Semiconductors vs Nasdaq 100

SMH (VanEck Semiconductor ETF) tracks the MVIS US Listed Semiconductor 25 Index, holding 25 of the largest US-listed semiconductor companies. Top holdings as of early 2026: Nvidia (~20 percent), TSMC (~13 percent), Broadcom (~8 percent), AMD (~5 percent), Qualcomm, ASML, Applied Materials, Intel, Micron. The fund launched December 20, 2011 and carries a 0.35 percent expense ratio.

QQQ (Invesco Nasdaq-100 ETF) tracks the Nasdaq-100 Index, holding 100 of the largest non-financial companies listed on the Nasdaq exchange. Top holdings: Nvidia (~9 percent), Apple (~7.6 percent), Microsoft (~5.7 percent), Amazon (~5.5 percent), Alphabet (~6.7 percent combined), Broadcom (~4.5 percent), Meta (~3.7 percent), Tesla (~3.5 percent). QQQ launched March 10, 1999 and carries a 0.18 percent expense ratio (reduced from 0.20 percent in December 2025).

SMH and QQQ have substantial overlap in AI-exposed names. Nvidia, Broadcom, and (through QQQ's broader tech names) Microsoft, Alphabet, Meta, Amazon all benefit from the AI capex cycle. The practical difference is that SMH concentrates on the chip manufacturers themselves (Nvidia, AMD, TSMC, ASML, Applied Materials) while QQQ includes the broader software, platform, and consumer-tech ecosystem (Apple, Microsoft, Tesla, Netflix, etc.).

Both ETFs have been AI-cycle beneficiaries since November 2022. SMH has risen approximately 3x from the ChatGPT release to April 2026; QQQ has risen approximately 2.1x over the same window. The SMH-to-QQQ ratio has therefore risen approximately 40 percent over three years, reflecting the semiconductor sector capturing disproportionate AI capex relative to the broader tech complex.

SMH outperforms QQQ when the AI story is in its capex-buildout phase: hyperscalers committing massive spending to GPU farms, data center construction accelerating, memory and networking demand surging. This has been the dominant pattern from late 2022 through 2024, with Nvidia's quarterly data center revenue growing from $3 billion annually to $194 billion (fiscal 2026 total). The $500 billion+ combined annual capex commitment from Microsoft, Google, Meta, Amazon, and Oracle has flowed directly into SMH constituents.

The mechanism: AI workloads require specific chip types (GPUs for training, custom ASICs for inference, HBM memory, switching silicon). Companies designing and manufacturing these chips (SMH holdings) capture the direct capex spending. Broader tech names (most QQQ holdings outside semis) benefit second-order through productivity gains or platform monetization, which takes years to show up in revenue. SMH therefore leads QQQ in the early capex-deployment phase.

QQQ outperforms SMH when the AI story transitions from capex deployment to monetization: software companies capturing value from AI-enhanced products, consumer tech companies benefiting from AI features, and the broader platform economy growing through AI productivity. This pattern historically emerges 2-4 years into a major tech capex cycle, when the infrastructure is mature and the revenue benefits accrue to end-market companies.

Through early 2026, QQQ is beginning to show signs of this transition. Microsoft Azure AI revenue, Google Gemini-related cloud revenue, and Meta AI-driven ad targeting all showed material contributions in 2025. Apple's Apple Intelligence rollout (delayed through 2024, broader deployment in 2025-2026) adds to this. If AI monetization accelerates in 2026-2027, QQQ should begin catching up to or exceeding SMH performance, which would reverse the three-year SMH outperformance pattern.

Nvidia is the single most important name in both ETFs. SMH weights Nvidia at approximately 20 percent; QQQ weights Nvidia at approximately 9 percent. Combined, the two ETFs hold roughly 29 percent of their weight in Nvidia across the two products. Nvidia earnings therefore drive both, though SMH is more exposed.

This means the SMH-QQQ spread is partially an Nvidia-beta bet. When Nvidia outperforms, SMH outperforms (20 percent weight vs 9 percent). When Nvidia faces specific challenges (competitive pressure from AMD/Broadcom, regulatory headwinds, China restrictions), SMH underperforms more than QQQ. Nvidia's August 2024 earnings miss, for example, hurt SMH more than QQQ on that day.

