Chip stocks entered Tuesday, Sept. 15, trying to recover from a sharp sell-off tied to renewed concerns that artificial-intelligence development could slow.
Nvidia (NVDA), Advanced Micro Devices (AMD), and other semiconductor stocks fell after AI-industry leaders raised concerns about the pace of frontier-model development.
The rebound was uneven around midday on Sept. 15. The PHLX Semiconductor Index was up about 0.3%, with AMD gaining roughly 1.85% to $502.55, Nvidia rising about 0.41% to $211.83, and Marvell Technology (MRVL) adding about 1.63% to $222.38.
Applied Materials (AMAT) fell about 1.36% to $418.45, and Lam Research (LRCX) dropped roughly 2.21% to $267.44.
Bank of America’s latest semiconductor forecast gives investors a much longer time horizon.
BofA semiconductor analyst Vivek Arya and his team raised their estimate for global semiconductor sales through 2030 in a Sept. 14 report shared with TheStreet.
BofA’s preferred names span compute with Nvidia and AMD, networking with Marvell, analog chips with Analog Devices (ADI) and onsemi (ON), and equipment with Lam Research and Applied Materials.
The firm rates all seven Buy. Its price objectives include $350 for Nvidia, $620 for AMD, $365 for Marvell, $385 for Lam Research, and $650 for Applied Materials.
BofA raises 2030 chip market estimate by nearly $470 billion
BofA now expects global semiconductor sales to reach $3.2 trillion in 2030, up from its prior estimate of about $2.7 trillion.
The new forecast implies an 18% compound annual growth rate from 2026 through 2030, compared with 14% previously.
Related: After $664 billion backlog, Oracle sends shocking message to staff
The industry would nearly double from the roughly $1.7 trillion in semiconductor sales BofA forecasts for 2026.
Memory chips and servers account for much of the increase.
BofA expects memory sales to reach about $1.85 trillion in 2030, up from $937 billion in 2026.
Server semiconductor sales are projected to increase from about $360 billion to $849 billion over the same period.
PCs and smartphones are moving in the opposite direction in the near term.
BofA expects semiconductor sales into PCs to fall about 9% in 2026 and smartphone semiconductor sales to decline roughly 9%, leaving data centers and memory responsible for a much larger share of industry growth.
AI servers and memory carry most of the growth
BofA expects server-chip sales to grow at about 24% annually through 2030, faster than any major end market in its forecast.
Memory is projected to grow about 19% annually over the same period.
More AI:
- Nvidia just made a move Wall Street wasn’t ready for
- Microsoft just took sides in AI policy fight
- OpenAI just disclosed something genuinely alarming
AMD provides one example of the server demand already reaching chip suppliers.
It reported record second-quarter revenue of $11.5 billion, up 50% from a year earlier. Its data-center revenue also more than doubled to $6.7 billion on demand for EPYC server processors and Instinct GPUs.
BofA’s $620 AMD price objective reflects the potential for further share gains in AI accelerators and server processors.
Execution of AMD’s MI400 rack-scale products and the timing of large AI projects remain risks to that target.
Marvell captures another part of AI infrastructure through custom processors and networking chips.
The company reported record fiscal second-quarter revenue of $2.74 billion, up 37% year over year, with data-center revenue rising 46%.
BofA’s $365 target for Marvell reflects improving visibility for major custom-chip projects and continued demand for networking and connectivity inside AI data centers.
Chip factories could spend $360 billion on equipment by 2030
Higher processor and memory demand also requires manufacturers to add factory capacity and more production steps.
BofA raised its 2026 wafer-fab equipment forecast to $156 billion from $144 billion.
It expects spending to reach about $210 billion in 2027 and roughly $360 billion by 2030, compared with its previous 2030 estimate of around $300 billion.
Wafer-fab equipment includes machines that deposit material onto silicon wafers, etch microscopic structures, and perform other steps needed to manufacture processors and memory chips.
Memory drives much of BofA’s latest equipment upgrade.
Related: Burry says AI leaders have ‘nothing to slow down’
The bank expects memory-equipment spending to increase from about $61 billion in 2026 to $85 billion in 2027, primarily because of additional DRAM investment.
High-bandwidth memory used with AI processors adds more manufacturing work.
HBM uses substantially more wafer capacity than conventional DRAM and requires additional stacking and advanced packaging steps.
BofA expects those changes to keep increasing the amount of manufacturing equipment required per wafer through 2028.
The spending directly affects Applied Materials and Lam Research.
BofA previously estimated that global cloud capital spending could reach $1.18 trillion in 2027, creating more demand for processors and memory and eventually pushing chipmakers to expand factories.
Applied Materials reported record fiscal third-quarter revenue of $9.12 billion, up 25% year over year.
The company also said it expected strong demand in DRAM, leading-edge foundry and logic, and advanced packaging.
BofA’s $650 price objective for Applied Materials assumes semiconductor equipment spending continues expanding through 2026 and 2027.
Lam Research has heavier exposure to the deposition and etch processes used in memory and advanced processors.
It reported record June-quarter revenue of $6.72 billion, up 15.1% sequentially.
BofA’s $385 target is based on Lam’s exposure to memory investment and leading-edge foundry and logic production, as well as the increasing number of etch and deposition steps required to manufacture more complex chips.
Equipment orders and memory prices are the key risks
The semiconductor index remains about 6% lower over the past five trading days, even after the modest Sept. 15 rebound, showing that investors have not fully dismissed concerns about the durability of AI spending.
Demand outside AI also remains uneven.
BofA expects wireless communications semiconductor sales to fall about 8% in 2026 and consumer semiconductor sales to decline about 7%. PC and smartphone demand are also expected to contract.
The bank has yet to see evidence of a broader AI hardware slowdown.
BofA said customer orders, long-term agreements, capacity commitments, and semiconductor pricing remain firm. It described 2027 as largely booked or contracted across compute, networking, and memory suppliers.
Those indicators provide concrete tests for the forecast.
Falling equipment orders would show that chipmakers are becoming less willing to add manufacturing capacity. Weaker memory pricing would point to softer supply-demand conditions. Reduced capacity commitments would suggest that customers no longer need as much future production.
A sustained deterioration in those three areas would be the clearest evidence that the physical buildout behind BofA’s long-term semiconductor forecast is beginning to slow.
Related: BofA’s $1.18T cloud forecast puts 3 chip stocks in focus