A news-triggered bloodbath: 'Breaking news: Reports indicate that nearly half of the data centers in the U.S., originally planned to start operations in 2026, are expected to face delays or cancellations.' Chips (NVDA/AMD/AVGO): Management has confirmed that combined orders for Blackwell + Rubin GPUs exceed $500B, with visibility extending through the end of 2026, forming a strong backlog protection. However, sentiment might take a hit, as AMD's Helios architecture deployment is closely tied to the data center construction timeline, posing relatively higher risks. Storage (MAmericaNDK): Micron's CEO has explicitly stated that HBM supply for the entirety of 2026 has been locked in through pricing and capacity agreements. Construction delays have virtually no material impact on MU—HBM operates on a "priority booking, queued delivery" logic. SNDK (NAND) is slightly weaker in follow-through and faces some pressure. Optical modules/interconnects (LITE/COHR/CRDO/FN): Most directly affected, as revenue for optical modules aligns with the "power-on and production" timeline of data centers. Since fiber optic transceivers are the "nervous system" of AI data centers, a slowdown in optical module orders is often seen as a leading indicator of GPU demand softening. CRDO, a high-beta pure-play copper interconnect company, has valuation priced for perfect execution, making it the most at risk under delay news. LITE, despite Nvidia's $2B strategic investment and multi-year procurement commitments, could face a 30–50% valuation reset if there’s any guidance miss or signs of demand softening.