Goldman Sachs says the AI boom is bigger than investors think

In a recent analysis, Goldman Sachs has raised eyebrows by projecting that the AI investment boom is poised to surpass current market expectations significantly. The firm notes that the anticipated slowdown in AI spending may not occur as soon as some investors believe, particularly among major tech companies known as hyperscalers. Goldman's analysts suggest that capital expenditures from these hyperscalers could soar to around $1.1 trillion by 2027, a substantial increase from the $920 billion forecasted by Wall Street. In a more optimistic scenario, this figure could escalate to $1.4 trillion. The bank emphasizes that the demand for AI computing power remains in its infancy, predicting a staggering 24-fold increase in token consumption by 2030, largely driven by the emergence of enterprise agents. This surge in token consumption is set to create a domino effect, necessitating greater investments in data centers, chips, networking gear, and power infrastructure. However, analysts caution that rising input costs could also lead to increased capital expenditures needed to meet token consumption demands. Goldman Sachs highlighted a critical concern: numerous companies have recently raised flags about the rising costs associated with AI tools, sparking questions regarding whether the productivity enhancements will justify the expenses of operating more advanced AI models. This tension reflects a broader dilemma in corporate America, where firms are aggressively investing in AI yet struggling to demonstrate tangible returns on their investments. One of the key indicators backing Goldman's optimistic perspective is the remarkable backlog reported by cloud service providers like Google Cloud and Amazon Web Services, which has surged from $358 billion to $832 billion in just six months. The analysts predict that the balance between AI supply and demand will not be achieved until at least the latter half of 2027, implying that elevated spending may persist longer than anticipated. Historical trends suggest that investors might be undervaluing the scale of the necessary investment, with AI-related expenditures projected to account for approximately 1.5% of GDP by 2026. This is notably lower than the peaks seen during previous investment booms in sectors like railroads and electrification, which reached 2% to 3% of GDP. Despite these promising projections, Goldman warns that financing is not the primary hurdle to future spending. Instead, physical constraints may pose significant challenges, including delays in data center projects and limitations in memory, power, and labor resources. In conclusion, Goldman Sachs believes that sustained capital expenditures will likely bolster earnings growth for companies involved in the AI infrastructure buildout, such as semiconductor and networking firms. However, the firm also cautions that the market for AI infrastructure stocks is becoming increasingly crowded, with rapidly rising valuations that could lead to greater volatility. Despite discussions of AI productivity gains by over half of companies during recent earnings calls, concrete evidence remains scarce, with only a small fraction able to quantify specific benefits to their bottom lines. This analysis comes at a time when tech stocks have faced pressure due to geopolitical uncertainties and fluctuating interest rate outlooks, leading to a notable sell-off in the Nasdaq 100 index.