AI Hardware vs Software Stocks: The Rotation Is On

The AI trade is splitting in two. When Alphabet raised its 2026 capital-spending forecast toward $200 billion after the close on Wednesday, July 22, 2026, the market’s reaction told the story of the moment: the companies that sell the picks and shovels of the AI buildout ripped higher, while the software companies that were supposed to ride AI to bigger margins were sold hard. It was a clean, visible rotation — and it is reshaping how investors are positioning for the back half of the year.

The move followed Alphabet’s Q2 report, in which Google Cloud revenue grew 82% year over year to $24.8 billion and management lifted its capex guidance by roughly $15 billion. Investors sold Alphabet itself on the spending, but they bought the beneficiaries of that spending with both hands.

Hardware surged, software slumped

The dispersion in a single session was stark. Among the AI infrastructure names:

• Super Micro Computer rose roughly 19.8%
• Dell Technologies gained about 9.4%

Both companies build the servers and systems that hyperscalers stuff into data centers. When a customer the size of Alphabet tells the market it will spend more, not less, those order books get more valuable overnight.

On the other side of the ledger, enterprise software fell:

• ServiceNow dropped about 6.5%
• Workday fell roughly 6.2%

The broad indices barely moved by comparison — the S&P 500 slipped 0.14% and the Nasdaq Composite lost 0.57% — which masked how violent the rotation was underneath the surface. This was not a risk-off day. It was a reallocation.

Why the money moved

Two theses are driving capital out of AI software and into AI hardware at the same time.

Capex is revenue for someone. Every dollar a hyperscaler commits to AI infrastructure lands as revenue for the companies that supply servers, memory, networking gear, and accelerators. With Alphabet, Microsoft, Amazon, and Meta collectively steering hundreds of billions toward data centers, the demand signal for hardware has rarely been clearer. The hyperscaler capex boom that spooked equity investors when it showed up as an expense line at the platforms is, for the supply chain, the biggest tailwind of the cycle.

Software’s AI premium is under scrutiny. The bear case on enterprise SaaS has two prongs. First, the sheer scale of AI capital spending raises the risk that it crowds out other IT budgets — if a customer is pouring money into compute, the discretionary software renewal gets harder. Second, and more structurally, agentic AI threatens the seat-based pricing model that companies like ServiceNow and Workday were built on. If AI agents can do the work of a licensed human user, the “price per seat” logic that underpins SaaS valuations starts to wobble. Investors are not waiting to find out; they are derating the multiples now.

A sharp reversal from mid-July

What makes the July 22 rotation notable is how quickly it flipped the prior week’s narrative. Just days earlier, on July 15, AI hardware stocks had been the losers in a broad selloff: Dell fell 14% in a single session, Hewlett Packard Enterprise dropped 8%, and Super Micro slid as well, dragged down by fears that hyperscaler roadmaps were slipping and that memory pricing had gotten ahead of itself. Analysts at one shop flagged memory costs as a specific risk; another argued the pullback was a gift.

Alphabet’s capex confirmation resolved that debate, at least for a day. The roadmaps were not slipping; the spending was accelerating. Hardware names that had been sold on demand fears were bought back on demand certainty. It is a reminder of how narrative-driven this tape has become — the same stocks can be a bubble one week and a bargain the next, depending on a single guidance number. Our look at the semiconductor selloff on a strong TSMC quarter captured the same whipsaw dynamic earlier in the month.

The valuation backdrop

None of this is happening from a standing start. The hardware names had already delivered extraordinary runs before the rotation — Dell entered the week up more than 200% year to date, and Intel is up 163% ahead of its own report today. Stocks that have doubled or tripled are, by definition, pricing in a lot of good news, which makes each new data point matter more and each reaction sharper.

That is the paradox of the current market: the businesses are performing, the spending is real, and the stocks still trade with the volatility of speculative bets. Wall Street’s more optimistic voices frame the AI infrastructure spend — which some estimate could exceed $1 trillion across the industry — as a multi-year demand pipeline that justifies the hardware run. The skeptics counter that no capex cycle in history has been linear, and that the software derating is an early warning that the returns on all this spending remain unproven.

What it means

The hardware-versus-software split is the most important structural story in the AI trade right now, and it reflects a market that has grown more discerning about where in the stack the value accrues.

Who wins: the arms dealers. Server makers, memory suppliers, networking vendors, and accelerator designers capture hyperscaler capex directly and immediately. As long as the spending guidance keeps rising, the demand story for hardware is intact — and the July 22 reaction shows investors will pay up for confirmation.

Who loses: any software company whose growth story depends on adding seats. The market is repricing the risk that agentic AI compresses seat counts and that AI capex crowds out software budgets. That does not mean SaaS is doomed — the winners will be those that reprice around outcomes and consumption rather than headcount — but the burden of proof has shifted, and multiples are contracting while the market waits for evidence.

What to watch next: the remaining Big Tech capex guides this earnings season, which will either extend or break the hardware demand narrative; Intel’s foundry results today as a read on leading-edge chip demand; and, most importantly, the first real signs of whether enterprises are actually cutting software spend to fund compute, or whether the SaaS derating is running ahead of the fundamentals. The rotation is a bet on a specific sequence — build first, monetize later. If the monetization shows up, the software names get their multiples back. If it does not, the hardware trade eventually runs out of buyers too.