For much of the past year, investing in artificial intelligence felt refreshingly simple. Buy the companies building AI infrastructure or the chipmakers supplying it, and both sides generally moved higher together. That trade is becoming more complicated. Alphabet (NASDAQ:GOOG | GOOG Price Prediction) and Meta Platforms (NASDAQ:META) continue raising their AI spending plans, pushing annual capital expenditures toward levels that would have seemed unimaginable just two years ago.
Yet as the price tag climbs into the hundreds of billions of dollars annually, investors are beginning to ask a more practical question: Who ultimately pays for all this spending, and when does it begin generating acceptable returns?
AI Spending Keeps Rising While Investor Confidence Slips
The AI investment race isn’t slowing down. Over the past several weeks, both Alphabet and Meta increased their 2026 capital expenditure outlooks during earnings discussions, adding to an already record-setting spending cycle. Combined with Microsoft (NASDAQ:MSFT) and Amazon (NASDAQ:AMZN), the four largest hyperscalers are now on pace to invest around $730 billion on AI infrastructure in 2026, with Wall Street already anticipating another increase in 2027.
| Company | 2026 AI Capex Outlook |
| Alphabet | $195 billion to $205 billion |
| Meta Platforms | $125 billion to $145 billion |
| Microsoft | $190 billion |
| Amazon | $200 billion |
These investments include data centers, networking equipment, AI accelerators, custom silicon, and the enormous power infrastructure needed to support them.
Granted, those investments continue fueling demand for companies such as Nvidia (NASDAQ:NVDA), Broadcom (NASDAQ:AVGO), Micron Technology (NASDAQ:MU), Taiwan Semiconductor Manufacturing (NYSE:TSM), and other semiconductor suppliers. But investors are becoming less convinced that the companies writing the checks will generate returns quickly enough to justify the spending.
The AI Trade Is No Longer Moving Together
According to research compiled by Kevin Gordon of Charles Schwab, the 30-day correlation between America’s largest capital expenditure spenders and the Philadelphia Semiconductor Index has collapsed to almost zero. That’s near the lowest reading in at least 4.5 years.
The change has been dramatic.
- April 2026 correlation: +0.78
- Average correlation since early 2022: +0.60
- Current correlation: Near 0.00
Since the beginning of June, semiconductor stocks have generally advanced while many hyperscaler stocks have struggled to hold recent highs.
That divergence tells investors something important. The market no longer views AI infrastructure builders and AI suppliers as the same investment. Chipmakers benefit immediately from rising orders. Hyperscalers, meanwhile, must eventually prove those investments translate into stronger revenue growth, expanding margins, and higher free cash flow.
Ironically, the companies creating AI demand now face tougher questions than the companies selling the hardware.
Investors Are Beginning to Focus on the Bill
Adding to those concerns, a recent Nikkei investigation reported that major hyperscalers collectively hold approximately $1.65 trillion in off-balance-sheet obligations tied largely to long-term infrastructure commitments used to support AI expansion.
While these commitments comply with accounting rules and are disclosed in regulatory filings, they highlight just how much future spending has already been committed outside traditional balance sheet debt.
Recent earnings reports and rising capital spending forecasts have coincided with pullbacks from recent highs among several hyperscaler stocks, suggesting investors are becoming more selective about companies whose profits remain years away.
That said, none of this means AI spending is ending. Quite the opposite. Demand for computing power continues accelerating, and the largest technology companies possess balance sheets capable of funding enormous investments. The debate has simply shifted from whether AI deserves investment to whether those investments will earn attractive returns.
Key Takeaway
In short, AI has entered a new phase. For nearly two years, investors rewarded almost every company connected to artificial intelligence. Today, Wall Street is drawing distinctions.
Semiconductor companies continue benefiting from immediate demand, while hyperscalers increasingly must demonstrate that hundreds of billions of dollars in annual AI spending can produce sustained profit growth. Recent data suggests investors are already making that distinction.
Regardless of how large AI ultimately becomes, the next winners won’t necessarily be the companies spending the most money. They’ll be the ones that prove they can convert those enormous investments into durable cash flow and shareholder returns. For smart investors, that’s likely to become the defining theme of the AI trade over the next several years.
Contact [email protected] for any questions or corrections.
Top Gaining Stocks
Top Losing Stocks