The CHIPS and Science Act, signed in August 2022, set aside $52.7 billion for US semiconductors, including $39 billion in manufacturing incentives run by the Commerce Department’s CHIPS Program Office. Those incentives pay for new fabs, packaging plants and suppliers of wafers, gases and equipment, mostly as direct grants, sometimes with loans. Most of the big awards were finalized in the last months of the Biden administration. The Trump administration kept the programme but changed how it works, most visibly by turning Intel’s grants into a government shareholding.

Co-packaged optics (CPO) moves the part that turns electrical signals into light off the front panel of a network switch and onto the same package as the switch chip. Today, almost every fiber link in a data center runs through a pluggable transceiver: a small module, about the size of a pack of gum, that slots into the front of a switch or a server network card. CPO gets rid of most of those modules.

CoWoS (Chip-on-Wafer-on-Substrate) is TSMC’s method for placing a large processor and several stacks of high bandwidth memory side by side on a shared slab of silicon wiring, then mounting that slab on a package substrate. Almost every leading AI accelerator is built this way: Nvidia’s Blackwell and Rubin GPUs, AMD’s Instinct chips, Google’s TPUs and most custom chips designed with Broadcom or Marvell. For the past three years the number of CoWoS wafers TSMC could process each month has capped how many of those chips could ship, more often than the supply of leading-edge wafers did.

The US can block exports of American chipmaking tools, but it does not make all the tools that matter. The world’s only maker of EUV lithography machines is ASML in the Netherlands. Japan’s Tokyo Electron is a top supplier of the coating, etching and deposition systems found in every advanced fab. Without the Netherlands and Japan on board, US equipment controls on China would leak badly. Both countries did sign on, starting in 2023.

The Entity List is the US Commerce Department’s register of foreign companies, institutes and people that need a special license before they can receive items subject to US export rules. It is run by the Bureau of Industry and Security (BIS) and it has become the main tool Washington uses to slow China’s progress in chips, supercomputing, AI and surveillance technology. Since 2019, BIS has added many hundreds of Chinese entries.

High bandwidth memory (HBM) is DRAM built as a tower. Eight, twelve or sixteen memory dies sit on top of a logic base die, wired together vertically, and the whole stack is placed a few millimetres from a GPU or AI accelerator inside the same package. The reason is bandwidth. One HBM4 stack moves at least 2 terabytes per second, more than three times what a 12-channel DDR5-6400 server socket manages (about 614 GB/s).

Huawei is the company the US has spent the most effort trying to contain. In 2019 it was the world’s largest telecom equipment maker and was about to pass Samsung as the biggest smartphone vendor. Seven years of US measures later it is still huge, still profitable and now building its own chips, operating system and AI accelerators. Both halves of that sentence are true, and the timeline below shows how.

Hybrid bonding joins two chips by pressing their polished faces together so that copper pads on one fuse directly with copper pads on the other. There is no solder, no bump and no gap. The name comes from the two bonds made at once: the insulating surfaces stick to each other, and the embedded copper pads weld into continuous metal. Because nothing sits between the dies, connections can be packed far closer together than solder allows, and stacks get thinner and cooler.

For years US export controls stopped American chips and tools from reaching Chinese military-linked users, while American venture capital could still fund the Chinese startups building rival chips and AI models. The outbound investment rules close that gap. Since January 2, 2025, US persons have been barred from some investments in Chinese semiconductor, quantum and AI companies, and must report others to the Treasury Department. This is a different tool from export controls.

A “3nm” chip has no important feature that measures 3 nanometres. Node names stopped describing physical sizes decades ago and now work as generation labels, each one promising a denser, more efficient chip than the last. This page explains where the names came from, how the big foundries’ labels line up, and which numbers actually tell you something. Last updated: 15 September 2026. Where the names came from Through the 1990s a node name roughly matched the length of a transistor’s gate.

This page tracks the main US rules that limit China’s access to advanced chips, AI accelerators and chipmaking equipment. It starts with the October 2022 package and runs to the guidance Washington issued in 2026. Entries are in date order, and each one says what changed and who felt it. Last updated: 15 September 2026. How the controls work The Bureau of Industry and Security (BIS) sits inside the US Commerce Department and writes the Export Administration Regulations (EAR).

Semiconductor headlines run on a small vocabulary that most readers use interchangeably and almost nobody uses correctly. The difference between a company being added to the Entity List and a company being sanctioned is the difference between a licensing requirement and an asset freeze. The difference between the de minimis rule and the Foreign Direct Product Rule is the difference between a control that stops at the US border and one that reaches into a Dutch or Taiwanese factory.

