Every new generation of artificial intelligence (AI) model typically soaks up more data center computing capacity than the last. The latest batch of reasoning models spend more time "thinking" in the background compared to their predecessors, in order to render higher quality outputs. This process chews through up to 1,000 times more tokens (words and symbols), requiring far more computing power.
That's why, according to Nvidia (NVDA +5.93%) CEO Jensen Huang, data center operators could spend between $3 trillion and $4 trillion upgrading their infrastructure to meet demand from AI developers between now and 2030. That will include millions of graphics processing units (GPUs), which are the primary chips used in AI workloads, in addition to a wide variety of networking and other hardware components.
The iShares Semiconductor ETF (SOXX +5.00%) is an exchange-traded fund (ETF) that invests in many of the hardware companies in line to benefit from that AI infrastructure spending. Its largest positions include Nvidia, Advanced Micro Devices (AMD +18.37%), and Broadcom (AVGO 0.52%), which have fueled significant returns for the fund over the last few years.
Here's how the ETF could turn a consistent investment of $500 per month into $1 million from here.
Image source: Getty Images.
A concentrated portfolio of the best AI hardware stocks
Some ETFs hold hundreds or even thousands of different stocks, but the iShares Semiconductor ETF holds just 30. It exclusively invests in companies that design, manufacture, and distribute semiconductors, and specifically those that stand to benefit from megatrends like AI.
The top three holdings in the ETF certainly fit the bill, and they alone account for 25% of the entire value of its portfolio:
|
Stock |
iShares ETF Portfolio Weighting |
|---|---|
|
1. Advanced Micro Devices |
9.97% |
|
2. Nvidia |
7.62% |
|
3. Broadcom |
7.42% |
Data source: iShares. Portfolio weightings are accurate as of Nov. 3, 2025, and are subject to change.
Since the AI boom started gathering momentum at the beginning of 2023, those three stocks have delivered a blistering median return of 529%. For some perspective, the benchmark S&P 500 (^GSPC +1.46%) index is up just 76% over the same period:
Nvidia is the undisputed leader in the market for AI data center GPUs. Its latest Blackwell Ultra chips are designed specifically for AI reasoning models, delivering up to 50 times more performance than its original Hopper-based chips, which were released in 2022. Demand continues to outstrip supply, which is fueling significant revenue growth for the company.
AMD is one of Nvidia's closest competitors in the data center market. The company recently signed a blockbuster deal with ChatGPT creator OpenAI, which could be worth $90 billion by 2030. The start-up will use AMD's new MI450 GPUs next year; they will come in a complete data center rack called Helios, which could be one of the most powerful AI hardware stacks in the world.
Then there is Broadcom, which supplies some of the best networking equipment for data centers. It accelerates processing speeds by facilitating the faster transmission of data between chips and devices, which is key in AI workloads. The company also makes custom AI accelerators for hyperscale tech giants, which have become a popular alternative to GPUs.
Outside its top three positions, the iShares ETF holds a number of other top AI hardware companies:
- Micron Technology, which supplies some of the most powerful high-bandwidth memory solutions for data centers. Both Nvidia and AMD use them in their latest GPUs.
- Qualcomm, which just entered the AI data center market to compete with AMD and Nvidia. However, the company is already a leading supplier of AI chips for computers and smartphones.
- Taiwan Semiconductor Manufacturing, which is the world's largest semiconductor fabricator. It manufactures chips on behalf of some of the industry's leading chip designers, including Apple, Nvidia, AMD, and Qualcomm.
NASDAQ: SOXX
Key Data Points
Turning $500 per month into $1 million
The iShares Semiconductor ETF has delivered a compound annual return of 11.9% since its inception in 2001. But that return accelerated to 27.2% per year over the past decade, as mega forces like cloud computing and AI have fueled incredible demand for advanced chips.
Here's how much investors could earn by putting $500 into the iShares ETF every month, based on three different returns over three different time periods:
|
Compound Annual Return |
Balance After 10 Years |
Balance After 20 Years |
Balance After 30 Years |
|---|---|---|---|
|
11.9% |
$115,980 |
$493,354 |
$1,726,565 |
|
19.5% (midpoint) |
$185,594 |
$1,466,332 |
$10,328,241 |
|
27.2% |
$310,147 |
$4,870,100 |
$72,021,083 |
Calculations by author.
To be clear, it's entirely unrealistic to expect any ETF to return over 27% per year forever, because the law of large numbers eventually becomes a headwind. For example, Nvidia's market capitalization has grown from $360 billion to $4.8 trillion since the start of 2023, but the company now serves almost every possible customer in the AI data center market, so it's becoming harder to grow its enormous revenue base. In fact, its revenue growth is actually starting to decelerate:
NVDA Revenue (Quarterly YoY Growth) data by YCharts
However, if data center operators really do spend $4 trillion on AI infrastructure by 2030 as Jensen Huang expects, there could still be plenty of growth ahead for the semiconductor industry overall.
Plus, keep in mind other technologies like the internet, the smartphone, enterprise software, and cloud computing have also driven significant demand for chips over the last two decades. And once the AI boom inevitably slows, new technologies like autonomous vehicles, robotics, and quantum computing could take over as the dominant sources of hardware demand.
Nevertheless, per the above table, the iShares Semiconductor ETF could turn $500 per month into $1 million in under 30 years even if its annual return reverts back to its long-term average of 11.9%.
