Warning signs for semiconductor market in AI boom

Semiconductor market researchers are at odds on the impact of AI on the supply of chips in 2026. While all see strong growth, some are predicting that the market has fundamentally changed, and that the market will top one trillion US dollars by the end of this year.

That is not possible as there is not enough capacity and there are underlying weaknesses, according to leading researcher Malcolm Penn of Future Horizons, who has just released his forecasts for the year.

“These are eye watering numbers and, with all due respect, this is not going to happen,” he said.

He sees a growth of around 12% for 2026, less than the 22% in 2025. The growth could be as high as 18%, but this is very different from other forecasts which are as high as 40% on the back of the boom in demand for AI chips.

“Any party is a good party, and enjoy it while it lasts, but prepare for the hangover,” he said. He points to overcapacity in legacy process nodes in China and weak global economies.

“We believe that the extrapolations of the forecasts are very dangerous, and we believe we are highly likely to see a correction, it’s inevitable, it’s only the timing that is unclear. When there is a correction, the growth goes negative so its could go from -8% to -30% depending on how fast it happens. The risk of negative growth this year is certainly on the cards.”

However the spending on building AI datacentres has driven other market analysis. The datacentre demand has been driving up the demand for graphics processor units (GPUs) from Nvidia in particular but also from AMD, as well as high bandwidth memory (HBM) and also driving up the prices.

“What is happening now is something categorically different,” said Ben Bajarin, chair of market researcher Creative Strategies in the US. “The semiconductor industry will never be the same. The shape of the category has been changed forever.”

“The artificial intelligence infrastructure buildout represents the largest total addressable market expansion the semiconductor industry has ever experienced—a giga cycle that dwarfs previous periods of growth in both absolute dollar terms and in the breadth of its impact across every segment of the value chain.

“The numbers tell a story of unprecedented scale. Global semiconductor revenues will climb from roughly $650 billion in 2024 to more than $1 trillion by decade’s end, with several forecasts now pulling the trillion-dollar mark forward into 2028-2029.

“This expansion is not driven by a single product category or geographic market. This is a fundamental restructuring of the industry’s trajectory, driven by infrastructure requirements that touch every category of semiconductor technology simultaneously,” he said.

This analysis is backed up by predictions from market research firms Omdia which is predicting the semiconductor market to top $1 trillion, not in 2030, but this year.

“Computing & Data Storage will lead all segments in semiconductor revenue growth rising 41.4% YoY in 2026 to exceed $500bn, due to high demand in data centre servers and other memory–intensive applications, as well as higher memory IC pricing. Notebook PC growth is being fuelled by AI-capable adoption and a large-scale enterprise refresh cycle. Collectively, the top four hyperscalers are expected to spend approximately $500bn on capital expenditures this year, with further growth expected,” said the researchers at Omdia. This will drive the market over $1 trillion this year, it says.

Industry veteran Penn at Future Horizons agrees that AI is important but not in the way that the industry expects at the moment, pointing to the market bubbles in DRAMs, dotcoms, and internet over the last decades.

“AI is a potential game changer but not as we know it. You are looking at seven to ten years before the first usable product, and that tells us the first AI killer app will be in 2030,” he said. “But the investment is in the trillions of dollars and the return on investment is in the millions. That’s not sustainable.”

He also points to the GPUs becoming obsolete quickly, so the investment in infrastructure is worthless before it has depreciated. “These chips age as gracefully as dead fish, in four years they are practically worthless,” he said.

Despite the extremely bullish forecast, Omdia is also hedging its bets in the forecast.

“Semiconductor revenue growth in 2026 is being driven by highly concentrated, AI-related demand, rather than broad-based consumer behaviour or industrial production trends that have historically influenced the market. Without the contributions of memory and logic ICs, overall semiconductor revenue growth would fall from 30.7% to only 8%, highlighting the nature of the demand driving recent market surges,” said Myson Robles-Bruce, Senior Principal Analyst at Omdia.

The forecast this week from TSMC as the leading foundry also supports a more conservative view. “We forecast the foundry 2.0 industry (which includes semiconductors and advanced packaging) to grow 14% year over year in 2026, supported by robust AI related demand,” said CC Wei, CEO of TSMC.

The company grew 35% last year on the back of the growth in demand for AI chips from NVIDIA and AMD but Wei is cautious.

“Entering 2026, we understand there are uncertainties and risk from the potential impact of tariff policies and rising component prices, especially in consumer related and price sensitive end market segments,” he said. “As such, we will be prudent in our business planning, while focusing on the fundamentals of our business to further strengthen our competitive position.”

However this is not stopping plans for a fab to build devices on the latest 1.4nm leading edge process in Taichung, Taiwan, set to be in production in 2028. This comes as the 1.6nm A16 process comes into production later this year alongside the 2nm N2 and N2P high performance processes.

“A16 is best suitable for specific high performance computing (HPC) products with complex signal routes and dense power delivery networks,” he said, “Volume production is on track for second half 2026 and we believe N2, N2P, A16 and its derivatives will propel our N2 family to be another large and long-lasting node for TSMC.”

This follows equipment going into its second fab in Arizona, with the third fab underway and a fourth in the planning stages, all for leading edge AI chips.

All this capacity is what worries Penn. “Capex still worries me. The spend is still stubbornly high. It looks like a capacity bubble,” he said. “If the AI bubble bursts the excess capacity at the leading edge is really going to hurt.”