Beyond the GPU: Why 2026 is the Year of the NPU Trade. – March 30, 2026

As the semiconductor industry pivots from a 'Training Era' dominated by GPUs to a 'Deployment Era' favoring cost-efficient Neural Processing Units (NPUs) and custom ASICs, investors are re-evaluating the structure of their sector exposure. This analysis compares three prominent semiconductor ETFs—the VanEck Semiconductor ETF (SMH.US), the SPDR S&P Semiconductor ETF (XSD.US), and the iShares Semiconductor ETF (SOXX.US)—to assess their technical posture amid this technological shift. The analysis focuses on how their different construction methodologies may influence performance and volatility.

The VanEck Semiconductor ETF (SMH.US), which is heavily weighted towards market-capitalization leaders, currently trades at $364.35. Its technical posture is bearish in the near term, with the price below its 50-day Simple Moving Average (SMA) of $400.26. However, it remains above its longer-term 200-day SMA of $341.76. Momentum indicators are soft, with a Relative Strength Index (RSI) of 37.97 suggesting conditions are approaching oversold territory. The Average Directional Index (ADX) at 22.88 indicates a weak underlying trend.

Technically, SMH.US has registered an overall signal score of -6, which falls into a neutral classification with low confidence. The price has broken below its pivot point of $384.61 and first support level at $376.53, a bearish development. The primary resistance level is now seen at $388.91. Critically, several strong bearish divergences on the MACD indicator have been noted, where price reached higher highs while the indicator formed lower highs, suggesting waning upward momentum. The current price action appears to be technically driven, as no specific fundamental catalyst was identified.

Employing an equal-weighted strategy that gives smaller firms the same influence as giants, the SPDR S&P Semiconductor ETF (XSD.US) is positioned to capture growth from a wider array of companies, including emerging NPU designers. The ETF is priced at $319.15. Unlike its peers, XSD.US maintains a bullish long-term trend, with its price above the 200-day SMA of $314.43. However, it has recently fallen below its 50-day SMA of $346.59.

Momentum is neutral, with an RSI of 42.53, while its ADX of 28.32 suggests a moderately stronger trend than its peers. XSD.US carries an overall signal score of -2, the best of the three, but is still classified as neutral with low confidence. The fund has breached its pivot point at $331.17 and first support at $325.06, with new resistance noted at $334.43. A significant number of strong bearish divergences on both the MACD and RSI indicators signal that its bullish trend may be losing underlying strength.

The iShares Semiconductor ETF (SOXX.US), another broad-based fund, is currently trading at $311.51. It shares a near-term bearish trend with SMH.US, trading below its 50-day SMA of $345.10 but holding above its 200-day SMA of $290.06. Its momentum profile is neutral, with an RSI of 41.26 and an ADX of 25.00 indicating a moderate trend strength.

SOXX.US has the lowest technical score of the group at -7, though this still results in a neutral overall signal with low confidence. The price has fallen through key support, including its pivot point at $332.42 and first support at $324.95, with overhead resistance now at $336.32. Similar to the other ETFs, multiple bearish MACD and RSI divergences have been flagged, suggesting that recent price peaks were not confirmed by momentum, pointing to underlying weakness.

While all three ETFs received a neutral overall signal with low confidence, key technical differences emerge. XSD.US is the only one of the three to maintain a bullish long-term trend, though this is tempered by a recent break below its 50-day moving average. It also has the least negative signal score at -2, compared to -6 for SMH.US and -7 for SOXX.US, suggesting a slightly better short-term technical posture.

A critical shared characteristic is the prevalence of strong bearish divergences across all three funds. This pattern, where price advances are not matched by momentum indicators, suggests widespread technical exhaustion in the semiconductor sector. Regarding volatility, the hypothesis that SMH.US’s concentration could lead to higher volatility is not reflected in recent data. The Average True Range (ATR) as a percentage of price is similar for all three: 2.94% for SMH.US, 3.13% for XSD.US, and 3.11% for SOXX.US. Any future volatility divergence would likely stem from thematic shifts rather than current trading patterns.

The choice between these ETFs hinges on an investor's outlook for the semiconductor industry.

• SMH.US may be favored by those with high conviction in the continued dominance of large-cap GPU leaders. Its concentrated nature offers direct exposure to the sector's biggest names but carries risk if the market rotates towards specialized NPU and ASIC designers.

• XSD.US is structured for investors seeking broader, diversified exposure. Its equal-weighted approach could mitigate concentration risk and capture upside from smaller innovators in the NPU space, aligning with the 'Deployment Era' theme.

• SOXX.US offers a general, diversified play on the sector's overall growth trajectory, without a specific bet on either market-cap concentration or equal-weighting. Its current bearish trend and weaker technical score warrant caution in the near term.

Given the low confidence in current signals and the widespread bearish divergences, a cautious approach is warranted for the sector. The technical weakness suggests the potential for price corrections across the semiconductor landscape.

Disclaimer: This article is for educational and informational purposes only and does not constitute investment advice, a recommendation, or an offer to buy or sell any securities. Investors should conduct their own research and consult with a qualified financial advisor before making any investment decisions. Past performance is not indicative of future results. Trading in financial markets carries inherent risks, including the potential loss of principal. The information and technical data presented herein are sourced from internal analytics tools, are based on available data as of the publication date, and may be subject to change. No guarantee is made regarding the accuracy or completeness of the information provided.