Nvidia (NVDA +0.78%) CEO Jensen Huang thinks that data center operators will spend $1 trillion every year on chips and infrastructure by 2028 to meet growing demand for computing capacity from next-generation artificial intelligence (AI) models.
That spending will be an enormous tailwind for Nvidia, which supplies the world's most powerful data center chips for AI development. But the benefits will also flow through to the company's competitors, not to mention suppliers of other data center hardware components. There is an opportunity for investors to profit from this tech revolution, and buying an exchange-traded fund (ETF) might be the simplest way to do so.
The iShares Semiconductor ETF (SOXX +1.42%) invests exclusively in suppliers of chips and components, and its top holdings happen to be three of the biggest names in AI: Nvidia, Broadcom (AVGO 1.05%), and Advanced Micro Devices (AMD +0.41%). The ETF has outperformed the broader stock market since its establishment in 2001, and here's how it could turn an investment of $250,000 into $1 million within the coming decade.
Image source: Getty Images.
The biggest names in AI hardware in one ETF
Some ETFs hold thousands of different stocks, but the iShares Semiconductor ETF holds just 30. It aims to offer investors exposure to companies that design, manufacture, and distribute semiconductors, primarily those that stand to benefit from megatrends such as AI.
Since the ETF was established in 2001, it has helped investors successfully navigate several tech revolutions driven by the internet, enterprise software, smartphones, and cloud computing. It's now heavily geared toward AI, and its top five holdings are among the biggest names in the hardware side of the industry:
|
Stock |
iShares ETF Portfolio Weighting |
|---|---|
|
1. Broadcom |
10.07% |
|
2. Nvidia |
8.74% |
|
3. Texas Instruments |
7.49% |
|
4. Advanced Micro Devices (AMD) |
7.30% |
|
5. Qualcomm |
5.83% |
Data source: iShares. Portfolio weightings are accurate as of June 4, 2025, and are subject to change. ETF = exchange-traded fund.
Nvidia's graphics processing units (GPUs) are the most popular data center chips among AI developers. The company's latest GPU architectures, Blackwell and Blackwell Ultra, are designed for a new generation of AI models capable of 'reasoning,' which means they spend time thinking in the background to generate the most accurate responses.
Jensen Huang says some of these models consume up to 1,000 times more computing capacity than traditional one-shot large language models (LMs), hence his lofty spending forecast mentioned earlier.
NASDAQ: NVDA
Key Data Points
Amazon, Microsoft, and Alphabet are three of Nvidia's biggest customers. They are seasoned data center operators because of their industry-leading cloud platforms, but they are now racing to build AI infrastructure to meet surging demand from developers.
But these companies are also trying to diversify their hardware portfolios by designing their own chips in collaboration with suppliers like Broadcom, which helps with the design and manufacturing processes. Broadcom is targeting a $90 billion market opportunity for its custom AI accelerator chips by 2027, with just three customers already on board and more in the pipeline. Plus, the company is a leading supplier of networking equipment, which helps to extract the most performance from AI chips.
Then there is AMD, which released a line of GPUs to compete directly with Nvidia in the data center. This year, the company will start shipping its latest chips based on its CNA (Compute DNA) 4 architecture, which was designed to rival Blackwell. AMD is also already a leader in AI chips for personal computers, which could be a major growth area in the future.
Investors will also find other leading AI chip stocks, such as Micron Technology, Taiwan Semiconductor Manufacturing, and Arm Holdings, outside the top five holdings in the iShares ETF.
Turning $250,000 into $1 million in the next decade
The iShares Semiconductor ETF has delivered a compound annual return of 10.4% since its establishment in 2001, outperforming the average annual gain of 7.9% in the S&P 500 over the same period. But the ETF has delivered an accelerated annual return of 20.9% over the past decade, driven by the accelerating adoption of technologies like cloud computing and AI.
If the iShares ETF continues to deliver annual gains of 20.9%, it could turn a $250,000 investment into over $1.6 million in the next decade. It won't be easy, but if AI infrastructure spending grows to $1 trillion per year by 2028, as Jensen Huang expects, it certainly isn't out of the question.
However, even if the ETF averages an annual gain of 15.6% over the next 10 years, that would be enough to turn $250,000 into $1 million:
|
Starting Balance |
Compound Annual Return |
Balance in 10 Years |
|---|---|---|
|
$250,000 |
10.4% |
$672,404 |
|
$250,000 |
15.6% (midpoint) |
$1,065,413 |
|
$250,000 |
20.9% |
$1,668,026 |
Calculations by author.
Last year, Huang said data center operators could earn $5 over four years for every $1 they spend on Nvidia's AI chips and infrastructure by renting the computing capacity to AI developers. If those economics are accurate, data center operators like Amazon, Microsoft, and Alphabet are likely to continue investing heavily in new infrastructure long into the future.
Plus, every new generation of AI models typically requires even more computing capacity than the last, so it's possible that Huang's spending forecasts will prove to be conservative when we look back on this moment. In any case, the iShares ETF could be a great addition to a diversified portfolio.
