Vivek Arya, an analyst at Bank of America, believes that the artificial intelligence craze has not cooled down, but is intensifying.
While AI skeptics point to eye-popping valuations as a reason for his candidacy, Arya said the industry is only at the “midpoint” of a decade-long transformation led by Nvidia (NVDA) and Broadcom (AVGO).
In a report titled “2026 Outlook: Choppy but Still Optimistic,” Arya predicted that global semiconductor sales will increase by 30% year-on-year, which will ultimately push the industry to exceed the historic US$1 trillion annual sales milestone in 2026.
Arya noted that there is a strong belief that the company has a “moat quantified through its profit structure.” In addition to Nvidia and Broadcom, he highlighted four other large semiconductor companies – Lam Research (LRCX), KLA (KLAC), Analog Devices (ADI) and Cadence Design Systems (CDNS) – as his top picks for 2026.
“I always say investing in semi-finished products is very simple,” Arya told reporters on a Dec. 19 conference call. “You don’t need any sell-side analysts to do this. Just take all your companies, sort them by gross margin, and buy the top five, and you can’t go wrong.”
Bank of America predicts that by 2030, the total potential market for artificial intelligence data center systems will exceed US$1.2 trillion, with a compound annual growth rate of 38%. AI accelerators alone represent a $900 billion opportunity.
Despite these staggering numbers, the market remains cautious because AI data centers are expensive. According to Bank of America, a typical 1-gigawatt facility requires more than $60 billion in capital expenditures, about half of which goes directly to hardware.
This raises the question: Can the return on investment be realized?
Read more: How to protect your portfolio from the AI bubble
Arya remains optimistic, viewing current spending as both “offensive and defensive.” In other words, Big Tech has no choice but to invest to protect its existing empire.
Arya said Nvidia is the world’s largest company by market capitalization and currently operates in “different galaxies.”
With Nvidia shares up more than 40% so far this year, Arya warned against comparing the artificial intelligence leader to traditional chipmakers. While the average selling price for chips is $2.40, Nvidia graphics processing units (GPUs) cost about $30,000.
While some worry that Nvidia’s market capitalization has reached its upper limit, Bank of America noted that free cash flow is expected to reach $500 billion over the next three years and that its valuation “remains very cheap” when adjusted for growth.
