Global semiconductor industry could grow 12% (18%) in 2026: Malcolm Penn

Future Horizons, UK recently organized the Industry Update conference.

Malcolm Penn, Founder and CEO stated that for the executive snapshot, in context, April 2025 US tariffs disrupted trade norms. However, worst case has been avoided. It matters, as tariffs were weaponized, and are still high. There is uncertainty-elevated efficiency, with losses rising. As for major economies, US growth is down, while inflation is rising. China has structural fragility. EU and emerging markets lack fiscal and financial support.

The bottom line -- economy has been resilient, but fragile. Confidence and stability are fastest path to stronger growth. The chip market needs a strong economy for sustainable growth.

Are unit shipments at the cusp of recovery? Nov. 2025 was 7.8 b/w, which was just above trend line, but 5% below 8.2b/w maxed out peak. 2H-2025 rebound was influenced by tariffs, memory shortage, and inventory builds. We had $108 billion worth of Covid-19 investments. It is still unclear whether Covid-19 boom excess inventory balance is fully depleted. We can't claim real market recovery until unit growth resumes.

For capacity, capex spend is still stubbornly high. Q1-2023 capex peak was at 19% of semiconductor sales, and three quarters after June 2022 crash, and that is normal. Q4-2025 capex is still not yet below the long-term trend line, and that is dangerous. Bleeding edge investments in logic and memory are held justifiable, but the AI bubble can be dangerous. Also, >7nm  capex overspend is not driven either by China and the Chips Acts.

Capex overspend culprit
China is the main capex overspend culprit. Capex share accelerated after June 2022 market collapse. 2025 spend was 4x the 2010-2015 level, along with decoupling, tariffs, and export controls driven. 2025 capex spend was 37% of the worldwide total. Taiwan was 1.5x, Korea 2x, USA 4.7x, Japan 5.5x, and Europe 17x. We have 3x justifiable market share level. Focus has been more on non-leading edge, but we are seeing progress on 5/7nm, especially from Huawei and SMIC.

ASPs are showing strong recovery since June 2022 crash. IC ASPs peaked at $1.35 in May 2022, and then collapsed in June 2022 to $1.11. Recovery started in May 2023, climbing to $2.19 in Nov. 2025. Logic was original ASP driver, and was now joined by memory and micro. Strong AI demand is the main reason, besides TSMC's logic SoC price hikes. Looking at IC ASPs by product sector, Analog has been dipping, and flat as a pancake. Micro, memory, and logic have all shown growth.

Long-term IC ASP trend is $1, or, Moore's second law. Current ASP boom is nothing to write home about. Rising ASP disruptions don't last forever. Moore's second law will kick in. Competition and capacity will increase, unless, there is a monopoly, such as Intel x86, and TSMC and Nvidia.

Jan. 2026 status shows market fundamentals are opaque. All four warning lights are blinking green, red, and amber. We have fragile global economy and geopolitical uncertainty due to Trumpeconomics. There is longer-term oversupply pressure on mature nodes, thanks to China and Chips Acts. 

The economy's GDP was 3.2% in 2025, and will be 3.1% in 2026. In units, recovery could be in prospect, but it is still below 2022 peak. ASPs are all trending up, except analog. Capacity is in excess, along with China threat. Capex spend is too high.

We are back in blue skies following brief dark side encounter. 3/12 gives near-term view on direction of growth. This gives an indication of which direction 12/12 will take. 3/12 curve is back, growing, following 12-month 2024 plateau. We are seeing a re-run of 2002-03, 2013-14, and 2020-21 bounce backs.

Looking at annualized semiconductor growth rates by sector, opto and discretes growth turned positive in Q2. Memory is back on growth trend. HBM capacity is reallocation driven. Logic is powering ahead, while micro / analog are 'steady as you go'.

Global semiconductor industry grows 22% in 2025
Jan. 2025 forecast for global semiconductor industry was 19% bull, and 11% bear, with outlook forecast at 15% , to $723.652 billion. Monthly annualized growth rates were trending steadily down. Excess capacity plagued the industry throughout 2025. China potential to flood the world with mature node technology was real. We had increasingly closed China market, along with embargoes and rise in Made in China.

What actually happened? The global semiconductor industry grew 22% in 2025, to $769.217. We had an unprecedented strong Q3 growth that added six points to the final number. Discretes, opto, and analog markets started to show early signs of recovery. China potential to flood the world with mature node technology did not materialize. There was no Q4 seasonal slowdown with AI-demand driven. Red flags were that 2025 growth rate was IC ASP-driven. Unit growth has yet to recover.

Global semiconductor industry to grow 12%, or, 18%  in 2026
For 2026, Future Horizons has forecast the global semiconductor industry to grow 12% to $813.104 billion. Bullish estimate for 2026 is at 18 percent to $863.121 billion. The next supercycle cannot happen before earliest 2029, and will be capex dependent. 

2026 bear outlook potential is -- if it crashes, it crashes. It is a sober reality. A correction is inevitable, only, the timing is unclear. Either the AI demand will tank, or the infrastructure will fail to keep pace with demand. If, and when it crashes, the growth would go negative. About -12% could be optimistic. There could be negative growth next year -- -8% minimum.

Key takeaways
Semiconductor 8.0 has shown relentless creative destruction. Four Asian, and six US companies made up the top 10 semiconductor companies in 2025. TSMC, Nvidia, Samsung, SK Hynix, and Broadcom made up top five. Bottom five were Intel, Micron, Qualcomm, AMD, and MediaTek. There is a natural and efficient way for companies and economies to evolve and grow.

As for CES 2026 overview, the show was decidedly underwhelming. There were robots, living aids, wearables, consumer durables that dominated the show. A common thread was intelligent sensing and processing to make them smarter. That's what Moore's Law has been doing for past eight decades -- more bang, same buck.

There is still no plan B for China-Taiwan advanced logic. TSMC will be building advanced 1.4nm fab at Taichung fab in 2028. Also, 1.6nm chips would be made in Kaohsiung next year. TSMC plans to start construction in Q4. China risk is still there. We are still doing nothing to take that away. 

Let's also look at the current AI reality. Human brain generates 20W, while hyperscaler generates 1GW to 20GW. 1GW or one nuclear reactor is sufficient to power 700k homes / 1 million people. The footprint is 1.5 times bigger than New York's Central Park. It is more than a giant stomach gorging indiscriminately on data, than brain. Investments are in trillion $, and RoI in million $. That's unsustainable!

Missy Cummings, Prof. Robotics & AI, George Mason University, USA, has remarked, "AI systems do not understand, imagine or reason under uncertainty." Current AI reality is that it makes errors and hallucinates. Malcolm Penn gave his own example of this on ChatGPT.

We have AI hype on steroids, and that keeps the bubble from probably popping. Nvidia said we may all end up making clothes for robots. Elon Musk said AI and robotics will surpass human doctor's capabilities by 2028.

The real AI market is yet to materialize! The over-hyped AI promise is yet to deliver a concrete, tangible, everyday product. Don’t hold your breath for a game-changer from Apple, given their recent capitulation to Google’s Gemini. A clue: It won’t be based on $30k plus Nvidia GPUs. The first killer ap could well be an AI camera-less smartphone.

AI could also pose threat to humanity. How will it actually destroy us? Stock market could probably destroy itself with AI. Overbuilt track and fiber was much useful!

Don't get drunk on AI hype or euphoria! It is a potential gamechanger, but not as we know it. History tells us that the first AI killer app could happen in ~2030 earliest.