Global Semiconductor Sales Hit $146.8 Billion in July 2026, Up 135% Year-on-Year

The Semiconductor Industry Association released its July 2026 monthly data on September 4, 2026, showing global semiconductor sales reached $146.8 billion — a 6.4% month-to-month increase from June's $137.9 billion and a staggering 135.1% jump compared to July 2025's $62.5 billion. It marks the 17th consecutive month of month-to-month growth, a streak the industry has never extended this far. Sales have already surpassed the highest-ever annual total with five months remaining in the year.

"The global semiconductor market grew on a month-to-month basis for the 17th consecutive month in July, and the industry has already reached its highest-ever annual global sales total just seven months into the year," said John Neuffer, SIA president and CEO, in a statement published by the trade group.

Americas Leads the Charge, but No Region Lags

The geographic spread tells its own story. Year-on-year growth in July was strongest in the Americas at 171.3%, reflecting the combined effect of domestic CHIPS Act investment and surging AI infrastructure orders. Asia Pacific posted 134.9% YoY growth, China 123.6%, Europe 85.2%, and Japan 50.8%.

Month-to-month the picture is more evenly distributed: Japan led at 9.2%, followed by the Americas at 8.7%, Asia Pacific at 6.8%, Europe at 5.2%, and China at 2.9%.

The figures are compiled by the World Semiconductor Trade Statistics organization as a three-month moving average — the methodology smooths month-to-month volatility and gives a cleaner signal on underlying demand. Applied to the current trajectory, the numbers suggest the market is nowhere near a plateau.

The AI Infrastructure Buildout Is Rewriting the Semiconductor Story

Behind the headline figure is a structural shift in what is driving demand. The growth isn't evenly distributed across chip categories. Memory — specifically high-bandwidth memory used alongside AI accelerators — is the primary engine of the current cycle.

According to data released by the Semiconductor Industry Association and analyzed by industry watchers, memory semiconductor sales grew approximately 305% year-on-year during the first half of 2026, substantially outpacing every other product segment. Logic chips also posted strong gains at roughly 45% year-on-year, but the scale of the memory surge is in a different category entirely.

SK hynix generated revenue of KRW 79.3 trillion in Q2 2026, up 51% sequentially and 257% year-over-year. The company attributed the performance to high-value products including HBM, AI server DRAM, and enterprise solid-state storage. Micron Technology reported fiscal third-quarter revenue of $41.46 billion, compared with $23.86 billion in the prior quarter — an increase of approximately 74%. Kioxia posted quarterly revenue of ¥1.77 trillion, up roughly 76% from the prior quarter, with SSD and storage revenue nearly doubling to ¥1.17 trillion.

The demand pattern has broadened beyond HBM. AI servers require large quantities of conventional DRAM alongside the high-bandwidth modules, and the growth of inference workloads is pushing demand for high-capacity NAND flash and enterprise storage. Smartphone and PC demand remains stable rather than declining — a normalization that contrasts with the 2022-2023 downturn.

"The global chip sales are expected to exceed $1.5 trillion in 2026," Neuffer said, a figure now considered conservative. Revised WSTS calculations incorporating the stronger-than-expected first half now point toward approximately $1.655 trillion for the full year, equivalent to roughly 108% annual growth.

Hardware Architecture Is Evolving to Match AI Demand

The memory surge is inseparable from changes in how AI accelerators are physically packaged. High Bandwidth Memory sits alongside processors in systems that train and run large models, and the bandwidth demands of modern AI chips require new interconnection approaches.

BGA — Ball Grid Array — packaging uses a dense array of gold solder balls to mount a chip to a circuit board. This format allows hundreds or thousands of electrical connections in a compact area, making it the standard for processors and memory modules that handle the bandwidth requirements of AI workloads. Intel, AMD, and Nvidia all use BGA-style packaging in their current AI-oriented silicon.

SK hynix has already begun mass shipments of HBM4, and Micron is shipping HBM4 in volume for its lead customer while developing the enhanced HBM4E variant for 2027 production. Every generation of HBM increases the memory bandwidth available to AI accelerators, directly affecting how fast large language models can be trained and served.

The continued expansion of GPU and AI accelerator shipments has ripple effects throughout the supply chain. Additional AI accelerator deployments require supporting memory, networking, power management, advanced packaging, substrates, and semiconductor manufacturing capacity. That's why the current cycle is creating growth well beyond the companies that directly manufacture GPUs.

Excluding a handful of major AI and memory suppliers, other semiconductor vendors among the top tier collectively grew revenue by a more moderate 9% sequentially in Q2 — a reminder that the boom, while historically notable, is concentrated in a specific slice of the industry.

Where the Market Goes From Here

The direction remains positive, though the most explosive sequential growth rates are beginning to normalize. Micron guided toward approximately $50 billion in fiscal Q4 revenue, compared with $41.46 billion in the prior quarter. Kioxia expects revenue of approximately ¥2.39 trillion in the September quarter, representing around 35% sequential growth — still extremely strong, but below its approximately 76% growth in the previous quarter.

The memory companies themselves are signaling that the growth curve is flattening at a high level rather than turning down. Supply agreements with hyperscale data center operators are longer-term in structure than the commodity cycles the memory industry has historically experienced, providing some floor to pricing even as the pace of sequential increases moderates.

WSTS projections for 2027 suggest the industry could reach approximately $2.1 trillion, representing another roughly 29% increase from the revised 2026 forecast. That projection relies on the WSTS Spring forecast assumptions for subsequent quarters and isn't a completely new scenario — but it indicates that even with growth normalizing, the baseline for the industry has moved permanently higher.

Previous semiconductor cycles were largely associated with a single major new computing platform: personal computers in the 1980s and 1990s, smartphones in the 2010s. AI infrastructure is the next such platform — but with a key difference. AI systems require massive semiconductor content across multiple chip categories simultaneously: GPUs, custom ASICs, HBM, DRAM, NAND, networking silicon, power semiconductors, and advanced packaging. That breadth is why the current expansion is unusually wide across the technology stack, even as it remains most visible in the memory and AI accelerator segments.

For semiconductor manufacturers, equipment suppliers, and the broader electronics supply chain, the question isn't whether the AI cycle will sustain — it's how quickly capacity can be added to meet demand that shows no sign of easing.

Related: Battery Tech powers the mobile edge devices running inference workloads, while Cloud & Edge Computing infrastructure hosts the data centers driving chip demand.