The VanEck Semiconductor ETF has returned 56 percent this year. The iShares Expanded Tech-Software ETF has lost 17 percent. A 73-point gap between two sectors that spent two decades moving in lockstep. Their rolling correlation has fallen from a long-term average of 0.71 to 0.29. In 32 of the past 60 trading sessions, the two ETFs moved in opposite directions, the highest count since both were created in 2001.
The divergence has a simple explanation. AI infrastructure spending drives chip demand. AI capabilities threaten the software applications that enterprise customers currently pay for. Same technology, two industries, opposite effects. One hundred percent of S&P 500 software stocks now trade below their 200-day moving average. Eighty-nine percent of semiconductor stocks trade above theirs. The software index has fallen roughly 30 percent since November. Snowflake, HubSpot, Cloudflare, Intuit, Atlassian, Workday: all down substantially. Investors have begun offloading software loans in debt vehicles at a discount, the credit market confirming what the equity market already said.
## The Other End
Brett Friedman at OptionStrat called it a potential regime shift: "Many software applications can be replicated with AI." If true, this is not a correction. It is software's Kodak moment, an entire category being disrupted by the technology it expected to benefit from.
But the market has not priced in the other end of the supply chain.
The companies buying the semiconductors are software companies. Microsoft, Alphabet, Amazon, and Meta are among the largest chip customers in the world. Alphabet alone spent $44.9 billion on capital expenditure in Q2 2026, enough to drive its free cash flow negative for the first time since 2004. Collective hyperscaler capex for 2026 is on pace to exceed $700 billion.
## The Closed Loop
The paradox: when software stocks fall, software companies' cost of capital rises. When cost of capital rises, capex budgets face pressure. When capex budgets face pressure, semiconductor orders shrink. The 73-point gap that rewards chip stocks is the same spending pattern that will eventually constrain the budgets those chip stocks depend on.
There is a deeper circularity. The hyperscalers' cloud divisions derive most of their revenue from hosting the software applications that AI supposedly threatens. When Snowflake's stock falls 30 percent, the cloud revenue it generates for AWS and Azure does not fall by 30 percent immediately. But the threat the equity market is pricing into software is a threat to the cloud computing platforms that fund chip purchases. The divergence is the supply chain arguing with itself.
## The Direction
The wedge will close. The only question is which direction.
If software companies figure out how to generate more revenue with AI than they lose to it, the software index recovers and the gap narrows from below. If they do not, the chip orders eventually slow and the gap narrows from above.
Corning's stock tripled in 1999 selling fiber-optic cable to carriers building the internet. The carriers went bankrupt. Corning lost more than 90 percent of its market value by 2002. The cable was real. The revenue to justify laying it was not.
The 73-point gap is the market's way of saying it believes in the tool but not the thing the tool was built to make. History does not say the wedge is wrong. History says the wedge is early. The market is pricing the chip before the application has proven it can pay for it. That has always been a sequence, not a destination.