Forecasting the global semiconductor market? “It’s impossible,” laughs Malcolm Penn. To make the point, the veteran industry forecaster produced a wide range of scenarios for the chip market in 2026.
His bullish scenarios range from an 80.3% rise to $1.4 trillion, a 95.3% rise to $1.55 trillion, and eye-popping 101% increase to $1.6 trillion.
Says Penn: “These scenarios assume the AI data centre growth does not slow down, crash, or burn, that the Middle East war does not trigger a global GDP slowdown or recession, that IC unit growth resumes, and there is no slowdown in average selling price growth in the first half of 2026.”
He added that there would need to be a strong recovery in non-AI markets including computer, smartphone, and automotive, and that DRAM Capex spend does not crash the HBM and non-memory AI markets.
His bear forecast scenarios range from 13.4% growth to $901.8 billion, based on a slowdown in Q2 this year, a 26.8% rise to $1.1 trillion, based on a Q3 slowdown, and an advance of 46.5% to $1.16 trillion, based on a crash in Q4.
These forecasts are based on continued Middle East conflict and a closure of the Hormuz Strait triggering a global economic slowdown, AI data centre growth stalling, leading to a collapse in advanced node logic and HBM demand, no IC unit growth, Capex spend triggering an oversupply in Q1 2027 or even Q4 2026, and IC ASP growth collapsing due to falling AI demand and/or memory over supply.
Penn’s concern is that the fundamentals that dictate the course of the semiconductor industry are not promising.
Notwithstanding the rise in stock markets, global GDP is now forecast at 3.1% down from 3.3%.
“The broader chip industry needs a strong economy to thrive,” he observes. “IC growth is currently fuelled by a speculative ‘Leap Of Faith’ AI FOMO vision.”
IC unit shipments are struggling to recover. “You can’t claim a true market recovery until IC unit growth resumes,” remarks Penn.
He also raises red flags over Capex spend. “It remains stubbornly high,” says Penn, citing China’s Capex spend at 37% of the worldwide total, which is three times its justifiable level given local demand.
China’s spend is 1.5% greater than Taiwan, 2x that of South Korea, 4.7x that of the USA, 5.5x that of Japan, and 17x that of Europe.
Average selling prices have made a strong recovery. Penn notes the current boom “is nothing to write home about, the industry’s been there, done that,” citing memory prices sliding from $20 to cents in the course of 12 months.
“Rising ASP prices do not last forever,” asserts Penn. “Either competition, attracted by the high margins intensifies, or capacity expands causing a glut.”
None of these fundamentals have stopped Penn’s forecasting peers issuing forecasts “on steroids”, leading to a $2.4 trillion market in 2035.
“Sorry,” sighs Penn, “with all due respect to my industry peers, that isn’t going to happen. Extrapolations are dangerous for your health, market booms never last.”
He thinks that either AI demand will tank or infrastructure won’t keep pace with demand and DRAM capacity bloating will trigger oversupply and collapse ASPs.
“A correction is inevitable, only the timing is unclear.”
To add a further douche of cold-water, Penn quotes the CEO of Hynix after analysts’ predictions of a $100 billion profit for the company in 2026: “That’s good news, but it could just as easily be a $100 billion loss.”
