The artificial-intelligence trade is starting to split into two distinct groups: the companies supplying the infrastructure needed to power AI, and the large technology companies racing to build AI into their products, platforms, and business models.
The first group includes the “picks and shovels” of the AI build-out: chips, data centers, power, cooling, networking, memory, and other critical components. The second group includes many of the best-known platform companies that are spending heavily to develop, deploy, and monetize AI capabilities.
That distinction matters because the economics of the AI boom may not be evenly distributed. Infrastructure companies may benefit more directly from the current capital-spending cycle as demand for computing power, storage, energy, and specialized hardware continues to rise. Meanwhile, the companies building AI applications still have to prove that their investments can translate into durable revenue growth, margin expansion, and shareholder returns.
Will AI infrastructure stocks outperform?
MarketWatch recently reported on analysis from Holt, UBS’s research unit:
“An ‘extraordinary’ shift is happening in the technology sector, with artificial-intelligence infrastructure stocks set to vastly outperform most of the ‘Magnificent Seven’ technology companies, according to UBS.
“Analysts at the Swiss investment bank’s research arm, Holt, led by John Talbott, observed that there’s been a notable gap between memory stocks, like Samsung, SK Hynix and Micron Technology and the AI hyperscalers, such as Amazon, Alphabet and Meta, in a note on Thursday.
“Forecasts for cash-flow return on investment (CFROI), which is used to assess the level of economic return compared with the cost of capital, have accelerated for AI infrastructure companies, while declining by 200 basis points in the past two years for the big tech companies as AI spending commitments have soared, they noted.
“For Talbott and his team, that group of infrastructure companies includes U.S. technology hardware—excluding Apple—and the semiconductor and semiconductor-equipment groups and global semiconductors, worth more than $1 trillion in market capitalization. …
“Three years ago, Holt’s framework ranked Apple, Microsoft, Alphabet, Meta and Amazon as the top five economic profit generators in the industry. For 2027, Nvidia, Samsung, SK Hynix, Micron and Alphabet are in the lead.”
The ‘Mag Seven’ lags semiconductor stocks
Charlie Bilello at Creative Planning added further dimension on the issue on July 1:
“The most dominant investing narrative in the first half of 2026 was as follows:
- “Buy the companies selling the shovels.
- “Sell the companies paying for all the shovels.
“And assume the spending never slows.”
FIGURE 1: SEMICONDUCTOR ETF VS. MAG SEVEN ETF (YTD RETURNS)
Source: Charlie Bilello at Creative Planning
“So enticing was this theme that we saw investors pour $25 billion into the memory ETF ($DRAM) in less than 3 months, the fastest ETF ever to hit that mark.
“And Micron Technology ($MU), one [of] the 3 largest holdings in that ETF (along with Samsung and SK Hynix), reported a 15x increase in quarterly profit to $28.2 billion. That nearly matched Apple’s quarterly profit ($29.6 billion) and is expected to surge past Apple in the second quarter.”
FIGURE 2: RELATIVE QUARTERLY NET INCOME FOR APPLE VS. MICRON
Source: Charlie Bilello at Creative Planning
Bilello then asks, “How long will this imbalance last?”—referring to pricing power shifting from companies like Apple to infrastructure companies. He also notes that this will be a key issue for the remainder of 2026, as hyperscalers “are issuing more debt, raising more equity, and burning free cash flow to fund the AI buildout.”
He adds, “But their share prices are starting to underperform as investors grow weary of the rising capital demands. If that continues, one would think a pullback in spending would be the result. And when that happens, the narrative changes instantly.”
Related Article: Beyond the headlines: The ‘Magnificent Seven,’ AI, and the technology product cycle
The estimated scale of AI-related capital expenditures
In May 2026, Goldman Sachs provided a lengthy, high-level look at the projected enormous growth of AI-related capital expenditures—and the assumptions and drivers behind them:
“The scale of these expenditures is enormous. Estimates of $4 trillion to $8 trillion of total capital investment over the next five years have featured prominently in recent market commentary. That capital is used to buy new chips, build new data centers, and construct new power, all in an effort [to] assemble sufficient computing infrastructure to meet the moment. Debates about whether this figure is ‘too high’ are usually framed around a demand-side question: Will AI adoption and monetization justify the spend?
“But there is an equally important supply-side unknown. The scale of required investment for the AI build-out is itself more uncertain than commonly assumed. Estimates rest on a number of assumptions that, if changed, can significantly increase or decrease the amount of capital required.”
While those questions may be unanswerable today, the sheer magnitude of the projected spending is impressive.
FIGURE 3: BASELINE AGGREGATE AI CAPEX ESTIMATES (BN)
Sources: Goldman Sachs Global Institute, Goldman Sachs Global Investment Research (see the original article for assumptions)
RECENT POSTS
