Big Tech’s AI spending spree isn’t done

The AI spending scare looks early

Semiconductor stocks got whacked because investors started wondering if Big Tech had already done its big AI shopping spree. Bank of America’s take is basically: not even close.

It now sees hyperscaler capex climbing above $1.2 trillion over the next 12 months, up from about $700 billion. Morgan Stanley is singing a similar tune, with roughly $800 billion in 2026 and $1.2 trillion in 2027. Translation: the AI money train may be running hotter, not cooling off.

It’s not just a Nvidia story anymore

For a while, the market treated AI like a one-stock parade with Nvidia out front and everyone else waving from the sidewalk. But BofA’s list of Buy-rated names says the spending wave is spreading across the whole stack:

• compute names like Nvidia, AMD, and Intel
• networking and interconnect plays like Marvell and Credo
• memory like Micron
• equipment makers like Applied Materials and KLA
• and, of course, Broadcom doing Broadcom things

That matters because when the trade broadens, the rally can get sturdier. Fewer “one horse, one track” vibes. More “entire stadium is buying popcorn.”

Why ETF investors should care

This is where semiconductor ETFs get interesting. Funds like SMH and SOXX already bundle a lot of the AI winners into one basket, while XSD leans more equal-weighted, which can help if the gains spread beyond the mega-caps.

And then there’s the memory angle. If AI demand keeps chewing through DRAM and HBM supply, specialized products like the Roundhill Memory ETF can get a little extra tailwind too.

Big picture: if hyperscaler spending really does keep ramping, the AI trade may have room to broaden from “chips that train models” to the entire ecosystem that builds, feeds, and ships them.