JP Morgan: The AI and semiconductor upcycle will extend to 2027, optimistic about the performance of Asian tech stocks next year

JP Morgan's latest research believes that despite ongoing market concerns about an artificial intelligence bubble, the AI-driven semiconductor upcycle is far from peaking, and its duration will exceed typical cyclical patterns and extend to 2027, providing strong support for the performance of Asian tech stocks in 2026.

According to news from the Wind Trading Desk, the bank raised its expectations in a report released on November 11, 2025, forecasting global semiconductor revenue to grow by 18% and 11% in 2026 and 2027, respectively. This judgment is primarily based on the steep early adoption curve of generative AI, the continued strong capital expenditure willingness of leading cloud service providers (CSPs), and the conservative capacity expansion strategies in key supply chain segments.

JP Morgan stated that the Asian tech market in 2026 will present a "delicate" balance, where investors will repeatedly worry about the peak of the cycle while continuously seeing upward adjustments in earnings per share (EPS) driven by AI infrastructure construction and price increases in certain components. The bank believes that the strong trend of earnings upgrades will ultimately drive Asian tech stocks higher in the coming quarters.

This viewpoint directly challenges the growing narrative in the market that "the AI bubble is about to burst," providing investors with a new perspective when evaluating the tech sector. JP Morgan believes that the unique nature of the current cycle makes it impossible to fully apply historical experiences and that more attention should be paid to its inherent structural drivers.

Market Divergence and Earnings Upgrades Coexist

JP Morgan stated in the report that although measured by classic indicators such as cycle duration and earnings growth, the current upcycle is nearing its late stage, this will not change its upward trend. The bank expects that in the first half of 2026, benefiting from the strong earnings growth of AI-leading companies, Asian tech stocks still have room for growth.

In the past three months, the EPS upgrade momentum of Asian tech stocks has significantly accelerated, driven by memory chips and other bulk tech commodities. The report pointed out that strong AI demand is "crowding out" the supply of the entire tech industry, leading to supply shortages in multiple areas, from advanced packaging (CoWoS), cutting-edge wafer foundry, and high bandwidth memory (HBM) to ordinary testing (OSAT), multi-layer ceramic capacitors (MLCC), and standard DRAM/NAND. As suppliers have increased capacity at a pace far slower than normal levels during this cycle, the price increases triggered by shortages will further drive corporate earnings upgrades.

Four Reasons Supporting the Extended Cycle

JP Morgan's analysts believe that several unique characteristics of this cycle distinguish it from previous general recoveries driven by global GDP expansion, and are sufficient to extend the upcycle to 2027 First, this is a "K-shaped recovery" cycle. Since mid-2023, the market has shown a divergence between AI and non-AI demand. AI-related demand has performed strongly, while technology demand in other areas is facing an adjustment trend.

Second, the adoption of generative AI is climbing along a steep S-curve, with its development trajectory more similar to the early stages of smartphones or public clouds. The report points out that smartphones took about 7 years, and public clouds took about 9 years before their annual growth rates fell below 20%. In contrast, generative AI is currently only in its fourth year and is expected to maintain a year-on-year growth rate of 50%-60% in 2026, indicating significant growth potential.

Third, leading cloud service providers have the ability and willingness to continue investing. Although there are concerns about the free cash flow (FCF) of smaller CSPs, JP Morgan's analysis shows that the capital expenditures of the top six CSPs are expected to grow strongly by 67% in 2025 and still increase by about 32% in 2026. Even assuming that free cash flow remains flat in 2027, these companies still have the capacity to support a further 13% increase in capital expenditures, indicating that their financial condition is sufficient to support the continued expansion of AI infrastructure.

Finally, the supply response in key areas is slow. In this cycle, capital expenditures in the semiconductor sector have been quite conservative, with supply highly concentrated among a few manufacturers such as Taiwan Semiconductor and SK Hynix. The report predicts that Taiwan Semiconductor's capital expenditures will only grow by 16% in 2026, while the DRAM industry's capital expenditures will only grow by 11%. This means that supply shortages in critical areas such as 3nm processes, CoWoS, and HBM will continue until 2026, prolonging the chip shortage.

Based on the above judgments, JP Morgan recommends that investors adopt a "barbell" investment strategy in 2026. One end should be allocated to leading AI enablers; the other end should be allocated to companies that benefit from price increases and margin expansion, with significant profit revision potential. The report maintains an overweight rating on Taiwan Semiconductor and lists it as a preferred stock. **