In May 2025, Richard Gordon, Vice President at market research firm Futurum Group, presented a conclusion in a report analyzing the semiconductor industry's long-term growth rate. The industry's compound annual growth rate (CAGR), which reached double digits in the 1990s, had declined as market size expanded, converging to roughly 6% over the most recent decade. Extrapolating from that 6% figure, he projected that annual sales would reach $1 trillion around 2032.
At the time, this view was close to industry consensus. Ajit Manocha, President of SEMI (Semiconductor Equipment and Materials International), stated at the February 2024 Industry Strategy Symposium that "$1 trillion is around 2030," noting that reaching that figure would require 109 new fabs (semiconductor manufacturing plants) to be built by 2026. Back in 2000, Wilfred Corrigan, CEO of LSI Logic, had predicted "$1 trillion within 10 years"—a forecast that missed completely when the market contracted following the dot-com bubble collapse of 2001. Since then, industry insiders had learned to approach growth forecasts with caution.
The semiconductor market moves in cycles of three to four years. During the 2019 memory downturn and the 2023 inventory correction phase, annual sales fell 10–20% year-over-year. The market has historically expanded gradually over the long term through repeated cycles of growth and contraction. This was the framework that had explained the industry for half a century.
Forty Years of Statistical Records Rewritten in Half a Year
The numbers now vividly demonstrate that this framework has broken down.
On August 6, 2026, the Semiconductor Industry Association (SIA) announced that global semiconductor sales for Q2 (April–June) reached $403.3 billion, a 35.1% increase from Q1 2026's $298.5 billion. Sales for June alone came to $134.5 billion, up 123.6% year-over-year and 9.7% month-over-month. This monthly data is compiled by the World Semiconductor Trade Statistics (WSTS) organization and calculated using a three-month moving average.
This growth rate has repeatedly broken records dating back to when WSTS began compiling statistics in 1986. According to an analysis by Semiconductor Intelligence, the 25% quarter-over-quarter growth in Q1 2026 was the highest in the more than 40-year history of WSTS statistics, surpassing the previous record of 20% set in Q2 2009 (during the sharp recovery following the Lehman Shock). Year-over-year, Q1 2026's 79.2% increase far exceeded the previous record of 60% set in Q1 2010.
| Metric | Previous Record | 2026 Result |
|---|---|---|
| Highest quarter-over-quarter growth rate | 20% (Q2 2009) | 35.1% (Q2 2026) |
| Highest year-over-year growth rate | 60% (Q1 2010) | 123.6% (June 2026) |
| Highest annual growth rate | 42% (1995) | 90% (2026 forecast) |
| Projected timing for reaching $1 trillion annually | Around 2032 (as forecast in 2025) | 2026 (expected actual result) |
On an annual basis, WSTS forecasts growth from $795.6 billion in 2025 (up 26.2% year-over-year) to $1,511.0 billion in 2026 (up 90% year-over-year). By 2027, the market is expected to reach approximately $1.9 trillion, a further increase of 27%.
Memory Is Driving Overall Growth in an "Anomalous Structure"
Looking at the structure of this growth, one segment stands out dramatically. According to WSTS's spring 2026 forecast, annual sales of memory products are expected to surge from $230 billion in 2025 to $803.9 billion in 2026—a year-over-year increase of roughly 250%. Against the overall market growth rate of 90%, memory alone is expanding at 2.8 times that pace.
Driving this rapid expansion in memory is demand for high-bandwidth memory (HBM), which is installed in accelerators used for AI training and inference. AI servers built on NVIDIA GPUs require large volumes of HBM; a single AI server rack contains more than 4,500 packaged chips, with semiconductors accounting for over 95% of its value. A joint report by SIA and Deloitte (June 2026) forecasts that annual semiconductor sales into AI-related data centers will grow roughly tenfold, from approximately $120 billion in 2022 to $1.2 trillion by 2028.
