Top artificial intelligence (AI) developers like OpenAI and Anthropic continue to launch new models, with each one being "smarter" and more capable than the last. However, every new model also soaks up more computing power than its predecessors, so it requires significantly more data center capacity.
Some of the biggest suppliers of AI infrastructure, chips, and components include Nvidia (NVDA 1.36%), Advanced Micro Devices (AMD 3.31%), and Micron Technology (MU 2.74%). Shares in those three companies have surged by an average of 119% in 2025 alone, crushing the S&P 500 index, which is up just 18%.
In other words, investors who had little or no exposure to the AI semiconductor space this year probably underperformed the broader market.
Fortunately, there's a simple way to buy a slice of this hypergrowth industry without having to pick winners and losers. The iShares Semiconductor ETF (SOXX 2.87%) is an exchange-traded fund (ETF) that exclusively invests in companies like Nvidia, AMD, Micron, and many of their peers. Here's how it could turn an investment of $250,000 into $1 million over the next 10 years.
Image source: Getty Images.
The best AI hardware stocks packed into one ETF
The iShares Semiconductor ETF solely invests in American companies that design, distribute, and manufacture chips and components, but primarily those benefiting from opportunities like AI. Its narrow focus is the reason its portfolio only includes 30 stocks. The fund is also quite top heavy, with its three largest holdings sporting a combined weighting of 22.7%:
|
Stock |
iShares ETF Portfolio Weighting |
|---|---|
|
1. Nvidia |
8.22% |
|
2. Advanced Micro Devices (AMD) |
7.62% |
|
3. Micron Technology |
6.88% |
Data source: iShares. Portfolio weightings are accurate as of Dec. 24, 2025, and are subject to change.
Nvidia's graphics processing units (GPUs) are the best chips on the market for developing AI models. Its current Blackwell Ultra lineup was designed to deliver sufficient computing power for the industry's latest reasoning models, like OpenAI's GPT 5.2, Anthropic's Claude 4.5, and Alphabet's Gemini 3. I think Nvidia stock is still cheap despite its 41% gain in 2025, which leaves plenty of room for more upside from here.
AMD is chasing Nvidia in the market for data center chips. The company's latest MI350 Series GPUs have captured some of its key rival's top customers, but they still fall modestly short in terms of performance. Next year, AMD plans to launch its new MI400 GPUs as part of a fully integrated data center rack called Helios, which could deliver 10 times more performance than its MI350 GPUs. This company might be a real threat to Nvidia at that point.
Then there is Micron Technology, which is one of the world's leading suppliers of memory and storage chips. Both Nvidia and AMD have integrated Micron's HBM3E (high-bandwidth memory) solutions into their data center GPUs to unlock their maximum processing speeds. Micron is already sold out of its entire 2026 supply of data center memory, including its upcoming HBM4E solution, which is poised to deliver improvements in both capacity and energy efficiency.
Outside of its top three positions, the iShares ETF holds several other top AI semiconductor stocks like Broadcom, Texas Instruments, and Taiwan Semiconductor Manufacturing.
Turning $250,000 into $1 million over the next decade
The iShares Semiconductor ETF is on track to end 2025 with an eye-popping 43% return. That isn't sustainable over the long run, but the ETF has delivered a compound annual return of 27.2% over the past decade, so it's no stranger to high growth. Even its more modest average annual gain of 11.8% since its inception in 2001 is still better than the performance of the S&P 500 over the same period.
NASDAQ: SOXX
Key Data Points
Here's how long it might take for the iShares ETF to turn an investment of $250,000 into $1 million, based on three different average annual returns:
|
Initial Investment |
Compound Annual Return |
Time to Reach $1 Million |
|---|---|---|
|
$250,000 |
11.8% |
13 Years |
|
$250,000 |
19.5% (midpoint) |
8 Years |
|
$250,000 |
27.2% |
6 Years |
Calculations by author.
According to Nvidia CEO Jensen Huang, annual spending on AI data center infrastructure and chips could reach $4 trillion by 2030. If that happens, the iShares ETF is likely to deliver compound annual returns in excess of 20% for the foreseeable future. But to be perfectly clear, that wouldn't be sustainable under normal circumstances because of the law of large numbers.
For example, Nvidia is already a $4.6 trillion company. If its market cap grew by 27.2% annually for the next decade, it would be worth almost $50 trillion. The output of the entire U.S. economy was less than $30 trillion in 2024, so although AI is a revolutionary technology, it's important for investors to temper their expectations.
Fortunately, the iShares ETF can still turn $250,000 into $1 million over the next 10 years even if its annual returns moderate to slightly below 20%. Per the above table, if investors are a little more patient, the ETF could also help them join the million-dollar club in 13 years even if its returns suffer a significant reversion back to their long-term average of 11.8%.
But I think its returns are likely to remain elevated for now, considering practically every chipmaker is experiencing more demand than it can possibly supply.
