Semiconductor stocks did something not seen since the dotcom bubble burst. What the charts show

SMH has broken to a 26-year high versus QQQ, signaling continued semiconductor leadership and suggesting the rally remains in a strong technical uptrend. Nvidia is highlighted as trading around 23.7x forward earnings despite revenue expectations rising from roughly $20B-$60B historically to about $200B in coming years, which the author frames as inexpensive relative to growth. The piece argues this setup supports increasing semiconductor exposure, with AI-related demand and Blackwell/Rubin product cycles reinforcing the constructive outlook.

Analysis

The key signal is not simply that semis are strong; it is that leadership has broadened inside the most crowded growth benchmark. When SMH outperforms QQQ at a new extreme, index-level passive flows become more self-reinforcing because every incremental dollar into broad tech implicitly increases the weight of the highest-beta component. That creates a convexity effect: if the semi complex keeps leading, the Nasdaq can stay elevated even if the rest of software and internet land more choppily.

NVDA’s setup is more interesting than the headline valuation implies. A 23x forward multiple is only "cheap" if the market believes earnings durability survives a post-buildout digestion phase; the real risk is not immediate multiple compression, but estimate revision breadth. If hyperscaler capex keeps rotating from training to inference and custom silicon, NVDA can still grow while the stock lags; if that happens, the equity becomes a cash-flow machine with a lower beta profile rather than a melt-up story, which would cap upside versus semis as a group.

The market is underpricing second-order beneficiaries and losers from the AI infrastructure cycle. Equipment, packaging, and memory can re-rate further if the buildout remains supply constrained, but a continuation of this move will also pressure adjacent GPU/accelerator and cloud-capex substitutes, especially firms whose AI narrative is more aspirational than monetized. For GOOGL specifically, TPU progress matters less as a product story than as a budget-arbitration tool: if internal silicon improves, it can slow external GPU demand growth at the margin without killing AI spend overall.

The contrarian view is that this is still a leadership thrust inside a narrow market, not yet proof of a durable secular regime change. A break above the cited technical objective would likely require another leg of capex acceleration and no meaningful digestion in megacap tech earnings expectations; absent that, semis can overshoot on momentum and then consolidate for months. The highest-probability failure mode is not a bubble burst, but a phase shift from "up and right" to "up, then long sideways" once positioning gets fully crowded.

Key Decisions for Investors

• Add to NVDA on pullbacks, not breakouts: use a 2-4 week scaling window and size for a 10-15% drawdown; upside remains favorable if estimate revisions stay positive, but chase risk is high after a relative-strength breakout.
• Express the theme via SMH vs QQQ long/short for 1-3 months: long SMH, short QQQ on a beta-neutral basis. If semi leadership persists, this captures the convexity without relying on broad-market multiple expansion.
• Pair NVDA long against GOOGL short only if you want a relative-value hedge on AI silicon monetization over the next 2-3 quarters; the trade works best if TPU adoption is framed as a capex efficiency tool rather than a direct GPU displacement story.
• Favor the semiconductor supply chain over the hyperscalers in the next earnings cycle: consider overweighting names tied to packaging, test, or lithography exposure versus generic mega-cap tech, because the incremental dollar of AI capex still flows hardest into constrained bottlenecks.
• Use call spreads rather than outright longs for upside participation in NVDA: a 6-9 month structure reduces theta bleed if the stock enters the expected consolidation phase while preserving exposure to another leg higher.