BofA lifts semiconductor market to $1.29T
- Bank of America’s Vivek Arya raised his 2026 semiconductor revenue forecast on April 8 to about $1.3 trillion, arguing AI demand is overwhelming old models. - The eye-catcher is memory: IDC now sees total 2026 chip revenue at $1.29 trillion, with DRAM alone jumping to $418.6 billion. - That matters because even bullish forecasts still run into real bottlenecks — HBM supply and advanced packaging remain the hard constraints.
Semiconductors are having one of those moments where the numbers stop sounding incremental and start sounding absurd. Bank of America just yanked its 2026 chip-market forecast up to about $1.3 trillion, and that is not just a bullish tweak — it is a signal that AI spending is pulling the whole industry into a different size class. The gap was whether this was still a normal upcycle with some AI frosting on top. It doesn’t look like that anymore. It looks like AI infrastructure has become the market. ### What actually changed? On April 8, Bank of America analyst Vivek Arya raised the firm’s 2026 global semiconductor revenue target from $1 trillion to $1.3 trillion. That is a $300 billion jump in just four months — huge for an industry this large. Arya’s argument is simple: AI and data-center demand are now driving most of the upside, especially in compute, networking, and memory. (finance.yahoo.com) ### Is BofA out on its own? Not really — but it is at the aggressive end. IDC published a fresh forecast on April 29 that also blew past the old “near $1 trillion” framing. IDC now sees 2026 semiconductor revenue reaching $1.29 trillion, up 52.8% from $842.8 billion in 2025. So the (finance.yahoo.com)than expected. (idc.com) ### Why is memory suddenly the center? Because AI servers eat memory in a way normal servers never did. Training and inference systems need huge amounts of DRAM, and the premium version is HBM — high-bandwidth memory stacked close to the accelerator so data can move insanely fast. IDC sa(idc.com)oming one of the main ways AI spending shows up in revenue. (idc.com) ### Why does packaging matter so much? Because the hard part is no longer only making a leading-edge chip. The hard part is stitching together compute dies, HBM stacks, and interconnects into one working AI package. That is where advanced packaging — especially TSMC’s CoWoS family — beco(idc.com)el is getting wider, but it is still the tunnel. (trendforce.com) ### Are those bottlenecks actually real? Yes. Micron said its entire calendar 2026 HBM supply has already been committed on price and volume, including HBM4. That is about as direct as supply-tightness gets. On the packaging side, industry trackers still describe CoWoS capac(trendforce.com)ply are still scarce. (earningscall.ai) ### Does everyone agree on the size? No — and that is important. WSTS, the industry trade group, is still more conservative, forecasting 2026 semiconductor revenue around $975 billion. That means the spread between mainstream industry forecasts and the new AI-heavy forecasts is now enormous. Basically, the fight is no longer over whether chips grow in 2026. The fight is over how much AI pulls forward spending from later years. (wsts.org) ### Who benefits if BofA is right? The obvious winners are Nvidia and Broadcom on the compute side, plus AMD and Marvell as AI-system spending broadens. But the quieter winners are the picks-and-shovels companies — Applied Materials, Lam Research, Cadence, and Synopsys — because bigger, denser, more expensive chips need more tools, more process st(wsts.org)es the whole stack. (finance.yahoo.com) ### Bottom line The real story is not “chips are booming.” It is that AI has turned semiconductors from a cyclical market into an infrastructure buildout story. BofA’s $1.3 trillion call may still prove too high. But the direction is clear — demand has outrun the old ceiling, and now memory and packaging are deciding how fast the industry can actually scale.
