BofA Forecasts Semiconductor TAM to Double to $3.2 Trillion by 2030
nashnova research
Bank of America projects the global semiconductor market will nearly double from $1.7 trillion in 2026 to $3.2 trillion by 2030, driven by memory demand, data-center buildout, and an auto/industrial recovery — but the forecast hinges on whether AI infrastructure spending keeps its pace.
The chip industry took 50 years to reach $1 trillion — what justifies doubling in five?
BofA analyst Vivek Arya's team sees global semiconductor TAM (total addressable market — the theoretical ceiling if every potential buyer spends) rising from $1.7 trillion to $3.2 trillion in under five years.
Three engines firing at once: memory-chip demand expansion, data-center buildout acceleration, and a cyclical recovery in auto and industrial.
This means → the growth story does not rest on a single AI bet; multiple downstream markets are pulling at the same time — a thicker foundation, but also more moving parts that can stall.
"Will AI spending slow down?" — what does BofA actually see?
The market's biggest worry is an AI-infrastructure spending slowdown. BofA says it sees no deceleration across four dimensions: customer orders, long-term agreements, capacity commitments, and chip pricing.
The numbers: DRAM and NAND prices are flat week-on-week; Nvidia's B200 GPU rental sits at $5.72 per hour, rising over the past two months and only under 10% below the March peak of roughly $6.10.
AMD echoed the view: no customer-order slowdown detected. In plain terms = two leading chipmakers are saying the same thing — orders haven't cooled, prices haven't cracked, capacity keeps getting signed.
Arya adds: 2027 capacity is essentially fully booked or contracted; 2028 looks similarly tight, with accelerating demand for CPU/XPU pairings and optical interconnects (a new way to move data between chips using light signals instead of electrical wires).
Where is the money flowing hardest?
Memory leads all segments: from $937 billion to $1.8 trillion, nearly doubling.
Servers jump from $359 billion to $848 billion, more than doubling; core semiconductors rise from $739 billion to $1.35 trillion.
By contrast, PCs ($55.8B → $68.7B) and smartphones ($71.1B → $77.8B) grow modestly. This reflects a decisive shift: the growth battlefield has moved from consumer electronics to data centers and AI infrastructure.
Equipment spending is doubling too — what does that tell us?
Wafer-fabrication equipment spending is projected to rise from $155.9 billion in 2026 to $359.8 billion by 2030 — more than doubling.
This means → equipment makers' order books swell before chipmakers' revenue does — to expand capacity, you buy the machines first.
In plain terms = the "picks and shovels" business is scaling just as fast as the chips themselves, and capex is accelerating even faster than revenue.
Which names does BofA highlight?
Top picks in compute: Nvidia and AMD. Networking: Marvell. Analog: Analog Devices (ADI) and ON Semiconductor (ON).
If market momentum returns, Micron, Lam Research, Applied Materials, and Intel are positioned to lead a rebound.
This means → BofA's stock-picking logic splits into two layers: near-term resilience via compute and networking leaders, upside leverage via memory and equipment names.
What is the biggest "if" behind this forecast?
BofA itself names the preconditions: whether AI infrastructure investment sustains its pace, and whether memory and GPU pricing can hold up against expanding supply.
In plain terms = the doubling story assumes money keeps flowing and prices don't crack. If the AI investment cycle peaks early, or capacity ramps crush pricing, $3.2 trillion becomes a ceiling, not a floor.
This reflects a broader pattern: even the most bullish sell-side forecasts embed an "if" in the fine print — read the number, but read the assumption behind it too.
市场有风险，内容仅供研究参考，不构成投资建议。