Economy at a Glance
Figure 1: Key Indicators and Trends
Source: Nasdaq Economic Research, FactSet
AI Stocks Are Dominating Market Returns as a Result of Strong Earnings
Artificial intelligence (AI) has been a big theme in markets lately. Returns have been so strong that some have openly wondered if AI is experiencing a bubble. But data shows that equity returns (Figure 2, bars) over the last couple of years have mostly been driven by earnings (Figure 2, circles), not multiple expansion.
The Nasdaq-100® is the major equity index with the most AI exposure, as it’s home to the “Magnificent 7” stocks, as well as Broadcom and AMD, among others. Since the start of 2024, it’s up over 50 percent, with 35 percentage points of that gain coming from earnings as AI has driven demand for these companies’ chips, cloud computing and AI chatbots. Other indexes with less AI exposure, such as the S&P 500, Europe’s Stoxx 600 and the S&P 600 small caps, have seen progressively weaker earnings growth and, in turn, weaker returns.
So, despite concerns about an AI bubble, equity returns have mostly been driven by real earnings.
Figure 2: Index Price and 12-Month Forward Earnings (% Change from 2024)
Data through November 7, 2027. Asia Pacific is FactSet aggregate. Sources: FactSet, Nasdaq Economic Research
Capex for AI Hyperscalers Has Nearly Tripled in the Last Two Years
To meet demand for AI services, AI hyperscalers have boosted capital expenditures, especially on data centers and the chips and power needed to run them. Total capex for the top five AI hyperscalers has nearly tripled in the last two years, according to JPMorgan (Figure 3), with the bulk of the increase going to AI investments. Forward-looking earnings data suggests this capex will continue to increase for at least the next couple of years.
Importantly, the companies spending on AI already have real earnings. Data shows they are currently spending only about 60 percent of operating cash flow on these AI investments.
Figure 3: Capex from the Major AI Hyperscalers
Source: MUFG
PE Ratios Are Falling for the Biggest AI Companies, and Rising for the Rest of the S&P 500
That’s a key difference from the Dotcom bubble, when many companies didn’t ever realize a profit. The Price-to-Earnings (PE) ratio for the five largest stocks in the S&P 500 has recently been falling, as earnings have increased faster than prices (Figure 4). Although, the remaining 495 stocks in the S&P 500 have seen their PE ratios rise close to their record highs, pushing up the overall S&P 500 valuation.
Figure 4: The Largest Stocks Trade at 28x, Below 2021 and Tech Bubble Levels
Five largest stocks today: NVDA, MSFT, AAPL, GOOGL, and AMZN; Source: Goldman Sachs Global Investment Research
AI Capex Increasingly Contributing to U.S. Economic Growth, and Likely to Continue
All of that additional capex is also now translating to the real economy. By one estimate, AI investment accounted for 92 percent of U.S. GDP growth in the first half of 2025. Moreover, spending on data center construction has increased 400 percent since 2019. As a result, by mid-2025, data center construction rivaled office construction in the U.S. (Figure 5).
Figure 5: Data Center Construction Spending on Pace to Surpass Offices
Spending is Seasonally Adjusted Annual Rate (SAAR); Sources: FactSet, Nasdaq Economic Research
Of course, there’s more to AI investment than just building the data centers. Research from Citi shows that AI equipment investment has increased 0.9 percent as a share of GDP since 2023. In comparison, this same type of investment rose over 1.25 percent during the Internet Revolution, but over a six-year period, indicating that there’s potentially more upside to AI investment.
Whether AI Investment Is “Too Much” Depends on Its Productivity Impact
With AI spending reaching such significant levels, some companies are questioning whether it will create stranded assets and wasted investments. Although that’s possible as leaders in the AI space emerge, it’s worth considering that AI could be transformative infrastructure. AI has the potential to rival the scale of other historic investment cycles, like railroads which peaked at 6 percent of GDP, automobiles, electricity and the internet (Figure 6). Although these estimates vary, they show that AI is still comparatively small as a percentage of GDP.
Figure 6: Historic Investment Cycles
Sources: Paul Kedrosky, Nasdaq Economic Research
Both facts seem to indicate that the development of AI could continue for a while yet. The hope for AI investment is that it unleashes productivity growth. For example, from the late 1990s to the early 2000s, PCs and the internet contributed to productivity growth that averaged around 3 percent per year, which is roughly double the pace from 2007-2019 (Figure 7).
Figure 7: U.S. Non-Financial Corporates Productivity 5-Year Average (Q/Q)
Source: MUFG
AI Remains Underleveraged by Companies
As things stand today, AI appears to be having little impact on the jobs market. That may be because data shows that companies’ AI adoption remains low. Census data shows that just 10 percent of U.S. businesses were using AI to produce goods and services as of late September 2025. For larger companies with over 250 employees, it was just 13 percent of businesses. However, Goldman Sachs indicates 37 percent of clients are using AI for regular production, and that number is projected to reach 50 percent next year.
For now, the soft labor market is likely due to the Fed’s 2022-2024 rate hike cycle and some budget strains finally showing on low-income consumers.
5 Questions Leaders Should Ask Now
- If our customer base includes lower-income consumers who face headwinds from rising inflation and slowing wage growth, how do we accommodate them?
- With interest rates falling, what are our plans for debt refinancing and interest costs?
- What are our dependencies on chips, power and critical inputs, and what contingencies do we have for cost spikes, delays, or export and control disruptions?
- With only about 10-13 percent of firms currently using AI, what is our roadmap to embed AI in core workflows and what measurable productivity targets will we hold management to?
- With the expectation that tariffs will remain—even if SCOTUS rules against the current implementation—how are we affected and how are we changing our supply chain?
