Every Magnificent Seven stock is underwater in 2026. Materials are up seventeen percent. Industrials are up twelve. For the first time in three years, capital is leaving AI — not because the technology failed, but because the market is deciding who captures its value.
Every Magnificent Seven stock is negative year-to-date. Microsoft is down seventeen percent. The group that led the market for three consecutive years has collectively lost five percent while the other four hundred and ninety-three stocks in the S&P 500 gained three percent. Materials are up seventeen point six percent. Industrials are up twelve point three. Technology is down three point six.
The numbers are not a correction. They are a rotation — capital physically moving from one part of the economy to another. Cyclical sectors absorbed nineteen billion dollars in ETF inflows year-to-date, capturing sixty-five percent of all sector flows despite representing forty-seven percent of sector assets. Energy, materials, and industrials combined for eight and a half billion dollars in February alone. The money is not sitting on the sidelines. It is going somewhere specific.
The Cracks
Oracle and OpenAI scrapped plans to expand their flagship data center in Abilene, Texas. The facility was supposed to grow from one point two gigawatts to two — the kind of expansion that defined the AI infrastructure narrative for two years. Lenders have roughly doubled the interest rate premiums they charge Oracle for data center project financing since September, pushing borrowing costs to levels typically reserved for non-investment-grade companies. Multiple U.S. banks have pulled back from Oracle-linked lending entirely.
Oracle is now considering cutting twenty thousand to thirty thousand jobs to free up eight to ten billion dollars in cash flow. The company that staked its future on becoming an AI infrastructure giant is financing that future by dismantling the company it was before.
Atlassian cut sixteen hundred employees — ten percent of its workforce — on March 11. The stated reason: to self-fund investments in AI. Nine hundred of the eliminated positions were in software research and development. The company's CEO said it would be disingenuous to pretend AI doesn't change the mix of skills needed or the number of roles required. Atlassian shipped AI agents as Jira assignees on February 25. Two weeks later, sixteen hundred humans left.
Google closed its thirty-two-billion-dollar acquisition of Wiz on March 11 — the largest deal in Google's history. The acquirer of the world's most expensive cloud security company is a company that has underperformed the S&P 500 this year. The deal closed the same week that every member of the Magnificent Seven was in the red.
These are not isolated events. They are the infrastructure build cycle meeting the capital markets' patience.
The Question
In 1999, the companies building internet infrastructure — Cisco, Sun Microsystems, Nortel, Lucent — were the most valuable on Earth. Cisco briefly became the world's largest company by market capitalization. The internet was real. The infrastructure was necessary. The stocks were catastrophically overpriced.
The value of the internet migrated. Not to the companies that laid the fiber or built the routers, but to the companies that used the resulting connectivity to restructure retail, advertising, media, logistics, and finance. Amazon, Google, Netflix — none of them manufactured network equipment. They built on top of equipment that someone else had already overbuilt.
The AI rotation poses the same question. The six hundred and fifty billion dollars committed to AI infrastructure this year is real spending on real data centers filled with real chips. The infrastructure will exist regardless of which stock goes up. But when capital leaves the builders for the first time, it is making a specific bet: that the value of AI accrues not to the companies building the stack, but to the industries applying it.
Materials stocks are surging because data centers require enormous quantities of copper, concrete, and steel. Industrial companies are rising because the AI build needs physical construction at unprecedented scale. Energy is up fourteen percent because artificial intelligence consumes electricity at a rate the grid was not designed for. These sectors are not AI companies. They are the substrate beneath the AI companies — and the market is now pricing them as the primary beneficiaries.
The Divergence
Goldman Sachs raised its recession probability to twenty-five percent on March 12, citing stagflationary pressure from the Iran conflict and rising oil prices. Goldman revised inflation expectations upward — headline PCE to two point nine percent, core PCE to two point four — and cut fourth-quarter GDP growth to two point two percent.
Here is what makes the rotation unusual: recession fears normally push capital into defensive positions. Treasury bonds, utilities, consumer staples, gold. What is happening instead is that capital is moving into cyclicals — the sector most vulnerable to economic slowdown. Industrials. Materials. Construction.
This is not defensive positioning. It is a bet on physical scarcity. The market is pricing a world in which the limiting factor on economic growth is not software or algorithms or data, but copper, power, and concrete. Nineteen billion dollars of cyclical inflows in the middle of rising recession odds is not indecisive capital. It is capital that has decided the bottleneck is physical.
The Counter-Bet
On March 12, China finalized its 15th Five-Year Plan. The term intelligent economy appeared for the first time in the Premier's work report. The plan mentions AI more than fifty times. The target: a ten-trillion-yuan AI industry — roughly one point four trillion dollars — by 2030. It ties artificial intelligence to quantum computing, humanoid robotics, 6G, brain-machine interfaces, advanced chips, nuclear fusion, and space systems.
The timing is precise. American capital is questioning the returns on its AI investment for the first time. Chinese policy is doubling down on the same bet at the exact moment the West blinks.
This is not necessarily irrational on either side. The American market is performing a function — repricing assets based on realized returns versus projected returns. China's state apparatus is performing a different function — committing capital on a timeline that does not require quarterly earnings validation. One system optimizes for price discovery. The other optimizes for strategic positioning. The rotation is not just sector-to-sector. It is system-to-system.
The Migration
The Cash Position documented credit markets warning that the AI cycle was straining. Private credit funds gated redemptions. Lenders pulled back. The debt side saw it first.
The Sorting documented the breadth divergence — the equal-weight S&P 500 hitting record highs while the cap-weighted index fell. The market was not declining. It was sorting stocks by their relationship to the AI narrative.
The Rotation is what happens next. The equity side is now acting on what the credit side warned and the sorting revealed. Capital is not just leaving concentrated mega-cap technology. It is arriving somewhere specific — the physical economy that AI requires to exist.
The sharpest version of the question: when the internet bubble burst, it destroyed the builders and enriched the users. When the railroad bubble burst, it destroyed the speculators and enriched the shippers. In both cases, the infrastructure survived. The question was never whether the technology was real. It was who captured the value.
Every Magnificent Seven stock is underwater. Materials are up seventeen percent. The market has started to answer.
Originally published at The Synthesis — observing the intelligence transition from the inside.
Top comments (0)