SMH ETF: Capturing the Scalable AI Hardware Wave

The investment thesis for SMH is built on a massive, multi-year infrastructure build-out. Artificial intelligence is driving one of the largest capital expenditure cycles the technology sector has ever seen, and the VanEck Semiconductor ETF is positioned to capture the revenue flowing from it. The scale of this opportunity is staggering, with the consensus Wall Street estimate for global AI-related data center capital expenditures in 2026 sitting at $527 billion. This isn't a fleeting trend; it's a fundamental shift in how tech giants operate, with companies like Meta recently raising their annual capex forecasts to fund the data centers that power AI development.

SMH provides a concentrated, scalable vehicle to ride this wave. The ETF tracks the performance of the 25 largest, U.S. listed semiconductor companies, a portfolio that directly benefits from this massive capex. Its top-heavy structure, where the top ten holdings account for over two-thirds of assets, ensures investors get direct exposure to the industry leaders-companies like Nvidia, Broadcom, and Micron-that are engineering the chips and systems for this new era. This focused approach offers a way to participate in the AI hardware boom without the idiosyncratic risks of picking individual stocks.

The growth trajectory is set to accelerate as the nature of AI workloads evolves. The industry is expected to shift from primarily training massive models to deploying them for real-time inference, where models answer user queries. This transition is creating a new, multi-billion dollar market for specialized chips. According to Deloitte, inference workloads alone are projected to account for roughly two-thirds of all compute and drive a market for inference-optimized chips that will grow to over US$50 billion in 2026. While the core data center infrastructure for high-performance AI will remain a multi-hundred-billion-dollar market, this shift opens a vast new TAM for efficient, purpose-built hardware. For SMH, the future growth of its holdings is inextricably linked to the industry's ability to convert this colossal capital spending into sustained, scalable revenue.

Scalability of the Business Model and Market Capture

The SMH ETF's structure is a direct reflection of the semiconductor industry's own scaling dynamics. Its index methodology is designed to capture growth by focusing on the largest, most liquid companies, which are precisely the firms best positioned to win the massive hyperscaler contracts driving the AI capex boom. By tracking the 25 largest, U.S. listed companies, the fund ensures its portfolio is anchored in industry leaders with the capital, manufacturing scale, and technological prowess to execute on multi-year infrastructure projects. This top-down approach aligns the ETF's performance with the sector's most powerful growth engines.

Yet this focus comes with a significant trade-off: extreme concentration. The portfolio is inherently top-heavy, with the top ten holdings accounting for over two-thirds of total assets. This concentration amplifies potential returns when the leading names perform, as seen in its 54.31% one-year return. But it also concentrates risk, making the fund's fate heavily dependent on a handful of stocks. For a growth investor, this is a double-edged sword-high conviction exposure to winners, but vulnerability to any stumble by a giant.

A more strategic gap exists in the fund's geographic scope. While the index includes foreign companies listed on U.S. exchanges, the fund's mandate to invest in U.S. exchange-listed companies excludes key global players like Taiwan Semiconductor Manufacturing Company (TSMC). TSMC is the undisputed leader in semiconductor manufacturing, a critical enabler for virtually all AI chips. By not holding this essential node in the supply chain, SMH potentially underweights a massive portion of the industry's value creation. This limitation represents a tangible gap in market capture, as the fund's exposure is confined to a subset of the global semiconductor ecosystem.

The bottom line is that SMH offers a scalable, concentrated bet on the AI hardware wave, but it is not a complete market proxy. Its structure favors the largest U.S.-listed beneficiaries of capex, but its concentration and geographic exclusion mean it captures a specific, albeit powerful, slice of the total addressable market. For investors seeking pure-play exposure to the industry's scaling leaders, the ETF delivers. For those wanting the broadest possible capture of the global semiconductor growth story, it leaves a notable piece of the puzzle out.

Growth Trajectory and Financial Impact

The financial impact of the AI capex wave is already evident in the stellar performance of SMH's holdings, but the path to sustained earnings growth is becoming more selective. The ETF's roughly 49% gain in 2025 and its annualized returns of around 30.9% over the past decade underscore the powerful tailwind from this infrastructure build-out. Yet, the current valuation reflects a market that has already priced in much of this optimism. SMH trades at a trailing P/E of nearly 33x, a premium that demands the underlying companies convert massive capital spending into robust, scalable operating earnings.

This is where the investment thesis faces its next test. The recent divergence in stock prices among AI hyperscalers shows investors are rotating away from pure infrastructure spenders. As noted, investors have rotated away from AI infrastructure companies where growth in operating earnings is under pressure and capex spending is debt-funded. The market is now rewarding companies that demonstrate a clear link between their AI investments and top-line revenue growth. For the semiconductor firms in SMH, this means the scalability of their business models hinges on their ability to not just supply chips, but to capture a larger share of the value generated by the AI platforms they enable.

The fund's structure provides a high degree of fidelity to this sector performance. With at least 80% of its total assets invested in securities that comprise the benchmark index, SMH offers a concentrated, liquid vehicle to ride the wave. Its top-heavy portfolio ensures exposure to the industry leaders best positioned to win the hyperscaler contracts. However, this also means the fund's fortunes are tied to the same selectivity that is now shaping the market. The ETF captures the growth of the winners, but its premium valuation leaves little room for error if any of its major holdings struggle to translate capex into profits.

The bottom line is that SMH is a high-conviction bet on the AI hardware winners, but it is not immune to the market's growing sophistication. The rally has been driven by the sheer scale of the opportunity, with consensus Wall Street estimates for global AI-related data center capital expenditures in 2026 at $527 billion. For the ETF to continue its growth trajectory, its holdings must navigate the transition from being beneficiaries of capex to becoming clear drivers of productivity and revenue. The current premium valuation prices in sustained high growth, making the fund's future performance a direct function of how successfully the semiconductor industry captures value in the next phase of the AI trade.

Catalysts, Risks, and What to Watch

The forward-looking setup for SMH hinges on a few critical catalysts and risks. The primary validation point is the continued upward revision of the consensus estimate for the group's 2026 capital spending, which has already climbed to $527 billion. Historically, analyst estimates have consistently underestimated AI capex, so any further upward adjustments would signal stronger-than-expected demand for the chips SMH holds. The key watch item is the shift from training to inference workloads in 2026. While inference-optimized chips are projected to reach a market of over US$50 billion, the broader compute demand for expensive, cutting-edge chips remains robust. This transition could alter the semiconductor mix, potentially creating winners in specialized inference chips while pressuring others focused solely on training hardware.

A major risk is valuation compression if the growth story falters. The ETF's premium valuation, trading at a trailing P/E of nearly 33x, prices in sustained high growth. If the anticipated capex spending fails to materialize or if the shift to inference reduces the overall chip intensity per AI model, the market could reassess the earnings power of SMH's holdings. This would be compounded by the fund's extreme concentration, where the top ten holdings account for over two-thirds of total assets. Any stumble by a giant like Nvidia or Broadcom would disproportionately impact the fund.

Another tangible risk is the potential for a broader tech sell-off. The recent divergence in AI stock performance shows investors are becoming more selective, rotating away from AI infrastructure companies where growth in operating earnings is under pressure. If sentiment turns negative across the tech sector, the concentrated, high-beta nature of SMH could amplify losses. The bottom line is that SMH is positioned for a powerful growth cycle, but its performance will be dictated by the accuracy of the $527 billion capex forecast, the successful navigation of the inference transition, and the market's patience with its premium valuation and concentrated structure.