- The PHLX Semiconductor Index posted its best-ever first 100 trading days on record Wednesday, climbing 82% YTD as chipmakers added roughly $5.7 trillion in combined market cap — with Sandisk up 570%, Intel up 200%+, and SK Hynix, Micron, and Samsung all crossing $1 trillion in valuation for the first time.
- A global memory chip shortage is the engine: the hourly cost to rent an Nvidia B200 has nearly doubled in three months (from $2.66 to $5.27), and daily spot rates for DDR5 memory have more than doubled to $22.50 since November — giving chipmakers extraordinary pricing power over hyperscalers and AI infrastructure builders.
- The rise of agentic AI has resurrected demand for CPUs in addition to GPUs: Intel’s data center segment hit $5.1 billion in quarterly revenue on AI agent demand, beating estimates and driving a 20% single-day stock surge in April.
- Despite the run-up, semiconductor stocks still look cheap on earnings: the SOX trades at ~26x forward earnings vs. a 10-year average of 21x — and Micron specifically sits at just 10x forward earnings, a fraction of the S&P 500’s 22x multiple, because profits have scaled faster than share prices.
What Happened?
Wednesday marked a new milestone in the 2026 chip rally: the Philadelphia Semiconductor Index hit its best-ever performance over a year’s first 100 trading days, climbing 82%. The gains are broad but concentrated in the memory and AI infrastructure segments. Sandisk has surged 570%. Intel has more than tripled on the back of agentic AI CPU demand. SK Hynix and Micron both crossed $1 trillion in market cap this week — joining Samsung, which hit the threshold earlier this month — becoming the first memory chipmakers ever in the club. Nvidia reported fresh records for revenue and income last week. AMD now commands a higher market cap than JPMorgan Chase. Private AI infrastructure deals hit $80 billion in Q4 2025 before moderating slightly this year. The SOX is now trading far above both its 50-day and 200-day moving averages — a technical condition that historically signals either a breakout or an overextension.
Why It Matters?
This is not a single-stock AI hype story — it is a fundamental supply-demand rerating across the entire semiconductor value chain. HBM for AI accelerators, DDR5 for servers, CPUs for agents, and specialized chips for inference are all simultaneously in short supply. The pricing data is concrete: rental costs for compute have nearly doubled in three months, and memory spot rates have more than doubled since November. For investors, the key distinction from the dot-com era is that the earnings are real: Micron at 10x forward earnings and SK Hynix at 7x are not speculative multiples — they are a valuation gap that has only recently begun to close as the market recognizes these companies as AI infrastructure rather than commodity cyclicals. The leveraged ETF launches in Korea this week add a speculative froth layer, but the underlying thesis is earnings-driven.
What’s Next?
Three risks can derail the rally: hyperscaler capex deceleration, the emergence of AI models requiring significantly less compute, or negative policy changes (export controls, tariffs on semiconductor equipment). HB Wealth’s chief market strategist flagged all three in a note Wednesday. The first risk is the most watch-worthy — any guidance reduction from Microsoft, Google, or Amazon on data center spending would hit HBM demand assumptions immediately. On the upside, SK Hynix’s pending US ADR listing could pull a new wave of institutional capital into the trade. The SOX trading at 26x forward vs. a 10-year average of 21x means the market is pricing in sustained above-trend earnings growth — reasonable if the AI buildout continues, but fragile if any of the three risk scenarios materialize.
Source: The Wall Street Journal
