Semiconductors are the beating heart of the artificial intelligence (AI) revolution. Graphics processing units (GPUs), AI accelerators, high-bandwidth memory, and networking equipment fill modern data centers, delivering the computing capacity developers need to create advanced AI software.
Data center spending is growing each year, and Nvidia (NVDA +2.33%) CEO Jensen Huang predicts it will top $1 trillion annually by 2028 as tech giants and start-ups alike battle for AI supremacy.
The iShares Semiconductor ETF (SOXX +2.39%) holds 30 different stocks that could be massive winners if Huang is right. It's currently down 16% in 2025 amid the broader market sell off that was triggered by simmering global trade tensions, but here's why investors might want to look past the short-term noise and buy it now.
Image source: Getty Images.
Every top semiconductor stock packed into one ETF
The iShares Semiconductor ETF was established in 2001, so it has helped investors navigate numerous hardware booms driven by technologies like the internet, the smartphone, enterprise software, and cloud computing. Most suppliers of chips and components are now focusing on AI, because that's where the demand has shifted.
The top five holdings in the iShares ETF feature some of the biggest hardware names in the AI space, and they represent 37.9% of the total value of the portfolio:
|
Stock |
iShares ETF Portfolio Weighting |
|---|---|
|
1. Broadcom |
8.69% |
|
2. Nvidia |
8.01% |
|
3. Texas Instruments |
7.49% |
|
4. Advanced Micro Devices |
6.97% |
|
5. Qualcomm |
6.81% |
Data source: iShares. Portfolio weightings are accurate as of April 25, 2025, and are subject to change.
Broadcom makes AI accelerators for three unnamed hyperscale customers which can be customized to suit their needs, so they are a great alternative to traditional GPUs from suppliers like Nvidia. Broadcom thinks it could capture up to $90 billion in annual revenue from those three customers alone by fiscal 2027. Beyond chips, the company sells data center switches and networking equipment to facilitate rapid processing speeds, which is also critical for AI developers.
Despite growing competitive threats, Nvidia's data center GPUs are still the benchmark in AI hardware. The company recently unveiled its Blackwell Ultra GB300 GPU, which delivers a 50-fold performance bump in certain configurations compared to its old Hopper-based H100. The Blackwell Ultra architecture was designed for "reasoning" AI models, which require up to 100 times more computing power than traditional large language models, according to Jensen Huang.
NASDAQ: NVDA
Key Data Points
Advanced Micro Devices (AMD) is an emerging powerhouse in the data center space. It has its own lineup of AI GPUs, and its latest MI355X is built on a new architecture called CDNA (Compute DNA) 4, which was designed to rival Nvidia's original Blackwell architecture. However, those chips won't ship to customers in high volumes until midyear, so Nvidia still has a significant first-mover advantage. But it's not all about the data center, because AMD is also a top supplier of AI chips for personal computers, which could be a big growth market in the future.
Outside of the above stocks, the iShares ETF holds a number of other prominent AI hardware names. They include Micron Technology, which supplies memory and storage chips, and Taiwan Semiconductor Manufacturing, which fabricates most of the GPUs designed by Nvidia and AMD.
The iShares ETF has a long track record of success
President Donald Trump imposed a broad 10% tariff on all imported goods from America's trading partners in early April, in addition to a series of higher "reciprocal tariffs" on specific countries. However, semiconductors are exempt from the reciprocal levies, mainly because leading the AI race is a matter of national security for the U.S. Nevertheless, trade tensions could drive an economic slowdown, which might affect demand for chips in the short term as data center operators reevaluate their budgets.
Data by YCharts.
But here's the good news: The iShares Semiconductor ETF has delivered a compound annual return of 10.4% since it was established in 2001, beating the average annual gain of 7.8% in the S&P 500 over the same period. Plus, the ETF has delivered an accelerated annual return of 20.9% over the last 10 years specifically, thanks to the broad adoption of technologies like enterprise software, cloud computing, and now AI.
The point is the iShares ETF has weathered a number of economic shocks during its 24-year history -- including President Trump's last tariff saga in 2018 -- while still delivering strong returns.
If data center spending does grow to an annual rate of $1 trillion by 2028 as Jensen Huang predicts, then those returns are likely to continue (if not accelerate further). As a result, this might be a great time to buy the iShares ETF, especially considering its year-to-date dip of 16%.
