Key Takeaways
- Hyperscalers are investing hundreds of billions in artificial intelligence infrastructure.
- Data centers require vast energy and water resources, raising execution and environmental risks.
- Valuations in AI-linked sectors are elevated, with price/sales ratios nearing tech bubble levels.
The race to build AI infrastructure has become one of the largest investment stories of our time, and investors may already be more exposed than they realize.
This is just one of the topics we cover in Morningstar’s 2026 Outlook.
Meta Platforms META is harnessing AI to deliver more relevant content and advertising across Facebook and Instagram, allowing it to charge advertisers higher prices. In the pharmaceutical industry, AI computing power is helping discover new molecules, now moving into clinical development.
Building the infrastructure to support this emerging technology is very expensive. Data centers are being constructed across the world to house millions of semiconductors—graphics processing units—that power AI computation. These installations are massive. Meta, for example, plans to build the world’s largest data center in Louisiana, covering a footprint comparable to most of lower and midtown Manhattan combined.
Tech Giants Leading AI Data Center Expansion
Meta is one of several “hyperscalers,” alongside Alphabet GOOGL, Microsoft MSFT, Amazon.com AMZN, and Oracle ORCL. These companies are building data centers on a massive scale, both for their own operations and to sell computing power to companies, governments, and AI chatbot providers.
Collectively, hyperscalers are spending hundreds of billions of dollars each year on capital expenditure, much of it dedicated to data center expansion.
To put this in perspective, their combined 2026 capex will be more than 4 times what the publicly traded US energy sector spends to drill exploration holes, extract oil and gas, deliver gasoline to its stations, and run large chemical plants. Amazon’s capex alone is greater than that of the entire US energy sector.
Of course, deploying hundreds of billions of dollars across hundreds of massive structures comes with challenges—especially execution risk. First, data centers are extremely power hungry—GPUs need lots of electricity to make their calculations—and the current energy grid isn’t ready for the surge in demand. Second, cooling is another issue. GPUs “run hot” and require large volumes of freshwater to keep equipment running. Some communities are already pushing back against planned data centers, concerned about the water supply.
There are also questions about the hyperscalers’ business models. Will individuals and companies pay enough to justify the buildout? Only 5% of ChatGPT users currently pay for the service. Many companies are exploring how AI can enhance their ability to generate revenue or reduce costs, but most are moving slowly, committing limited resources today. During the dot-com era, “clicks” and “eyeballs” drove valuations that weren’t backed by profits. Could we be witnessing something similar with chatbots? People may like chatbots when they’re free, but how much will they actually pay per month to use them?
Hidden Cost Pressures Threaten AI Stocks
Another risk is cost. Data centers may turn out to be more capital-intensive than the hyperscalers currently assume. GPUs and servers account for approximately 35% of capex, and hyperscalers are assuming a useful life of five to six years. If their useful life turns out to be shorter, more spending will be necessary, which may leave hyperscalers below their planned return on investment goals, potentially slowing the pace of new construction.
While AI is still early in real-world usage, US equity investors are already heavily exposed.
Morningstar Indexes, in consultation with Morningstar Equity Research, developed the Morningstar Global Next Generation Artificial Intelligence Index, which tracks companies most leveraged to AI. US stocks in this index make up more than 30% of the value of the Morningstar US Target Market Exposure Index. If you own mutual funds or exchange-traded funds linked to broad US equity indexes, you’re already invested in AI.
AI Stocks Are Trading Near Tech Bubble Peaks
We also see AI’s impact on sector valuations. Communication services (home of Alphabet and Meta) and information technology (home of Nvidia NVDA, Microsoft, and other AI plays) are trading at price/sales ratios near or above tech bubble peaks. Now, because each sector earns more profit from every dollar of sales today compared with the tech bubble, the price/earnings ratios are elevated but haven’t reached those tech bubble levels.
Morningstar Equity Research tracks the price/fair value of its Global Next Generation Artificial Intelligence Index, which includes US and global stocks tied to the AI theme. The index currently sits above fair value, having ranged from 74% to 114% of fair value since 2023.
Investing in the AI Age? Remain Balanced
AI is here—but it’s still early in real-world adoption. Most investors’ portfolios are already significantly exposed to AI-related stocks, whether that be hyperscalers, semiconductor companies, or related plays. Morningstar Equity Research views the group as fairly valued overall while reflecting different upside and downside scenarios, with many stocks in 3-star territory—the middle of our Morningstar Rating range. This is consistent with Morningstar’s Global Next Generation Artificial Intelligence Index, which sits near fair value.
Over the past 10 years, the stock market has grown increasingly concentrated in AI-related names. Ten years ago, Nvidia, Microsoft, Amazon.com, Meta, Broadcom, Alphabet, and Oracle were 9.7% of the Morningstar US Target Market Exposure Index. Today, their weight has almost tripled, with that group accounting for 28.7%.
This means even those investors with substantial exposure to a broad market index are heavily exposed to AI-driven returns. For those seeking to reduce concentration risk or lacking the appetite for such exposure, we recommend diversifying into US value and small-cap stocks, which currently trade at a discount to our fair value estimates and have far less AI exposure, or shifting portfolios to stocks in selected foreign equity markets.
