McKinsey Raises 2030 Semiconductor Market Forecast to $2.3 Trillion, with Price Gains Replacing Volume Growth as Key Driver
nashnova research
McKinsey raised its 2030 global semiconductor market forecast from $1.6 trillion to $2.3 trillion, a 44% jump; wafer shipment growth stays flat — nearly all the new value comes from higher prices and a richer product mix, not more chips sold.
How much bigger is the new forecast, and where does the extra money come from?
McKinsey now projects a 19% CAGR for 2025–2030, up from 13% previously — far above the 9% historical average over the past decade.
Global wafer shipment growth stays at roughly 7% per year, unchanged from the prior forecast.
This means → the extra $700 billion in market size comes not from selling more chips, but from each chip fetching a higher price and the mix shifting toward premium products.
In plain terms = the pie is bigger not because there are more slices, but because each slice costs more.
Who is spending the money that pushes chip prices up?
The five hyperscale cloud providers plan combined capex of roughly $800 billion in 2026, rising to $1 trillion in 2027, overwhelmingly directed at AI infrastructure.
Global semiconductor sales growth in H1 2026 hit the highest since the mid-1980s; July set a 40-year record, after which industry revenue crossed $1 trillion for the first time.
This reflects a shift: the AI arms race has turned chips from an abundantly supplied industrial commodity into a scarce strategic resource — money alone no longer guarantees immediate supply.
Why are advanced-node chips getting more expensive, not cheaper?
The server and data-center chip market is projected to grow from $330 billion in 2025 to $1.2 trillion by 2030 — a 29% CAGR.
Historically, wafer prices fell after launch: 28 nm wafers dropped roughly 10% per year, 150 nm about 25%. Today, 3 nm and 5 nm wafer ASPs are rising at 2%+ per year.
This means → the old rule — "new node arrives, older nodes get cheaper" — is broken. AI demand is too strong and advanced capacity too scarce for prices to come down.
By 2030, price increases, volume growth, and product-mix upgrades are expected to add roughly $710 billion in advanced-node chip value.
How much have memory prices risen, and why so sharply?
Since early 2023, DRAM prices have risen roughly sixfold, with ASPs back to their 2011 peak.
In plain terms = AI training and inference devour high-capacity, high-bandwidth memory — especially HBM (high-bandwidth memory, stacked DRAM designed specifically for AI accelerators) — and supply simply cannot keep up, pushing prices to historic extremes.
McKinsey expects memory supply may begin catching up with demand around 2027, after which prices should ease — but likely remain above historical norms.
Even so, memory is still expected to contribute roughly $560 billion in incremental market value.
How much do legacy-node chips get?
Legacy-node chips — FPGAs (field-programmable gate arrays, chips that can be reconfigured for different AI tasks), interconnects, and power electronics — also benefit from AI infrastructure buildout.
But while shipment volumes rise, ASPs are expected to keep falling — contributing only about $100 billion in incremental value by 2030.
This means → legacy nodes follow the old "more volume, lower price" playbook, while advanced nodes and memory follow the new "flat volume, rising price" logic — two completely different growth stories.
Can this forecast actually materialize, and what should we watch?
Two variables matter most: whether AI infrastructure spending holds at current intensity, and whether capacity expansion breaks the current supply-demand balance.
If hyperscalers cut capex, or if capacity ramps faster than expected, the pricing thesis weakens.
This reflects McKinsey's core judgment: the semiconductor market is shifting from a "sell volume" era to a "sell price" era — but how long that lasts depends on the AI investment cycle itself.
市场有风险，内容仅供研究参考，不构成投资建议。