BofA lifts semiconductor TAM to $2T by 2030

Bank of America has pushed its semiconductor market forecast materially higher, and the scale of the revision helps explain why chip names remain central to the AI buildout story. In an April 7 research note, BofA analyst Vivek Arya and colleagues raised their estimate for the industry’s total addressable market to about $1.3 trillion by 2026, up from a prior $1 trillion view, and said the market could reach roughly $2 trillion by 2030. (moomoo.com) The revised forecast is not just a call on one product cycle. BofA tied the move to a broader data-center expansion spanning AI accelerators, networking gear and memory, with industrial demand as a secondary support. (goldmansachs.com) The bank also framed 2025-2030 growth at about 20% annually, versus roughly 9% over the past decade. ### Why did BofA move the number so much? The April 7 note said AI and data-center demand were the core reasons for the upgrade. BofA highlighted computing, networking and storage as the main engines, rather than treating semiconductors as a single undifferentiated market. (moomoo.com) That matters because the forecast implies growth is spreading beyond GPUs alone. BofA’s published themes included AI computing, semiconductor equipment, analog chips, consumer electronics and EDA software, suggesting the bank sees the expansion reaching suppliers across the stack. (moomoo.com) ### Which parts of the chip market are doing the heaviest lifting? Memory was one of the clearest swing factors in BofA’s note. The bank said the memory market could be exceptionally strong in 2026, with sales up 168% year over year, including 183% growth for DRAM and 151% for NAND, helped by AI-driven HBM demand, tight supply and higher pricing. (moomoo.com) Networking is the other major piece. Yahoo Finance’s report on the note said Arya named Nvidia, Broadcom, Marvell and AMD among the companies best positioned to capture AI-related demand, particularly as hyperscalers keep expanding training and inference infrastructure. (moomoo.com) ### Where does optical networking fit into this? Goldman Sachs last week put a much larger number around the networking side of the AI trade. Goldman said the total addressable market for AI networking could rise nearly ninefold to $154 billion by 2028, from about $15 billion in 2026, spanning both scale-up and scale-out architectures. (moomoo.com) Goldman said scale-up networking would account for about 69% of that opportunity, or roughly $106 billion. Its research pointed to rising optical content as GPU clusters get larger and data-transfer demands increase inside and between racks. (finance.yahoo.com) ### How aggressive is BofA compared with other industry forecasts? McKinsey said in a January analysis that most market forecasts put the semiconductor industry at roughly $630 billion to $680 billion in 2024 and around $1 trillion to $1.1 trillion by 2030. Against that backdrop, BofA’s $2 trillion by 2030 call sits well above more conventional industry ranges. (goldmansachs.com) That gap does not prove BofA is wrong, but it does show how much of its thesis rests on AI infrastructure sustaining a much faster growth rate than the industry delivered over the last decade. BofA’s own note explicitly ties that acceleration to data-center compute, networking and storage. (bwcfoworld.com) ### What should investors watch next? Nvidia’s earnings remain the next obvious checkpoint for this thesis, because spending on accelerators, networking and memory is flowing through the same AI data-center budgets that BofA and Goldman are modeling. Company commentary from Nvidia, Broadcom, Marvell, AMD and memory suppliers will give the next read on whether orders are broadening from training chips into optical links and HBM-heavy systems. (mckinsey.com) For now, the most concrete markers are already on the table: BofA’s April 7 note set the semiconductor TAM at $1.3 trillion by 2026 and $2 trillion by 2030, and Goldman’s May research put AI optical networking alone at $154 billion by 2028. (moomoo.com) (finance.yahoo.com)