Hyperscaler AI Spending Could Hit $1.1 Trillion by 2027

Big Tech's AI infrastructure spending may dwarf current Wall Street expectations, according to new analysis from Goldman Sachs that projects hyperscaler capital expenditures could reach $1.1 trillion in 2027—roughly $180 billion more than consensus estimates.

The investment bank's analysts argue that current forecasts of approximately $920 billion are too conservative, with a bullish scenario pushing spending as high as $1.4 trillion. The projection was detailed in a Wednesday research note first reported by Business Insider.

Token consumption drives the forecast

Goldman's core thesis rests on explosive growth in AI computing demand. The firm forecasts token consumption will increase 24-fold through 2030, primarily fueled by enterprise AI agents that require continuous processing power.

Each token consumed demands computing resources, which cascades into requirements for data centers, semiconductors, networking gear, and power infrastructure. Rising input costs further amplify the nominal dollar amounts needed to support this consumption, according to the analysts.

Demand signals already appear robust. Google Cloud and Amazon Web Services reported a combined backlog of $832 billion in the first quarter, more than doubling from $358 billion six months earlier. Goldman doesn't expect AI supply and demand to balance until at least the second half of 2027.

Why it matters

This spending trajectory has major implications for the AI infrastructure supply chain, but also highlights a growing tension in corporate America. While companies are investing aggressively in AI capabilities, quantifiable returns remain elusive. Goldman's analysis found that only 11% of companies discussing AI productivity in first-quarter earnings calls provided specific metrics, and just 2% quantified an earnings impact. The gap between investment and proven ROI could eventually test investor patience, even as physical bottlenecks—not financing—constrain the buildout.

Physical constraints, not capital

Goldman argues that money won't be the limiting factor for AI infrastructure expansion. Instead, physical bottlenecks pose the real challenge. Data center projects face delays, and memory, power, and labor have emerged as key constraints.

The bank notes that AI-related investment represented roughly 1.5% of GDP in 2026. Historical infrastructure booms—railroads, electrification, automobiles—peaked at 2% to 3% of GDP, suggesting room for further growth.

Valuation risks emerge

While Goldman expects continued earnings growth for AI infrastructure suppliers, the firm warns that parts of the sector are becoming crowded. Valuations for many stocks have expanded rapidly, with share prices outpacing earnings revisions and increasing volatility risk.

The analysis comes as technology stocks face pressure from geopolitical concerns and interest rate uncertainty. The Nasdaq 100 fell 2% on Wednesday and dropped 6% since Friday's sell-off began, though the index remains up 13% year-to-date.

These details were first reported by Business Insider.