented on the new findings from the Bureau of Economic Analysis (BEA). He posted on LinkedIn: “We’re seeing a pivotal shift in the US economy: investment is shifting away from residential investment and towards computers.”

“The AI investment boom is massive,” he added alongside a graph from the BEA demonstrating the investment figures.

However, Shapiro clarified that residential investment tends to be more sensitive to borrowing costs and has been “held down by high interest rates.” He contrasts it to the tech sector, where investment such as data centre development usually relies “less on borrowing and more on cash.”

This setup may change, and so may the results. In his post, Shapiro said that the trend has been reversing – tech investment has started to focus on borrowing as cash supplies are finite, and costs across the board are currently high.

“Cash reserves will eventually run out at the current pace of spending,” Shapiro added in a comment. “Which is why the hyperscalers are relying more and more on borrowing.”

The companies behind the investment drive are household names: Alphabet, Amazon, Microsoft, Meta, Oracle, and SpaceX. In its report, S&P Global estimated that the total capital expenditures from the six companies were around US$ 470 billion in 2025. It anticipates that this will grow to $870 billion in 2026 and will overtake $1.3 trillion in 2027.

However, the investment boom comes despite some companies reporting losses. OpenAI’s 2025 financials were revealed in June and showed that the firm suffered operating losses of around US$20.9 billion during that year.

As conversations around the “AI bubble” continue, South Korea’s NICE Credit Rating (NICE Investors Service) estimates that Amazon will need to raise another US$94.1 billion to meet its 2027-2028 fiscal targets. It expects Oracle to require US$55.7 billion to meet its targets.

Learning from the housing crisis

For both AI developers and detractors, the term “AI bubble” will feel reminiscent of the US’ 2000s “housing bubble”, which contributed to the 2008 financial crash.

In a nutshell, during this period, high mortgage repayments for borrowers led to high-risk loans which offered better returns than government subsidies. An unexpected number of borrowers could not pay these loans back, and so the bubble popped when foreclosures began, and securities became undervalued.

Forwarding over a decade, AI investment has surged, and residential housing in the US has struggled with stagnation since 2022 – since the Covid-19 pandemic rush.

Globally, countries are reporting high mortgage rates and record-high interest rates that are holding back economic mobility for many.

Homeowners with comparatively low mortgages are unwilling to sell, tightening the housing market further. Increasing borrowing costs are putting housing developers off building; projects may easily go vastly over budget or become financially unfeasible.

As Shapiro noted, hyperscalers are beginning to focus on borrowing as they also become impacted by limited cash flow due to high costs. There’s a lot of room here for unregulated and risky investments, as companies engage in the AI race without oversight.

Developers have called for brakes, but the rhetoric understandably tends to surround the human impact of unrestricted AI progress rather than how it could lead to its own bubble pop. The housing market’s negative path is one the infrastructure industry could follow if not careful.

As AI unprecedentedly overtakes housing as the largest investment in the US economy, it will take mindful growth to avoid the current – and historical – pitfalls experienced by the housing market.

RELATED STORIES

• Data centres at a crossroads: Can capital keep up with demand?
• The $1tn AI build-out: Power, capital and contractors under pressure