Semiconductor stocks took a hit recently on fears that Big Tech's AI spending had peaked. But Bank of America is pushing back hard, and its numbers are eye-popping. The bank now expects hyperscaler capital expenditures to top $1.2 trillion over the next 12 months, up from roughly $700 billion. Morgan Stanley is on board too, projecting about $800 billion in 2026 and $1.2 trillion in 2027.

For ETF investors, the question isn't whether AI spending is slowing—it's which funds are best positioned to ride this wave.

AI Spending Is Broadening Beyond GPUs

Bank of America's analysts have identified nine Buy-rated semiconductor stocks with at least 30% upside, and here's the kicker: they're not all about Nvidia. The list spans compute, memory, networking, and semiconductor equipment, which tells you something important about where the AI trade is heading.

The names include Marvell Technology Inc. (MRVL), Micron Technology Inc. (MU), Nvidia, Broadcom Inc. (AVGO), Advanced Micro Devices Inc. (AMD), Credo Technology Group Holding Ltd. (CRDO), Applied Materials Inc. (AMAT), KLA Corp. (KLAC), and Intel Corp. (INTC).

That breadth is a strong argument for diversified semiconductor ETFs rather than betting on a single winner.

Broad Semiconductor ETFs Cover the Entire AI Supply Chain

• The VanEck Semiconductor ETF (SMH) holds many of Bank of America's highest-conviction names, including Nvidia, Broadcom, AMD, Micron, Applied Materials, and KLA. Its market-cap-weighted portfolio gives you concentrated exposure to AI leaders while also capturing equipment makers that benefit from rising fabrication spending.
• The iShares Semiconductor ETF (SOXX) offers broader exposure across chip designers and equipment companies, making it well positioned if AI gains extend beyond mega-cap names.
• The SPDR S&P Semiconductor ETF (XSD) uses an equal-weighted methodology, giving greater weight to mid-cap semiconductor companies that could shine if the AI rally broadens.

Memory and Equipment Remain Key AI Bottlenecks

Micron stands out with 88% implied upside in Bank of America's estimates, highlighting memory as one of AI's biggest supply constraints. That's why specialized funds like the Roundhill Memory ETF (DRAM) are gaining traction. This fund focuses on memory-chip manufacturers and related suppliers, and it's been one of this year's standout AI-themed ETF launches as investors zero in on the memory segment.

The bullish outlook extends upstream too. Applied Materials and KLA underscore a simple reality: every additional AI accelerator requires more wafer fabrication, inspection, and manufacturing capacity. Broad semiconductor ETFs like SMH and SOXX already provide meaningful exposure to these equipment leaders.

The AI trade is no longer just about GPUs. Higher hyperscaler spending supports demand across memory, networking, chip equipment, and manufacturing, reducing the reliance on any single company. Semiconductor ETFs bundle all those exposures into one portfolio, making them a convenient way to play the trend.

So, if you're wondering whether the semiconductor rally has legs, the answer from the banks is a resounding yes—and the ETFs are ready to capture it.