The always insightful work of Malcom Penn at Future Horizons was contained in his March 2026 update, reproduced with permission below.
Executive Summary
January’s WSTS Report saw a further $4.0 billion upward revision to last month’s reported numbers, bringing the total up to US$795 billion, up 26.1 percent vs. 2025. This 0.6 percent annualised increase was solely attributable to ICs, up 39.9 percent year-on-year, more specifically to Memory, up 29.9 percent, and Logic, up 38.8 percent, all in turn driven by the red-hot AI-datacenter explosion.
In sharp contrast, Analog and Micro were up just 8.7 and 7.9 percent respectively, with Opto and Discretes even lower at 4.7 percent and 3.2 percent.
Whilst IC value-growth was eye-watering, unit growth was only single digit, re-enforcing the fact 2025’s growth was ASP, not demand, driven.
At this stage in the cycle, as we enter 2026, we would all be well-advised to remember the two golden rules of semiconductor ASPs. First, Moore’s second law, “The long-term IC ASP value is a dollar” and second, the historical industry observation, “The long-term IC ASP growth is zero.” Unless, or course, “It’s different this time!”
Market Outlook
Q3 and Q4’s strong double-digit growth rates were some of the strongest on record, lifting the full-year growth to 26.1 percent, bringing the total market value to just shy of US$ 800 billion. And with no sign yet of this growth momentum slowing, we can expect to see humdinger first quarter as well.
With still no sign yet of any slowdown in this growth, a strong first quarter would blow our plus 12 percent forecast, with an upside of 18 percent, clear out of the water, with an upside now north of 40 percent, potentially adding around US$300 billion to 2025’s blockbuster chip sales.
But … if the AI market tanks, overnight the current inexhaustible end-market demand for AI-hyperscaler investment would plummet, triggering a collapse in high-performance processor and HBM sales, ricocheting into other support products, from power discretes to microcontrollers and analog, with the rest of the chip market overwhelmed by the tsunami, unable to push back against the collateral damage and fallout.
On the global economic front, the world outlook is currently being stress-tested by a convergence of shocks, namely: the Middle East war; the emergence of AI as a disruptive technology; soured loans starting to pop up in the booming private-credit industry; a softening US job market; and stubbornly high inflation.
Each shock alone might be manageable, but together they are creating fragilities in global markets that no single policy lever can fix, making it different and more difficult than the rout sparked Donald Trump’s 2025 tariff rollout.
Do not expect the economy to provide a strong foundation for 2026 chip market growth.
There is no upside to the current economic outlook, other than hope that the multiple downside risks do not materialise.
A downturn and correction in chip market growth is inevitable, only the trigger and timing uncertain. If the trigger is a correction in AI demand, the downturn will happen in 2026. If that demand stays strong throughout 2026, then the crash will come early in 2027, once the additional memory capacity comes on stream.
The only forecast certainty right now is “Everyone’s chip market forecast for 2026 is wrong!”
Our overall message for 2026 is clear; enjoy the party if you can but proceed with extreme caution and do not be distracted by 2025’s headline dollar growth. Growth is ASP not demand driven and that can reverse just a quickly as it came.
To paraphrase the wise words of SK-hynix Chairman Chey Tae-Won, “2026’s US$300 billion sales growth could just as easily turn into a US$300 billion decline.”
