Will Meta Rent Out Computing Power and Shock the Market? The AI Arms Race Isn't Over, but Markets Are on Edge

As the second-quarter earnings season approaches, markets expect Google, Microsoft, Amazon, and Meta to report impressive AI investment results. According to market estimates compiled by Visible Alpha, the combined capital expenditures of these four companies for the quarter ending in June will reach $168 billion, a 74% year-on-year increase, setting another record high.

However, these massive investments are beginning to erode corporate cash flows and stock prices. So far this year, only Alphabet, Google’s parent company, has outperformed the S&P 500 index. The other tech giants have seen pressure on their free cash flow due to continued investment in AI infrastructure, prompting the market to reassess whether this 'spending race' is nearing a turning point.

From Internal Use to External Rental: Meta’s Cloud Expansion Sparks Market Speculation

What’s truly rattling markets isn’t the high capital spending itself, but recent signals that differ from past trends.

Reports suggest Meta is evaluating the use of its vast AI computing infrastructure to launch enterprise cloud computing services, renting out some of its processing power to external clients.

If true, this would mark Meta’s first attempt to generate commercial revenue from its large data centers, which were previously used solely for internal AI model training. While Amazon, Microsoft, and Google have dominated the enterprise cloud market for over a decade, this would represent a new business model for Meta.

Madison Rezaei, an analyst at Bernstein, pointed out that in terms of scale, Meta’s AI infrastructure is already on par with major cloud providers. The company currently has about 20GW of computing capacity, with approximately 14GW more expected in the coming years, placing its data center scale at a globally leading level.

Renting Computing Power May Not Signal Retreat—It Could Improve Investment Efficiency

The market is sensitive because renting out computing power often implies a company has excess capacity beyond its own needs.

Meta CEO Mark Zuckerberg was asked about this during the shareholder meeting in late May. He stated that the company still believes all its computing power has a purpose and has not yet launched cloud services. However, if future construction exceeds actual demand, “renting out computing power” could be a viable option.

This has led the market to speculate whether Meta is experiencing “overbuilding.” However, most Wall Street analysts believe this does not mean Meta is stepping back from the AI arms race.

Brent Thill, an analyst at Jefferies, believes Meta is not reducing AI investment but rather seeking to convert its large-scale infrastructure—built ahead of schedule—into a new revenue stream, improving asset utilization and avoiding idle, expensive data centers.

The Wall Street Journal original: Will the AI spending war eventually see someone back down?

Meta Remains the Most Aggressive AI Investor

In fact, among the four tech giants, Meta has been the most aggressive in AI investment in recent years.

Zuckerberg established the Meta Superintelligence Labs, aiming to pioneer superintelligence (Superintelligence), and continues to significantly expand data centers and GPU clusters.

Market estimates suggest Meta’s capital expenditure this year will exceed half of its annual revenue, with free cash flow potentially turning negative for the first time—reflecting the company’s choice to continue heavy investment rather than cut back.

Therefore, even as the market discusses renting out computing power, most analysts believe this is merely preserving future commercialization options, not a sign that AI investment has entered a contraction phase.

AI Supply Chain Heavily Reliant on Capex—Any Signal Can Trigger Sell-Off

Nonetheless, the market remains highly sensitive to any news that might suggest a slowdown in AI spending.

The AI infrastructure boom over the past two years has driven rapid growth across the entire supply chain, benefiting companies in GPUs, servers, networking equipment, memory, and storage systems. This has significantly elevated the importance of semiconductor stocks in the U.S. market.

According to S&P Global Market Intelligence, chip companies now account for about 18% of the S&P 500’s market capitalization, far higher than the 5% level five years ago, reflecting AI’s role as a key driver of the U.S. stock market.

Therefore, if the market begins to suspect that large tech companies are slowing their investment pace, the entire AI supply chain could face repricing.

The Real Test Is Coming—Markets Await Earnings to Reveal the Next Step in AI Investment

The second-quarter earnings reports to be released in the coming weeks will be a crucial indicator for assessing the AI investment cycle.

Based on previous plans announced by the four companies, Google, Amazon, Meta, and Microsoft are expected to spend a combined $710 billion on capital expenditures this year. Even if this exceeds $1 trillion by 2027, the growth rate will clearly slow compared to this year.

This means AI investment may continue to hit new highs, but growth rates will gradually decelerate. For investors, the key is not necessarily which tech giant will cut spending first, but that as investment amounts grow larger, even a slight slowdown in growth could prompt the market to reassess the profitability outlook for the AI supply chain.

The AI arms race may not end in the short term, but as the spending base grows, market focus will gradually shift from “how much is being invested” to “whether returns are being generated.” This engine that has driven the global tech stock rally is now being tested on efficiency and profitability.