The first trading day of 2026 left no room for doubt. While many investors were still recovering from the champagne, chip stocks surged worldwide and set new records.
Taiwan Semiconductor and Nanya Technology reached all-time highs in Taipei, while Samsung and SK Hynix did the same in Seoul. The immediate triggers were optimistic expectations around AI demand, stronger-than-expected chip export figures from Korea, and PMI data pointing to renewed industrial growth in South Korea and Taiwan. The semiconductor supercycle continues unabated.
The term “supercycle” deserves explanation, as it runs counter to the traditional classification of chipmakers as cyclical stocks. Historically, semiconductor companies moved in line with the economic cycle: demand rises in good times and collapses during recessions.
That demand was primarily driven by the PC cycle and later the mobile phone cycle. Inventory cycles were notorious, profit swings spectacular. Suppliers to this capital-intensive sector were, if anything, even more cyclical. These companies could lose money for years, only to post spectacular profits in a single year. Investors learned to buy these stocks on the dip and sell at the peak. That playbook was simple and successful for decades.
Chips are the new steel
But chips are no longer discretionary consumer products. They form the critical infrastructure of the global economy. That is precisely the transition we are witnessing today. Artificial intelligence, cloud computing, the electrification of transport, industrial automation, and the digitalization of virtually every sector, including defense, have transformed semiconductors from a cyclical product into a structural growth core. The chip is the new steel, the new oil, indispensable to economic progress.
The figures from New Year’s Day underscore this phenomenon. SK Hynix reached an all-time high for the first time since the late nineteen nineties. The company had been trapped for decades in brutal boom-bust cycles, but has definitively broken out of this pattern. High Bandwidth Memory, the specialized memory chips that power AI servers. Without HBM there is no ChatGPT, no Gemini, no Claude. Samsung and SK Hynix are extending their HBM3E contracts at significantly higher prices, while new customers are paying an even higher rate. This is due to structural scarcity in a market where supply and demand are fundamentally out of balance.
Three factors behind the supercycle
Why could this supercycle last much longer? Three factors stand out.
First, we are still at the very early stages of the AI adoption curve. Current demand is largely driven by hyperscalers training their models: Microsoft, Google, Amazon, and Meta are expanding their computing power exponentially. But training is only the first phase. The next wave, inference, the actual deployment of AI models in production for hundreds of millions of users, will be even larger. Every chatbot query, every AI-generated image, every smart search result requires processing power.
After that comes edge AI: smartphones running AI locally, autonomous vehicles making real-time decisions, industrial robots operating without a cloud connection. Each phase requires different chips, but all of them primarily require more chips. And after edge AI, the next horizon awaits: AI agents that autonomously perform tasks and communicate with each other. The demand curve is not flattening anytime soon; it is steepening.
Second, capacity expansion is inherently slow and extremely costly. A state-of-the-art chip fab costs 30 to 50 billion dollar and requires four to five years to build. The bottleneck lies with ASML, the Dutch company that is the only one in the world producing the extreme ultraviolet lithography machines needed for the most advanced chips. Each machine costs more than 350 million euro, weighs 150 tons, and is produced in limited numbers. TSMC, Samsung, and Intel all face waiting lists.
Even with record investments, production capacity is growing more slowly than demand. It takes years for new fabs to become operational, and by the time they are, demand has already increased further. This structural mismatch supports prices and margins for years to come. Overcapacity, the traditional death blow for chip cycles, is not on the horizon for now.
Third, geopolitics are permanently reshaping the playing field. The United States has drastically restricted Chinese access to advanced chips and chipmaking equipment. China is investing hundreds of billions in its own production capacity to break that dependency, with mixed success but undiminished urgency. Europe is trying to build strategic autonomy and is attracting Intel to Germany. Japan is rediscovering its semiconductor ambitions with the Rapidus project.
This fragments the global market and creates parallel demand curves. Where a single globalized value chain once sufficed, multiple economic blocs are now building their own infrastructure. Duplication of capacity means more demand for equipment, more demand for raw materials, more demand for talent, and higher prices across the board.
Do not treat semiconductors as cyclical stocks any longer, but as a structural core of the portfolio. Volatility will remain, quarterly earnings and inventory levels will always cause fluctuations, but the underlying trend points in only one direction. The digital economy rests on silicon, and demand for that silicon will only increase in the coming years. The semiconductor supercycle is not a hype and not a bubble. It is the new reality of the technology market, and 2026 promises to be the next chapter.
Han Dieperink is chief investment officer at Auréus Vermogensbeheer. Earlier in his career, he was chief investment officer at Rabobank and Schretlen & Co.