The AI boom isn't slowing down. It's just getting repackaged. Bank of America's latest semiconductor outlook makes the case that if you're nervous about chasing individual chip stocks at current prices, ETFs might be your best bet for riding what could be a historic wave of AI spending.

In its report "2026 Year Ahead: choppy, still cheerful," cited by Yahoo Finance, BofA analyst Vivek Arya argues that the semiconductor industry is only halfway through a decade-long transformation. His prediction? Global chip sales will jump 30% year over year, finally pushing the sector past the $1 trillion annual revenue milestone by 2026.

The High-Margin Winners Take All

Arya's investment approach is refreshingly simple: rank semiconductor companies by their gross margins and bet on the leaders. The logic is that companies with pricing power and operational efficiency will capture the lion's share of AI infrastructure spending. His six top picks for 2026 are Nvidia Corp (NVDA), Broadcom Inc (AVGO), Lam Research Corp (LRCX), KLA Corp (KLAC), Analog Devices Inc (ADI), and Cadence Design Systems Inc (CDNS). These companies don't just lead their markets—they dominate them, typically holding market shares between 70% and 75%.

This "margin moat" strategy maps nicely onto how major semiconductor ETFs are already structured. Funds like the VanEck Semiconductor ETF (SMH) and the iShares Semiconductor ETF (SOXX) hold substantial positions in Nvidia and Broadcom, while also giving investors exposure to the equipment makers and chip designers who profit every time a tech giant writes another massive check for AI infrastructure.

Why a Trillion Dollars Actually Matters

According to BofA, AI data center systems alone represent a total addressable market exceeding $1.2 trillion by 2030, growing at a compound annual rate of 38%. AI accelerators account for roughly $900 billion of that opportunity, which puts semiconductors at the absolute center of AI expansion.

But here's where it gets interesting: the sheer cost of building out AI infrastructure has made some investors skittish. A single 1-gigawatt data center can require more than $60 billion in capital expenditures, with nearly half of that flowing directly into hardware. The obvious question is whether the returns will justify the spending. Arya's view is that Big Tech doesn't really have a choice. The investment is both offensive and defensive—it's about protecting existing platforms as much as building new ones.

Nvidia and Broadcom Anchor the Trade

Nvidia remains the gravitational center of most semiconductor ETFs. Arya cautions against valuing it like a traditional chipmaker, noting that while the average semiconductor sells for around $2.40, Nvidia's GPUs command prices around $30,000. Despite the stock's massive rally, BofA believes the valuation still makes sense when you factor in the growth trajectory.

Broadcom has become equally essential to AI infrastructure, thanks to its custom silicon business serving hyperscalers who want to reduce their dependence on Nvidia. That strategic shift is increasingly reflected in ETF weightings.

ETFs Smooth Out the Bumps

Arya warns that the path to $1 trillion won't be a straight line, and no individual stock is immune to risk. But semiconductor ETFs let investors express long-term confidence in AI's growth while spreading exposure across multiple winners. As the AI cycle matures and evolves, that diversification might prove particularly valuable for investors who believe in the thesis but don't want to bet everything on one or two names.