Don't be fooled by $ growth for semicon! 2025 will be challenging!!

At the IFS 2025 Update, Malcolm Penn, Founder & CEO, Future Horizons, stated that 2024 global semiconductor market recovery has been flimsy, so far. The 2024 ‘recovery’ was driven by ASP, and not units. Demand is still devilishly weak. 

Future Horizons updated the May 2025 forecast for the global semiconductor industry at +8% (+/- 3%), to $681.109 billion. Bear growth is 5% at $661.940 billion.

Forecast
In Jan. 2025, the global semiconductor industry was forecast to grow 15% (+/- 4% in 2025. The year will have 15% growth for the global semiconductor industry to $723.652 billion. Bear shows 2025 growth at 11% to $699.009 billion.

Monthly annualized growth rates are trending steadily down. Excess capacity will plague the industry throughout 2025. China has real potential to flood the world with mature node technology. We also have an increasingly closed China market, with embargoes and Made In China activities.

May 2025 update suggests it is not quite Armageddon! But, not fact off. Economy GDP growth was downgraded to 2.8% vs. 3.3%. No recovery is in sight for units. They are now 16.9% vs. 5.9%, and below long-term trend. ASPs have weakening growth trend, except that logic has been increasing. Capacity is in excess, and China threat is still real.

Growth is still AI infrastructure driven. ‘Real markets’ are yet to recover. Worsening global GDP outlook could trigger a global recession. There is longer-term oversupply threat on mature market nodes (China Plus Chips Acts).

Data reality is that stormy seas are on the horizon. Death Cross had happened in Dec. 2024. We have stormy weather ahead. Will we see potential (indeterminable) period of froth, or have a full-blown collapse?

8 percent growth in 2025?
Future Horizons has updated May 2025 forecast for global semiconductor industry at +8% (+/- 3%), to $681.109 billion. Bear growth is 5% at $661.940 billion.

Annual growth rates are still strong. However, not so the units, and broader recovery is weak. Excess capacity will plague the industry throughout 2025, going on to 2026. As mentioned, China has the potential to flood the world with mature node technology.

Watch out for a slowdown or hiccup in the AI infrastructure demand! And, don't be fooled by the $ growth! 2025 will be challenging!

2026 outlook
As of now, the outlook for the global semiconductor industry is 1 percent for 2026, at $687.691 billion. This won't improve till 2029.

There are too many variables for anything other than an uwag, but, there will be very low single-digit growth at best. This could easily go negative if AI demand collapses, or slows. Trump 2.0 could also tip the world into recession, keeping it at knife's edge.

There can be no return to real growth before 2027.  The next supercycle will not be before 2029, and that's dependent on China and capex. 

Takeaways
So, how long can China stay at their impressive pace? Chinese wafer foundries currently have capacity expansion plans for 2025.

However, can they solve sub-5nm litho hurdle? Most electronics are still made in China, and there is strong internal OEM need. Unlike the West, OEM-generated value is trickling down to chip firms (eg., Huawei, et al). 

China is still producing 75% of laptops, 70% of TVs, and 60% of mobile phones. Since China is currently blocked from access to the ASML EUV machines, they need to solve three basic litho problems -- optics, overlay , and throughput.

There was status quo prospective impact of tariffs and reshoring. Apple has strong production footprint in China. There 2,700 different parts / 187 suppliers / 28 countries / 74 tiny screws by hand (1mm and up). You can’t reshore that overnight, no matter how much you wanted to.

Technology roadmaps keeps on trucking. EUV lithography issues are there. Bottom side power delivery / top side signal routing are being addressed. Better organization of resources is also required. Unlike frontend, backend does not readily scale. Increasing the volume is quite expensive, and scaling.

Node transitions are challenging, eg., GAA. It is a three-horse race (now ‘2’)? TSMC N2 (GAAFET), Intel 18A (RibbonFET), and Samsung SF2 (MBCFET) are leaders.

Comparing Intel's 18A and TSMC's N2, Intel is likely faster, but TSMC is likely denser. Intel has more technically advanced GAA plus backside power. Yields are lower for now, but it puts them ahead of the learning curve game.

Despite positive $ growth numbers, down cycle still has a long way to go. Proceed with extreme caution!