Logic products are also expected to grow 37% in 2026, with microprocessors up 20% and analog up 10%—segments outside of memory are expanding steadily as well. Still, the fact that the center of gravity for growth is so heavily skewed toward memory marks a qualitative departure from past semiconductor cycles.
$886.7 Billion in Hyperscaler Capital Expenditure
Demand-side figures also suggest that this growth is being supported by real demand rather than speculative expectations. According to TrendForce's August 2026 estimates, combined 2026 capital expenditure (CapEx) for nine major companies—Google, Amazon, Meta, Microsoft, Oracle, ByteDance, Tencent, Alibaba, and Baidu—will exceed $886.7 billion, up roughly 90% year-over-year from approximately $467.0 billion in 2025. Of this, the five North American hyperscalers account for roughly 90%.
Breaking down the investment, spending on GPUs and AI accelerators accounts for the largest share at approximately $310 billion, followed by server and rack infrastructure at $130 billion, data center buildings at $90 billion, and power and cooling at $80 billion. In 2027, combined CapEx for the nine companies is projected to reach approximately $1.3 trillion, a further increase of nearly 50%.
John Neuffer, President and CEO of SIA, stated: "Global chip sales are expected to exceed $1.5 trillion in 2026, and Q2 sales significantly exceeded Q1. Strong sales in the Americas, Asia Pacific, and China continue to drive market growth."
The "Geography of AI Demand" by Region
Looking at year-over-year growth rates by region in June 2026, the geographic concentration of AI demand comes into sharp relief.
| Region | June 2026 YoY | June 2026 MoM | 2025 Full-Year YoY |
|---|---|---|---|
| Americas | 160.9% | 9.6% | 31.4% |
| Asia Pacific/Other | 124.4% | 9.8% | 45.4% |
| China | 112.8% | 10.4% | 17.9% |
| Europe | 75.2% | 7.7% | 6.7% |
| Japan | 39.0% | 8.6% | -4.3% |
The Americas' 160.9% growth rate reflects the concentration of AI-related semiconductor demand and cloud infrastructure investment in North America. WSTS's forecast also projects the Americas to grow 112% for the full year 2026, the highest of any region.
Japan posted the lowest year-over-year growth rate among all regions at 39%, but given that the full-year 2025 figure was a 4.3% decline, this can be read as a turn toward recovery. Europe also accelerated from 6.7% growth for full-year 2025 to 75.2% growth in June 2026, confirming that this growth is a broad-based expansion in demand rather than one confined to specific regions.
Three Uncertainties Remaining Beyond 2027
WSTS forecasts the 2027 market at approximately $1.9 trillion, up 27% year-over-year, and does not anticipate a slowdown in growth. However, at least three uncertainties remain regarding whether this growth trajectory can be sustained.
First is the durability of memory pricing. Roughly half of 2026's growth stems from the memory segment, particularly demand for and price increases in HBM. The memory market has historically experienced large swings in supply-demand balance, and it always carries the risk that prices will collapse once supply catches up with demand. During the DRAM price decline of 2018–2019, memory sales fell approximately 40% from their peak—a precedent worth noting.
Second is the constraint of power infrastructure. The construction boom in AI data centers is increasingly running up against the physical limits of power supply capacity. The SIA-Deloitte report forecasts cumulative investment of $4 trillion in AI data center construction from 2023 through 2030, but whether this investment proceeds as planned depends on grid reinforcement and securing generation capacity in each region.
Third is the issue of "normalizing" the growth rate. The 90% growth in 2026 is built on top of the 2025 base of $795.6 billion. Even the 27% growth forecast for 2027 far exceeds the semiconductor industry's historical CAGR of 6%. Determining whether this level represents a "new normal" or a singular event produced by a temporary concentration of AI investment will require at least two to three years of accumulated data.
The semiconductor industry has reached the $1 trillion market that it once called a "decade-away future" in just two years. This fact alone clearly shows that AI is fundamentally transforming the structure of semiconductor demand. The next question is whether this structural shift represents a cyclical boom—or a rewriting of the industry's growth model itself.