Global semiconductor sales smashed through $1 trillion in cumulative revenue by August 2026, hitting $159.7 billion in that month alone—a 144.3% year-over-year jump that nobody in the industry saw coming this fast.
The Semiconductor Industry Association (SIA) announced the milestone on October 5, capping 18 consecutive months of growth. August sales rose 8% from July's $147.9 billion, with the Americas leading regional gains at 180.9% year-over-year. WSTS data, compiled as a three-month moving average, shows no sign of the streak ending before 2027.
The $1 trillion threshold matters because it's not just a number—it's a signal that AI infrastructure spending has permanently reset the industry's baseline. SIA president Michael Robbins said annual sales are now projected to exceed $1.6 trillion this year, a figure that would have seemed aggressive even 18 months ago. The growth isn't concentrated in one region or product category either—it's broad-based, spanning memory, logic, and analog chips across every major market.
The Memory Squeeze Behind the Numbers
The revenue surge is being driven disproportionately by memory chips. DRAM and NAND prices have climbed sharply through 2026, with Nomura estimating Q1 increases of 90% and 60% respectively. AI processors like Nvidia's H100 and Huawei's Ascend series demand ever-larger pools of high-bandwidth memory, and manufacturers can't build fabs fast enough to keep up.
Deloitte's 2026 semiconductor outlook noted that an estimated 1.05 trillion chips were sold in 2025 at an average selling price of $0.74 per chip. This year, that ASP is climbing as AI servers consume premium components at volume. The mismatch between memory supply and AI-driven demand has turned the chip cycle from a predictable boom-bust pattern into a sustained seller's market.
The shortage isn't just about capacity—it's about the right kind of capacity. TSMC's CoWoS advanced packaging lines, which stack memory and logic chips together for AI accelerators, remain in undersupply through Q3 2026. The company's capital budget for 2026 sits at $52–56 billion, with CoWoS capacity slated to double to 260,000 wafers by the end of 2028. TrendForce reported that TSMC is also developing an "EMIB-like" approach as it races to diversify beyond CoWoS.
Memory module makers are feeling the squeeze too. Innodisk and Apacer both posted September 2026 revenue growth, citing AI demand as the primary driver. Advantech, an edge AI hardware supplier, topped $1 billion in third-quarter revenue. These numbers confirm that the AI hardware boom isn't limited to GPU makers—it's rippling through every layer of the semiconductor supply chain.
Singapore's $800M Bet on the Chip War's Next Phase
Singapore has quietly become one of the world's most important semiconductor hubs, producing roughly one in ten chips globally and one-fifth of all manufacturing equipment. On October 6, the city-state launched SG Semiconductor, a national identity program designed to keep that position from eroding.
The initiative, announced by Trade and Industry Minister Dr. Tan See Leng at the Innovate Together 2026 forum, comes with fresh commitments. Applied Materials and A*STAR will expand their joint lab for AI and advanced computing packaging. GlobalFoundries is deepening silicon photonics R&D on 300mm wafers. KLA is establishing a new process control partnership to improve yields.
"Singapore's strength in semiconductor innovation has been built through decades of sustained investment and close partnership across public research, universities and industry," said Beh Kian Teik, Chief Executive of A*STAR. "SG Semiconductor brings this collective endeavour under a national identity."
Singapore's Economic Development Board Managing Director Jermaine Loy framed the branding effort as a practical tool for companies evaluating where to invest next. "As semiconductor technologies become more complex, companies are increasingly seeking trusted locations where innovation, manufacturing and talent come together," Loy said.
The S$800 million commitment runs through 2030 under Singapore's RIE2030 strategy. It covers seven technology areas, including advanced packaging, power electronics, and silicon photonics—all critical for AI infrastructure. A new S$60 million national R&D centre for power electronics is also set to open by April 2026, focusing on partnerships between academia and industry.
The timing is deliberate. As the US and China tighten export controls and diversify supply chains, Singapore is positioning itself as the "trusted location" that multinationals need. The semiconductor desk has tracked Singapore's rise for years, but the SG Semiconductor launch marks the most aggressive attempt yet to formalize that role.
China's Tool Stockpile and Washington's Next Move
While demand booms, the supply side is roiling with tension. A late-September report from the Centre for Technology & Statecraft found that Chinese fabs had amassed an estimated 343 immersion DUV lithography systems by early 2026. About 270 of those were ASML's Twinscan NXT:1980i, machines that can be adapted to produce 7-nanometre logic chips and advanced memory.
Chinese plants spent over $13 billion acquiring roughly 90 NXT:1980i machines in 2024 and another 89 in 2025, according to the report co-authored by former US export control official Nicholas Brown and former White House technology adviser Saif Khan. China accounted for 41% of ASML's net system sales by shipment destination in 2024.
Those numbers have revived calls in Washington for a complete ban on immersion DUV exports to China. The bipartisan Match Act, introduced in April, seeks China-wide restrictions on sales and servicing for chipmaking equipment China can't easily make at home. Senators renewed pressure for its passage in September, and Khan publicly endorsed the bill earlier this year.
Sales to China have since moderated—dropping to 19% of ASML's total in Q1 2026 and 14% in Q2—but the inventory buildup means any future supply cutoff would have limited near-term impact. The CTS report projected that if China continues importing about 90 NXT:1980i systems annually, its DUV fleet could support up to 434 million AI chips equivalent to Nvidia's H100 by 2035. Halting new imports in 2027 would cap that ceiling at 153 million units, it said.
The Netherlands began requiring export licenses for ASML's NXT:2000i and more advanced systems in September 2023, extending the requirement to the NXT:1980i a year later. Shipments of that model can still be approved case by case, leaving what critics call a "major loophole." SCMP reported that the CTS authors recommended a complete China-wide ban on all immersion DUV equipment, alongside restrictions on replacement parts and servicing.
ASML itself has denied selling its most advanced EUV tools to China. In June, the Dutch company said it had never shipped an EUV lithography machine to China, after a Bloomberg report raised questions. But the DUV stockpile is real, and it's growing.
The semiconductor industry has entered uncharted territory. Global revenue is growing at a pace that matches the most optimistic forecasts, but the geographic concentration of supply—and the political pressure to keep it that way—means the next decade will be defined as much by trade policy as by transistor density. The $1 trillion milestone isn't just a finish line. It's the starting gun for a new phase of the chip war, one where the battlegrounds are memory supply chains, packaging capacity, and export control language. The companies and countries that manage those tensions best will write the next chapter of the AI age.