Global semiconductor sales reached a record $120.6 billion in May 2026, according to new data released Monday by the Semiconductor Industry Association, as AI infrastructure demand continued to pull more of the chip market into the same upcycle.
The May figure was up 9.2% from April and 104.1% from May 2025, when sales were $59.1 billion. SIA said the monthly total was the highest it has ever recorded and marked the 15th consecutive month of month-to-month growth. The numbers are compiled by World Semiconductor Trade Statistics and reported as a three-month moving average, which makes the jump harder to dismiss as a one-month spike.
The strongest point in the report is not just the headline total. Sales rose across every major region, with China up 10.7% from April, Asia Pacific and all other regions up 9.2%, the Americas up 8.6%, Europe up 7.3%, and Japan up 6.4%. On a year-over-year basis, the Americas were up 132.2%, Asia Pacific and all other regions rose 118.9%, China climbed 88.8%, Europe increased 60.7%, and Japan grew 23.8%.
Why the May chip number matters
For readers tracking AI, the useful signal is that semiconductor growth is no longer confined to a few high-profile accelerator suppliers. AI data centers need GPUs and custom accelerators, but they also need high-bandwidth memory, DRAM, NAND, networking chips, controllers, power components, transceivers, CPUs, DPUs, and a long tail of analog and foundational semiconductors.
That wider stack is why monthly chip-sales data has become a proxy for the physical buildout behind AI. A model-training cluster or inference cloud is not just a row of premium GPUs. It is a dense system of compute, memory bandwidth, storage, networking, power delivery, cooling controls, and management hardware. When demand rises broadly across regions, it suggests buyers are not only ordering headline accelerators but building the rest of the infrastructure needed to put them to work.
SIA’s June report with Deloitte made that point more directly. The study estimated that semiconductors account for more than 95% of the content value of a leading AI server rack and more than half of the total capital expenditure required to build and operate an AI data center. It also projected annual revenue from semiconductors used in AI data centers could reach $1.2 trillion by 2028, a nearly tenfold increase over four years.
The regional spread makes the cycle harder to read
The May data also complicates the simple version of the AI chip story. If growth were concentrated only in one region or one supplier class, the market would be easier to explain as a narrow accelerator boom. Instead, the report shows strong growth in the Americas, China, and Asia Pacific, plus continued gains in Europe and Japan.
That does not mean every chipmaker benefits equally. Memory suppliers, networking vendors, foundries, packaging providers, equipment makers, and custom-silicon partners all sit in different parts of the cycle. Some benefit from urgent AI buildouts; others may face timing mismatches, capacity constraints, or pricing pressure as buyers chase specific parts of the stack. The same demand that lifts high-bandwidth memory can strain ordinary DRAM supply. The same rush to build AI data centers can create bottlenecks in networking, power components, substrates, and advanced packaging.
That is why the May report matters more as a capacity signal than as a blanket verdict on every semiconductor stock or supplier. A record monthly total shows that the industry is moving at unusual speed, but it does not erase the operational questions around who has enough advanced capacity, who can qualify parts quickly, and who can turn orders into shipped systems without hitting energy, packaging, or memory limits.
What to watch next
The next useful test is whether the growth rate keeps broadening or starts to expose sharper constraints. SIA’s June 5 release for April sales endorsed a WSTS forecast that global semiconductor sales would grow 90% to $1.5 trillion in 2026 and exceed $1.9 trillion in 2027. May’s record keeps the industry on that aggressive path, but the details will matter: memory pricing, AI accelerator availability, regional export controls, power and data-center delays, and the ability of foundries and packaging lines to absorb demand.
For AI companies, the semiconductor market is becoming a delivery schedule. Model capability, cloud pricing, enterprise AI deployment, and consumer-device features increasingly depend on whether the chip supply chain can provide the right mix of compute, memory, and networking hardware at scale. For chipmakers, May’s record is a sign of extraordinary demand. For everyone buying AI infrastructure, it is also a reminder that the AI boom is now being measured in wafers, racks, memory bandwidth, and regional capacity as much as in model benchmarks.
