Goldman Sachs has made a significant call that flips the standard AI investment thesis on its head. The bank is now advising clients to favor the giant cloud companies, known as hyperscalers, over the semiconductor firms that have been the stars of the AI boom so far.
The core of their argument is a paradox created by the massive spending on AI infrastructure. This spending, or Capex, directly boosts the revenues and stock prices of semiconductor companies—the "picks and shovels" of the AI gold rush. For the hyperscalers footing the bill, however, this huge outlay raises immediate questions about Return on Investment (ROI). This investor concern has suppressed their stock valuations, even as they pour billions into building the future of AI. It's a classic case of one group's expense being another's income, and the market has so far rewarded the income earners.
This dynamic has been building for months, driven by several key factors. First, the recent wave of earnings reports from Alphabet (Google), Meta, and Amazon all featured staggering increases in capex guidance for 2026 and beyond. While this spending is essential for building next-generation AI, it immediately worried investors about short-term profitability. This "spending shock" is precisely what has kept hyperscaler stock multiples compressed, creating the entry point Goldman now sees.
Second, real-world physical constraints are beginning to matter more than financial projections. Recent reports indicate that nearly half of all planned data center projects in the U.S. face potential delays due to power grid limitations and equipment shortages. This is a critical development because it directly links the revenue of semiconductor companies to the actual, physical construction pace of these data centers. A delay for a hyperscaler isn't just a postponed project; it's a direct and immediate revenue risk for a chipmaker waiting to ship its products.
Finally, market sentiment and current valuations have created the perfect setup for this relative-value trade. Investor surveys, like the one from Bank of America, show that "AI bubble" or overinvestment fears are a top concern. This deep-seated skepticism is already baked into the hyperscalers' stock prices. It means that any tangible proof of AI monetization—better ad targeting, popular AI agents, or profitable enterprise services—could lead to a rapid and significant re-evaluation of their worth. In stark contrast, many semiconductor stocks have already seen huge gains and are priced for near-perfection, leaving far less room for positive surprises.
In essence, Goldman's call is a bet that the narrative is about to flip. The market's focus will shift from who sells the equipment to who makes money from the AI services built on it. Once hyperscalers start showing tangible profits from their enormous investments, their currently suppressed valuations could unlock substantial upside, while the red-hot semiconductor sector may finally cool down.
- Hyperscalers: Massive-scale cloud computing providers that offer services like computing and storage. Key examples include Amazon Web Services (AWS), Google Cloud, and Microsoft Azure.
- Capex (Capital Expenditure): Funds used by a company to acquire, upgrade, and maintain physical assets such as property, buildings, technology, or equipment like data centers.
- ROI (Return on Investment): A performance measure used to evaluate the efficiency or profitability of an investment. It measures the amount of return on an investment relative to its cost.