Ajit Manocha, President and CEO of SEMI, a trade association for the semiconductor manufacturing supply chain, declared that the industry is in a "super cycle that only goes up." Having been in the industry for four decades, he noted that the typical boom and bust cycles of the past 20 to 30 years have been replaced by this continuous growth. Manocha projects semiconductor industry revenue to reach between $1.8 trillion and $2 trillion, attributing this decade's growth to AI and the next's to quantum computing. His optimistic outlook is based on what SEMI members, who supply equipment and materials to fabs globally, see in their order books, indicating robust demand.
A key driver of this super cycle is the aggressive investment in memory capacity. Manocha highlighted that every memory company is actively investing, with demand for memory, for both storage and compute, picking up significantly. Micron Technology's fiscal Q3 results underscore this trend, with revenue reaching $41.46 billion, a 345.72% year-over-year increase, and a GAAP gross margin of 84.6%. Micron's CEO, Sanjay Mehrotra, stated that DRAM and NAND industry demand continues to significantly exceed supply, a situation expected to persist beyond calendar 2027.
However, the aggressive capacity expansion raises concerns for some analysts. Historically, memory cycles end when contract prices spike, leading producers to build new fabs. These fabs deliver wafers two to three years later, often coinciding with a cooling of initial demand, causing an oversupply. Micron is already increasing its fiscal 2027 quarterly capital expenditure above its fiscal Q4 pace of roughly $10 billion, with its Idaho ID1 fab slated for first wafer output in mid-calendar 2027. While Strategic Customer Agreements, covering about 20% of Micron's DRAM volume with take-or-pay commitments, mitigate some risk, they may not prevent non-HBM DRAM pricing from declining.
The current chip cycle is supported by several structural changes. Hyperscaler capital expenditure from companies like Amazon, Microsoft, Google, and Meta, committing hundreds of billions annually to AI infrastructure, provides stable demand. High-bandwidth memory (HBM) requirements create supply constraints, keeping prices elevated as AI models demand more HBM per unit of compute. Additionally, government-backed fab investments and geographic diversification of manufacturing, seen in the US CHIPS Act and similar legislation in Europe and Japan, add resilience by spreading capacity. Global semiconductor revenues hit $368 billion in Q2 2026, marking a 35% quarter-over-quarter increase and a 104% year-over-year gain, with full-year 2026 growth forecasted between 90% and 112%, potentially reaching $1.6 trillion.
Despite the optimism, risks remain, including supply chain vulnerabilities in rare earth materials and advanced lithography equipment, and regulatory uncertainty from export controls. Should hyperscaler capital expenditure moderate due to slower AI adoption or an economic slowdown, demand could shift faster than committed capacity. The surge in AI chip demand has also attracted new entrants beyond Nvidia, including AMD, Intel, and hyperscalers developing custom silicon. While Manocha believes the cycle structure has changed with a higher floor and smaller downturns, the aggressive capacity additions for memory raise questions about the long-term sustainability of current margins, impacting companies like Micron compared to equipment suppliers like Applied Materials or AI compute providers like Nvidia.
