The Magnificent Seven are having a rough year. As of April 12, 2026, every major AI hyperscaler has lost ground: Microsoft is down approximately 23%, Meta is down roughly 12.9%, Amazon has shed about 7.5%, Apple is off around 6.9%, Nvidia has slipped approximately 5%, and Alphabet is down about 2.5%. These are snapshot figures from investor journalism, the precise numbers shift daily, but the directional story is consistent across multiple sources.
The infrastructure layer tells a different story. Vertiv, which makes thermal management and power systems for data centers, is up approximately 61.8% year-to-date. Micron Technologies is up roughly 32.3% over the same period. Those aren’t modest beats. That’s a spread of roughly 84 percentage points between the best and worst performers in the broader AI equity universe, as of the same date.
Why the divergence? The hyperscalers are absorbing the cost side of an unprecedented AI capital expenditure cycle. Microsoft, Amazon, Google, and Meta have committed hundreds of billions to build AI infrastructure, data centers, training clusters, networking fabric. That spend hits earnings before the revenue materializes. Investors appear to be repricing that risk. The companies actually supplying the physical layer, cooling, power conversion, memory, collect their margins regardless of which hyperscaler wins the model race. The YTD performance gap suggests the market is currently pricing infrastructure execution over frontier AI development, a rotation visible across multiple infrastructure-adjacent names.
Vertiv’s underlying business supports the valuation move. The company reported organic order growth of 252% year-over-year in Q4 2025, with diluted earnings per share up 200% over the same period, according to Motley Fool’s analysis. That’s not a stock story built on AI enthusiasm. Those are hard operating metrics from a company whose customers are building data centers at scale.
This pattern extends beyond the two infrastructure names in today’s data. Recent pipeline coverage has tracked CoreWeave’s infrastructure partnership with Anthropic and Oracle’s enterprise AI platform expansion, both examples of capital flowing toward the infrastructure and deployment layer rather than the frontier model layer. The Alaska data center brief captured federal appetite for physical AI infrastructure. Taken together, these aren’t isolated data points. They’re a consistent signal across multiple cycles.
What to watch: The hyperscaler earnings cycle will be the next stress test. If Q1 2026 results show AI-driven revenue acceleration, particularly in cloud, enterprise software, and advertising, some of this discount may reverse. If capex continues to expand without corresponding revenue lift, the infrastructure-over-hyperscaler trade could extend further. Watch for capital expenditure guidance revisions and any language from CFOs about AI-specific revenue attribution.
The takeaway for investors and enterprise strategists isn’t that AI is failing. It’s that the equity market is in the middle of a pricing argument about where value accrues in an AI buildout. Right now, the answer it’s reaching is: not at the top of the stack.