AI's Next Decade May Not Belong to One Company — Bank of America Bets on the Chipmakers Behind the Scenes

That thesis, laid out by longtime BofA semiconductor analyst Vivek Arya, has put a fresh spotlight on a group of companies that don't always make headlines the way Nvidia does, but whose fingerprints are on nearly every piece of AI infrastructure being built today: Broadcom, AMD, Lam Research, Cadence Design Systems, and Marvell Technology.

Why Bank of America Thinks This Is Bigger Than One Chip Company

Bank of America expects the total semiconductor market to hit $2 trillion by 2030, which would work out to a 20% compound annual growth rate through the end of the decade — more than double the roughly 9% pace the industry averaged over the previous ten years. That's a striking jump, and it's part of why BofA has been steadily raising its near-term forecasts too. The bank now sees the global semiconductor market reaching $1.3 trillion in 2026, up from a $1.0 trillion forecast just four months earlier.

Zooming in specifically on AI data centers, the numbers get even bigger. Bank of America estimates that the total addressable market for AI data center systems will exceed $1.2 trillion by 2030, growing at a compound annual rate of roughly 38%, with AI accelerators alone representing a $900 billion opportunity.

Arya's reasoning goes beyond just bullish math. He's argued that this AI investment cycle looks different — and more durable — than past tech spending booms. The current wave of spending is being fueled by hyperscalers, sovereign AI projects, and growing enterprise adoption, with capital expenditures potentially tripling between 2025 and 2030. Unlike previous cycles, he points out, this buildout is backed by strong, reliable cash flow from major cloud providers and government-backed buyers rather than speculative borrowing. He's noted that the top five U.S. cloud companies are expected to put roughly 25% of their 2025 sales toward capital spending, supported by operating cash flow margins above 30%.

There's also a structural reason AI spending may hold up better than past hardware cycles: everyday consumers don't need to buy new devices for AI to reach them, meaning tools like ChatGPT can scale to more users instantly without requiring a costly hardware refresh on the consumer side.

Meet the Companies BofA Says Are Powering the Boom

Nvidia and Broadcom: The "Brain" and the "Nervous System"

Nvidia remains the most recognizable name in AI chips, and BofA still sees room for it to run. One recent BofA analysis described Nvidia as trading at a price-to-earnings growth ratio of roughly 0.6x — a notable discount compared to the broader S&P 500's multiple of about 2x — while projecting the company could generate around $500 billion in free cash flow over the next three years. Arya has also pushed back on comparisons between Nvidia and traditional chipmakers, noting that while the average semiconductor sells for about $2.40, a single Nvidia graphics processing unit can sell for roughly $30,000.

If Nvidia is the brain of modern AI infrastructure, BofA has described Broadcom as its nervous system. Broadcom has shifted from being a diversified components supplier to a core pillar of AI infrastructure, with its market capitalization climbing to nearly $1.6 trillion.

AMD and Marvell: The "AI Compute" Leaders

Among individual stock calls, BofA has been leaning further into what it calls "AI compute" leaders, specifically naming Marvell Technology and AMD as companies it's doubling down on. Both companies have carved out growing roles in custom AI chips and the networking components that move massive amounts of data between processors — an increasingly critical piece of the AI infrastructure puzzle as data centers scale up.

Lam Research: The Equipment Behind the Chips

None of this hardware gets built without the machines that manufacture it, and that's where Lam Research comes in. BofA has described Lam Research as a leader in wafer fabrication equipment, calling it essential to the increasingly complex manufacturing process behind next-generation chips. The bank has also flagged opportunities in chipmaking equipment broadly, pointing to both Lam Research and Applied Materials as key beneficiaries of rising manufacturing complexity.

Cadence: The Software Behind the Chip Designs

Perhaps the least talked-about name on the list, Cadence Design Systems plays a foundational role that most consumers never see. BofA has called Cadence the software backbone of chip design, noting its "catch-up" potential as more companies outside the traditional chip industry attempt to design their own custom silicon. The bank has also pointed to a looming rebound for electronic design automation software companies like Cadence and Synopsys as chip designs grow more complex. Arya has framed this moment as a period of "leading logic intensity," where the sheer complexity of modern chip designs is forcing a ramp-up in the specialized tools needed to build them.

Why BofA Thinks "Picks and Shovels" Beat Picking Just One Winner

Bank of America has highlighted eight key beneficiaries of its 2026 semiconductor upgrade — Nvidia, Broadcom, Marvell, AMD, Applied Materials, Lam Research, Cadence, and Synopsys — spanning AI compute, networking, manufacturing equipment, and chip-design software. As a group, these stocks have significantly outperformed the S&P 500 over the past 12 to 18 months, and analysts expect them to remain central beneficiaries of AI data center spending in 2026, even with periodic pullbacks as expectations reset.

Arya has even suggested that picking winners in this sector doesn't require deep expertise. He's told reporters that investing in semiconductors can be as simple as sorting companies by gross margin and buying the top performers, adding that these companies typically hold roughly 70% to 75% market share in their respective niches — a level of dominance that's fairly normal for leaders in the tech sector.

That said, BofA isn't promising a smooth ride. The bank has cautioned that the path to a trillion-dollar AI data center market will likely be "bumpy," largely because of the extremely high construction costs involved in building modern AI data centers. Outside of AI-focused chips, the broader consumer electronics market — PCs and smartphones — is expected to stay relatively weak until 2027, which could limit upside for some of the more traditional semiconductor names.

What This Means Looking Ahead

Bank of America expects AI data center capital expenditures to keep growing at 25% to 30% per year, pushing the total AI data center market to somewhere between $1.2 trillion and $1.4 trillion by 2030. The bank also anticipates a significant rebound for the memory chip market in 2026, continued strength for wafer-fab equipment makers, and steady, less-cyclical growth for the electronic design automation and intellectual property side of the business.

For everyday investors who don't want to bet everything on one stock, BofA's own research suggests that diversified semiconductor and technology ETFs — such as SOXX, SMH, XSD, PSI, and QQQ — offer a way to gain broad exposure to this trend, with the option to layer in individual positions in the strongest names rather than concentrating risk in a single company.

The bigger takeaway from Bank of America's outlook is a shift in how to think about the AI boom altogether. It's not just about who builds the flashiest chatbot or the most advanced model — it's about who's supplying the picks and shovels for the entire industry. And according to BofA, that list runs a lot deeper than just one household name.