Goldman Sachs is challenging growing concerns that the artificial intelligence investment boom may be approaching its peak, arguing instead that spending on AI infrastructure could substantially exceed current Wall Street expectations.

According to a new research note, the firm expects major cloud providers and hyperscale technology companies to continue accelerating investments in data centers, computing infrastructure, networking equipment, and energy systems required to support next-generation AI applications.

The bank estimates hyperscaler capital expenditures could reach approximately $1.1 trillion in 2027, significantly above the current Wall Street consensus of roughly $920 billion. Under a more optimistic scenario, spending could climb as high as $1.4 trillion.

AI Demand Continues to Expand

At the center of Goldman’s thesis is the belief that AI adoption remains in its early stages.

The bank forecasts that AI token consumption could increase twenty-fourfold by 2030, fueled largely by the emergence of enterprise AI agents capable of automating increasingly complex business tasks.

As AI systems process larger volumes of data and handle more sophisticated workloads, demand for computing power is expected to grow substantially. This trend creates a powerful tailwind for companies supplying semiconductors, networking hardware, cooling systems, power infrastructure, and data center services.

Goldman also notes that rising input costs are increasing the amount of capital required to support growing AI workloads, further boosting overall infrastructure spending.

Cloud Backlogs Signal Strong Future Demand

One of the strongest indicators supporting Goldman’s outlook comes from cloud computing providers themselves.

Combined backlog figures reported by Google Cloud and Amazon Web Services reached approximately $832 billion during the first quarter, up dramatically from $358 billion just six months earlier.

These expanding backlogs suggest customer demand continues to outpace available infrastructure capacity, reinforcing the view that significant investment will be needed to satisfy future AI-related requirements.

Goldman does not expect supply and demand conditions within the AI infrastructure market to fully normalize until at least the second half of 2027.

Historical Comparisons Support Larger Buildout

The firm also argues that investors may be underestimating the potential scale of the current AI investment cycle.

AI-related spending represented approximately 1.5% of GDP during 2026. Historically, major technological revolutions such as railroads, electrification, and the automobile industry generated investment peaks equivalent to between 2% and 3% of GDP.

If AI follows a similar historical trajectory, current spending levels may still be far from their eventual peak.

Physical Constraints Could Become the Biggest Challenge

While funding availability remains strong, Goldman believes the primary limitations on future AI expansion may come from physical infrastructure bottlenecks rather than capital shortages.

Data center construction delays, shortages of advanced memory components, electricity supply constraints, and labor availability have all emerged as challenges for companies attempting to expand AI capacity.

These constraints could slow deployment timelines while simultaneously increasing overall investment requirements.

Productivity Gains Remain Difficult to Measure

Despite massive spending commitments, questions remain about the financial returns generated by AI investments.

Goldman found that while more than half of public companies discussed AI-driven productivity improvements during first-quarter earnings calls, only a small percentage provided measurable evidence of those benefits.

Just 11% quantified productivity gains, while only 2% reported a direct impact on earnings.

This gap highlights the ongoing debate among investors regarding how quickly AI spending will translate into meaningful financial results.

Implications for Investors

Goldman believes continued capital expenditure growth should benefit companies tied directly to AI infrastructure development, including semiconductor manufacturers, networking suppliers, cooling technology providers, and energy-related businesses.

However, the bank also cautions that many AI-related stocks have experienced substantial valuation expansion. In several areas of the market, share prices have risen faster than earnings forecasts, increasing the risk of volatility if investor expectations become too aggressive.

Recent market weakness, fueled by geopolitical uncertainty and concerns over future interest rates, demonstrates how quickly sentiment can shift even within high-growth sectors.

Outlook

Goldman Sachs’ analysis suggests that the AI infrastructure cycle may still be in its early innings despite already unprecedented levels of spending. Rising cloud backlogs, growing enterprise adoption, and increasing demand for AI computing power support the case for continued investment growth through at least 2027.

While questions remain regarding long-term returns on AI spending, the scale of infrastructure development currently underway indicates that artificial intelligence could continue reshaping technology investment trends for years to come.