A strong increase in artificial intelligence spending by the largest American technological companies has propelled American actions from new records. However, Goldman Sachs analysts warn that a dramatic investment withdrawal could eliminate a large part of these earnings.
In a research note published this week, strategists led by Ryan Hammond underlined how capital expenses of hyperscalers such as Microsoft, Amazon, Alphabet, Meta and Oracle fueled the growth of income through the IA supply chain – Semiconductor giants like NVIDIA with cloud infrastructures and equipment suppliers.
Record expenses, register assessments
Goldman estimates that the capital expenses of hyperscaler have already reached $ 368 billion in 2025, exceeding previous forecasts. These expenses, mainly for data centers, GPU and Cloud capacity, have translated directly into real income for flea manufacturers and service providers, feeding the boom of AI actions.
Register For Tekedia Mini-MBA Edition 18 (Sept. 15 Annual made for access to Blurara.com.
Tekedia ai in Business Masterclass opens registration.
Join Tekedia Capital Syndicate and Co-Investment in large world startups.
Register For Tekedia Ai Lab: from technical design to deployment.
The five largest companies in the S&P 500 – NVIDIA, Microsoft, Apple, Alphabet and Amazon – are currently negotiated at a 28 -year price price, noble according to historical standards but still well below peaks of 40 in 2021 and 50 during the DOT -COM 2000 bubble. Goldman stresses that, unlike past speculative mania, today’s evaluations are supported by today tangible income.
The risk of slowing down expenses
Despite this, analysts warn that the market is vulnerable to a sudden change. If the hyperscaller investment returns to the levels of 2022, suppliers of IA equipment and services could lose 30% of the 1 dollars billion planned in S&P 500 sales growth in 2026. In such an “extreme scenario”, the Goldman model suggests that the multiple global of S&P 500 could fall from 15 to 20 percent.
“The sustainability of this rally is directly linked to the momentum of Capex,” wrote Hammond and his team, noting that the enthusiasm of investors depends strongly on the hyperscalers that maintain their pace of aggressive investment.
Echoes of past technological booms – but with a difference
Goldman does not stop to label the current rally a bubble. Unlike the dot-com era, the income generated by AI demand are immediate and measurable. NVIDIA, for example, has reserved record quarterly sales from data center chips, while cloud suppliers have declared to accelerate the adoption of customers of AI services.
The historical parallel, however, is instructive. In the early 2000s, telecommunications carriers paid hundreds of billions of dollars into network accumulations, only to suddenly reduce expenses when revenues failed to make projections – a collapse that left equipment suppliers with excess and free fall capacity. Similarly, the boom of the composition of the Cloud of the 2010s saw an arms race in the investment of the data center, although in this case, the demand finally caught up and reshaped the corporate software industry.
The wave of expenditure on today shares elements of the two: a massive initial infrastructure construction with uncertain long -term demand curves, but also short -term monetization than previous cycles.
Momentum always upwards – for now
Despite the concerns about sustainability, the rally continues. AI actions increased by 32% in 2024 and added 17% until now in 2025. Analysts expect the growth in spending to slow down the end of 2025 or 2026, but the main technological companies have always revised higher advice, which suggests that the inflection point could be more distant than fearing.
Currently, Wall Street remains divided. The bulls argue that AI is a transformative technology still in the first rounds, while skeptics warn that the markets tariff perfection. Goldman’s message to investors landed somewhere between the two: the increase is real, but the risks too.
