Back to News
Market Impact: 0.6

SanDisk Up 558% This Year as Chip Stocks Outpace the 1999 Dot-Com Bubble

SNDKINTCMU
Artificial IntelligenceTechnology & InnovationCorporate EarningsCorporate Guidance & OutlookCompany FundamentalsMarket Technicals & FlowsInvestor Sentiment & PositioningDerivatives & Volatility

Semiconductor stocks have staged an extreme rally, with SanDisk up 558% YTD, Intel up 239%, and the S&P 500 semis adding about $3.8 trillion in market cap over six weeks. The article argues the move is partly justified by fundamentals, citing Micron's revenue rising from $15.5B in 2023 to an expected $107B in 2026 and SanDisk Q3 FY26 revenue of $5.95B, up 251% YoY. It also flags froth and risk, noting SOXL is up 320.99% YTD and leverage/reset mechanics can amplify volatility and decay.

Analysis

This is less a broad semiconductor trade than a narrow squeeze in the memory and foundry complex, where earnings revisions are still outrunning valuation discipline. The key second-order effect is that AI capex is increasingly pulling demand through the entire memory stack, not just GPUs: as inference workloads become persistent, buyers need more DRAM/NAND for throughput, caching, and model persistence, which supports MU and SNDK even if accelerator spend eventually normalizes. The market’s real vulnerability is positioning, not fundamentals. When leveraged retail proxies like SOXL become a crowded vehicle, the marginal buyer is no longer fundamental capital but reflexive momentum capital, which means any pause in upside revisions can trigger forced de-risking disproportionately fast. That sets up a discontinuous drawdown risk over days to weeks, even if the underlying business cycle remains strong for months. INTC is a different animal: the stock can keep levitating on strategic optionality, but the path is much more dependent on narrative and policy than on near-term operating leverage. That makes it a poorer vehicle for expressing the AI memory trade and a better one for a relative-value hedge against the more expensive beneficiaries. The consensus is probably underestimating how quickly memory margins can mean-revert once supply responds; the highest-beta winners may be the most vulnerable if lead times normalize by mid-2026. The contrarian view is not that AI demand is fake, but that investors may be paying peak multiple for mid-cycle cash flows. In prior hardware cycles, the biggest losses were not from demand collapse but from supply arriving just late enough to crush pricing after the best growth numbers were already printed. That creates a strong asymmetry: fundamentals can stay good while equities still underperform if the market starts discounting normalized rather than peak earnings.