Semiconductor demand cycles typically run 2-4 years in length, driven by end-market application waves: PC cycles (2000-2008 era), mobile cycles (2007-2015), cloud infrastructure (2015-2020), and now AI (2022-present). In each cycle SMH has outperformed during the buildout phase and underperformed during the inventory correction phase.

Major post-1999 SMH drawdowns: 2000-2002 dot-com (-85 percent), 2008-2009 financial crisis (-60 percent), 2014-2015 China slowdown (-25 percent), 2018 trade war (-30 percent), 2022 inflation shock (-38 percent). The 2022 drawdown was particularly severe for SMH given its duration sensitivity; by contrast QQQ fell 33 percent in 2022. SMH tends to have 15-25 percent higher peak-to-trough drawdowns than QQQ, reflecting the semiconductor sector's greater cyclicality.

Three years into the AI capex cycle, markets are asking whether the buildout phase is sustainable. Signs of maturation: Nvidia's year-over-year revenue growth has decelerated from 265 percent in Q4 FY24 to 62 percent in Q3 FY26 (still extraordinary but clearly slowing). AMD and Broadcom have captured incremental market share in AI ASICs. Hyperscalers have begun commissioning their own custom silicon (Google TPU, Amazon Trainium, Microsoft Maia) to reduce Nvidia dependence.

Signs of continued buildout: total data center capex is still growing (Nvidia management projects 40 percent annual growth through 2030), enterprise adoption of AI is accelerating beyond hyperscalers, and China's own AI infrastructure investment is creating parallel demand for alternative chip suppliers. The SMH-QQQ ratio has been approximately flat through 2025-2026, suggesting markets see the cycle as mature rather than ending. A sustained decline in the ratio would signal the monetization phase is dominating; a renewed rise would signal continued buildout dominance.

SMH holdings extend beyond chip designers to include semiconductor equipment makers: Applied Materials, Lam Research, KLA Corporation, and ASML (through its ADR). These companies supply the machinery used to manufacture chips. Their performance tracks global semiconductor capex cycles but with a roughly 6-12 month lag behind chip company revenue.

ASML specifically supplies the extreme ultraviolet (EUV) lithography machines required for advanced node (3nm, 2nm) manufacturing. TSMC, Samsung, and Intel are the primary buyers of ASML machines, which cost approximately $200-400 million each. ASML's order book and capacity utilization are leading indicators for the semi cycle; strong orders predict continued chip demand 12-18 months out. Through 2025-2026, ASML orders have been strong but China-specific restrictions (US export controls on EUV to Chinese fabs) have constrained its potential.

Both SMH and QQQ carry significant Taiwan and China risk, but the exposure differs. SMH is more directly exposed through its 13 percent TSMC weight and the concentration of semiconductor production in Taiwan. Any escalation around Taiwan (Chinese invasion, blockade, or regional conflict) would hit SMH harder than QQQ. US-China chip export controls primarily affect Nvidia's China revenue (both ETFs affected but SMH more so given its AI-chip concentration).

For risk management, investors who want US-tech exposure while reducing Taiwan/China geopolitical risk can use XLK (US Technology Select Sector SPDR, excludes TSMC and reduces Nvidia weight) or SOXX (iShares Semiconductor ETF, caps individual names at ~8 percent). SOXX specifically reduces Nvidia concentration and TSMC weight versus SMH. The XLK-SMH-SOXX triangle offers a range of geopolitical-risk-adjusted semi/tech exposures.

The primary signal is the SMH-QQQ spread trajectory. A sustained outperformance by SMH above the April 2026 level would indicate continued AI buildout dominance. A shift toward QQQ outperformance would signal AI monetization taking over as the market's narrative. The switch typically happens around the mid-point of major tech capex cycles, which historically has been at 40-50 percent of total cycle revenue, a point the current AI cycle may be reaching.

Secondary signals: Nvidia's fiscal 2027 data center revenue guidance (next earnings May 2026), hyperscaler capex commentary in Q1 2026 earnings season (Microsoft, Google, Meta, Amazon, Oracle), ASML order book (measures forward semi capex), and Taiwan/China geopolitical developments. If the Iran-Hormuz situation expands to include any Taiwan-related tension, SMH is the most at-risk mainstream US tech ETF. Conversely, a resolution of geopolitical tensions combined with sustained AI enterprise adoption would support continued SMH outperformance through 2026-2027.