One selloff across the AI supply chain
On July 28, 2026, AMD fell about 8.1%, data-center power and cooling supplier Vertiv about 6.3%, and electrical-equipment maker Eaton about 3.1%. AI-power beneficiaries Constellation Energy, Vistra and GE Vernova also declined roughly 3.7%, 5.4% and 5.3%. AP reported that Micron lost 8.9% and Applied Materials 7.8% in the same session.
The synchronized move across industries points more to a repricing of the entire AI-infrastructure trade than to one company-specific event. A single trading day is not evidence of long-term demand, and the drivers overlap.
Why prices fell
- Elevated expectations and profit taking: In parts of the AI hardware and power chain, share prices had moved faster than reported earnings. When valuations require repeated upside surprises, even modest uncertainty can trigger a large correction.
- Proof of return on AI capex: The question has shifted from whether hyperscalers will spend to when hundreds of billions in chips, facilities and power become cash flow and productivity. Cheaper models and efficiency can expand usage, but may also reduce equipment needed per unit of compute.
- Rates and financing costs: Data centers, generation and grids tie up capital for years. Higher long-term rates reduce the present value of future profits and increase required returns on project debt and power contracts.
- Competition and supply risk: Progress in Chinese semiconductor capabilities raises concern that scarcity premiums could fade. There is still insufficient evidence, however, that leading suppliers will lose their position abruptly.
What will decide the next phase
More useful than a short-term rebound is whether orders convert into revenue and cash. For chips, watch data-center revenue growth, accelerator shipments and gross margin. For power and cooling equipment, watch new orders, book-to-bill, backlog cancellations and margin. For electricity suppliers, examine counterparty quality, plant availability, transmission and interconnection approvals, and cost of capital.
Structural AI demand remains, but profits will not arrive at the same pace for every company. Near-term volatility is likely to remain sensitive to rates and earnings. Over the medium term, the market should increasingly separate companies that convert announced gigawatts and investment into operating capacity, revenue and free cash flow from those that do not.
This is market analysis based on public information, not a recommendation to buy or sell any security.