Meet the Super Semiconductor ETF With 40% of Its Portfolio Parked in Micron, AMD, Broadcom, Nvidia, and Intel

Running artificial intelligence (AI) software requires an astronomical amount of computing capacity, which is typically delivered via large, centralized data centers. This infrastructure is powered by thousands of specialized chips and networking components, which are supplied by semiconductor companies like Micron Technology, Advanced Micro Devices, Broadcom, Nvidia, and Intel.

Nvidia CEO Jensen Huang believes that data center operators could spend up to $4 trillion on AI infrastructure per year by 2030, so this is the biggest financial opportunity in the semiconductor industry's history.

The iShares Semiconductor ETF (NASDAQ: SOXX) is an exchange-traded fund (ETF) that exclusively invests in suppliers of chips and computing components. It has delivered a blistering return of 330% since the AI boom started gathering momentum in early 2023, thanks to its large positions in Micron, AMD, Broadcom, Nvidia, Intel, and a host of other semiconductor giants.

Here's why more upside might be ahead for this ETF over the long term.

The world's best semiconductor stocks packed into one ETF

ETFs can hold hundreds, or even thousands, of different stocks, but the iShares Semiconductor ETF holds just 30. It only invests in U.S. companies that design, manufacture, and distribute chips and components, particularly those that stand to benefit from megatrends like AI. Below are its top five holdings, which represent 40% of the value of its portfolio.

Data source: iShares. Portfolio weightings are accurate as of May 14, 2026, and are subject to change.

Graphics processing units (GPUs) are the main data center chips used to develop and deploy AI models, and Nvidia's Blackwell series currently dominates the industry. The company will up the ante later this year when it starts shipping its Vera Rubin systems, which the company says will be so powerful that developers can train AI models using 75% fewer GPUs (compared to Blackwell).

However, Nvidia faces growing competition from AMD and Broadcom. AMD has released a portfolio of data center GPUs for AI workloads, but it's going in a slightly different direction with its upcoming MI450 AI accelerators, because they can be customized to suit the needs of specific customers. When configured in the company's Helios data center rack, MI450 accelerators can deliver a staggering 36 times more performance than its last generation of GPUs.