Goldman Sachs: AI Spending Will Hit $1.2 Trillion by 2027

Every few months, the AI industry produces a story that quietly rewrites an assumption everyone had stopped questioning.

For the past year, that assumption was simple: Big Tech is spending too much on AI infrastructure. Wall Street analysts warned of a bubble. Investors fretted about the absence of a clear payoff. The narrative was one of unsustainability -- a spending spree that would eventually correct itself.

Goldman Sachs just published numbers that flip that script on its head.

The investment bank projects that Amazon, Alphabet, Microsoft, Oracle, and Meta will spend a combined $1.2 trillion on AI infrastructure in 2027 alone. That is more than 50 percent above the roughly $800 billion projected for this year and well above Wall Street's consensus of $1.1 trillion, according to Bloomberg, citing Goldman strategist Ryan Hammond.

Relative to GDP, it would be the biggest investment cycle since railroad construction in the 19th century.

The real story is not the peak, it is the slope

The headline number is staggering, but the hidden detail is more revealing. The growth rate of AI infrastructure spending is already collapsing.

The pace drops from nearly 100 percent year-over-year growth in 2026 to 54 percent in 2027 and then to just 12 percent in 2028. The spending is not accelerating -- it is reaching a visible peak. The buildout is front-loaded, and the deceleration tells us something important about how hyperscalers themselves view the timeline for AI adoption.

Goldman previously flagged in June that consensus estimates were far too low. The latest projection confirms that the Wall Street narrative of "AI spending is unsustainable" was always looking in the wrong direction. The spending is real, it is massive, and the companies doing it have already committed to levels that dwarf what most analysts thought possible.

The debt question nobody wants to answer

The uncomfortable detail buried in Goldman's analysis is how this spending is being funded. Capital expenditure already exceeds what the five companies generate from ongoing operations. The gap is being closed with debt.

Goldman expects more debt financing as the buildout continues. That creates a structural vulnerability: if AI revenue does not materialize fast enough, the companies face rising interest costs on top of their already enormous capital commitments.

To recoup the $1.2 trillion in outlays, the companies would collectively need about $300 billion a year in AI-related revenue. Current earnings still fall short, though there are encouraging signs. Cloud revenue growth jumped from 25 percent in 2024 to 48 percent in the second quarter of 2026. Whether that trajectory is steep enough to cover the debt payments before rates rise further is the open question that will define the next phase of the AI buildout.

Bottlenecks in power, labor, and memory chips could slow things further. Goldman flagged all three as constraints that could push timelines out and costs up.

What this means for the AI industry

The scale of this spending has two implications that matter beyond Wall Street.

First, the AI infrastructure buildout is now the single largest economic commitment in the technology sector, measured by any historical standard. It surpasses the dot-com cable buildout, the fiber optic boom of the early 2000s, and the mobile infrastructure expansion that followed the iPhone. If Goldman's projection is even close to accurate, the next 18 months will see more capital deployed into AI compute than into any technology category in history.

Second, the concentration of spending among five companies means that the future of AI infrastructure is increasingly controlled by a handful of balance sheets. If one of those companies stumbles -- a downgrade, a strategic pivot, a regulatory intervention -- the knock-on effects on the supply chain for GPUs, data centers, and energy infrastructure would be immediate and severe.

The AI industry has been asking whether the technology lives up to the hype. The more urgent question, now that Goldman has put a number on the investment wave, is whether the financial architecture supporting that hype can hold together long enough for the revenue to arrive.

This is not a story about a bubble. It is a story about a bet so large that the institutions making it cannot afford to lose -- and are borrowing money to double down.

Sources

• The Decoder: Goldman Sachs expects Big Tech to spend $1.2 trillion on AI infrastructure by 2027
• Bloomberg (via The Decoder): Hyperscaler Capex projections by Ryan Hammond
• CNBC: Debt-hungry AI companies face increased risk as bond yields spike