AI sovereignty is the capacity of a state to develop, deploy and govern artificial intelligence systems without depending on infrastructure, models or rules controlled by another state. The term entered mainstream policy vocabulary around 2023 and has since been applied to national compute programmes, domestic model development, data residency law and regulatory regimes. It is used loosely, and that looseness matters: the phrase bundles together at least four distinct claims with radically different price tags and success rates.

Cadence put up quarterly revenue of $1.58 billion, up 24.2 percent year over year, and raised its annual guidance to a range of $6.26 to $6.34 billion against a consensus near $6.21 billion. The stock added more than four percent after hours. The number that matters is not the 24.2 percent. It is the guide raise, and specifically its timing: it landed on the same session in which ASML fell more than seven percent, Applied Materials more than six, and Lam Research more than seven on a report that a Chinese state-backed manufacturer has begun building deep ultraviolet lithography machines.

A report that a Chinese state-backed company has begun manufacturing deep ultraviolet lithography machines took ASML down more than seven percent, Applied Materials down six and a half, and Lam Research down seven. By the time Korean markets opened, Samsung Electronics was off more than eleven percent, SK Hynix more than twelve, and the KOSPI more than nine. The sequence looks like one story propagating outward from a single headline. It is at least two stories, and the market priced them as if they were the same one.

I’ve defended Micron against its rivals often enough that I don’t reach for the company’s press office as a source of good faith. Legacy chipmakers cry protectionism every time a customer finds a cheaper part, and most of the time the complaint is about margin, not national security. Every instinct I have says discount this warning too. I can’t. Not this time. Apple’s pitch to the Trump administration — sourcing Chinese memory for products sold outside the United States — has been framed as a carve-out, a geographic footnote that leaves the domestic industry untouched.

I generally treat “confidential IPO filing” stories as noise — every ambitious Chinese tech company floats a Hong Kong listing rumor at some point, and most of them either stall or get quietly repriced down before they ever ring a bell. I don’t think MetaX’s filing is noise, and the reason has nothing to do with MetaX’s chips. The Shanghai-based GPU maker confidentially filed for a Hong Kong listing targeting a year-end IPO, part of what reporting describes as a surge of fundraising from rivals in the same window China’s vice premier reportedly warned domestic AI companies that resisting local chips made them traitors to the state.

I want to like the “AI is deflationary” story. It’s clean, it’s optimistic, and every model launch this year has come with a chart showing cost per token falling off a cliff. I’ve written versions of that chart myself. The chart is real. It’s just not the whole curve. Qualcomm telling customers it’s raising prices by double digits, after exhausting its ability to absorb supplier costs, landed in the same news cycle as Anthropic pricing Opus 5 at roughly half of what Fable 5 costs for comparable capability.

Every dollar of AI capital expenditure eventually lands somewhere concrete: a wafer, a switch, a cooling loop, a power purchase agreement, a SaaS seat. Tracking that flow from the hyperscaler earnings call down to the individual equity ticker is the only way to separate durable exposure from sentiment-driven noise. The chain below runs twelve tiers deep rather than the usual eleven, because the packaging layer that sits between memory and networking has become too structurally important to leave folded into a generic equipment bucket.

The case for owning Marvell almost always opens with custom AI chips, and that is the part of the story where Marvell is structurally second. The company and Broadcom are the two names that turn hyperscaler in-house silicon designs into manufacturable chips, the application-specific processors that Amazon, Microsoft, Google and Meta increasingly favor over off-the-shelf GPUs for inference. But Marvell holds roughly 15 percent of that market against Broadcom’s 55 to 60 percent, and Counterpoint expects Broadcom to command about 60 percent of custom AI silicon by 2027.

The technology relationship between the United States and China did not break in a single moment. It eroded through a decade of escalating restrictions, retaliatory measures, investment screenings, and legislative maneuvers — each move accelerating the next. What began as targeted actions against individual companies has become a structural reorganization of the global technology supply chain. This timeline documents that process from its earliest institutional signals through the present. 2015 January — The Obama administration’s Department of Commerce adds CETC (China Electronics Technology Group Corporation), a state-owned defense electronics conglomerate, to the Entity List, citing its role in supplying military electronics.

Something subtle but important is happening in the global AI landscape, and you could almost miss it if you only focus on the usual headlines about chips, models, and hyperscalers. At NVIDIA GTC 2026, Taiwan showed up not just as the world’s manufacturing backbone, but as an increasingly confident architect of full-stack AI systems. A delegation of 16 high-growth startups, organized through Startup Island TAIWAN’s Silicon Valley Hub with backing from the National Development Council, didn’t just exhibit—they embedded themselves into the ecosystem in a way that signals a shift in strategy.