Musk's TeraFab orders chip tools, ASML's China share falls

Samantha: Welcome to The State of Tech, The European Edition, Monday November nine, 2026. I'm Samantha Lawrence.

Bob: And I'm Bob Russell. Today: Elon Musk's chip factory TeraFab starts ordering machines, TSMC eyes a new kind of memory for Nvidia, export controls redraw who buys chipmaking tools, a new push to fix quantum computing's error problem, and two things to look up yourself: NASA's X-70 Fractal test plane and Roche's AXELIOS sequencing platform. Let's start with Musk.

Elon Musk's TeraFab chip factory reportedly starts ordering equipment from Japanese suppliers for Texas.

Samantha: Elon Musk's chip factory plan, TeraFab, is moving off the drawing board. According to reports, the project is placing substantial orders with Japanese suppliers of chipmaking equipment. The factory is planned for Texas, and the goal is to supply Tesla and SpaceX with their own computing chips. The ambition is enormous. By 2032, TeraFab wants to turn out computing chips equivalent to one terawatt every year, and reach a million silicon wafers a month. A wafer is the thin disc of silicon that chips are cut from.

Bob: Those numbers are the story. Chip factories normally take years to plan, build and fill with machines, and the reporting describes TeraFab as a direct challenge to those traditional timelines. Placing equipment orders is the first concrete step that turns a concept into a real factory. The purpose is clear too: chips for AI and robotics, made for Musk's own companies. That adds a major new player to the advanced chip market, a market where a small number of manufacturers have long set the pace.

Samantha: It's worth being clear about what is and isn't known. The orders are reported, not formally detailed, and 2032 is still six years away. A million wafers a month is a target, not capacity that exists today. What the orders do show is that the plan now has real money behind it, flowing to equipment makers in Japan. For those suppliers, a customer with ambitions of this size could mean busy order books for a long time, if the project keeps moving at this pace.

Bob: And that brings in Europe. The continent is home to key chip equipment suppliers, and TeraFab's global shopping list could shift competition among them. So far the reports name Japanese suppliers, not European ones. When a project of this scale chooses where to buy, the suppliers left out feel it as well. There's a supply chain question too. If TeraFab delivers, a lot of new chip capacity sits in Texas, but it is designed first to serve Tesla and SpaceX, not outside customers.

Samantha: The bottom line: TeraFab is no longer just an idea floated by Musk. With equipment orders reportedly going out, the question shifts from whether it gets built to how fast. For European listeners, two things are worth watching: which suppliers win the next rounds of orders, and whether a factory built to feed two companies' AI and robotics plans changes how everyone else competes for the same scarce chipmaking machines.

TSMC drafts plans to make advanced stacked memory chips at Nvidia's request.

Bob: TSMC, the world's largest contract chipmaker, may be moving into the memory business. According to reports, the Taiwanese company has started drafting plans to produce an advanced type of memory chip, known as 3D DRAM. And it's doing this at the request of its largest customer, Nvidia. This kind of memory is the fast working memory that processors use while they crunch data. The 3D version stacks the memory in layers, so more of it fits into the same space.

Samantha: The background is a shortage. AI systems need huge amounts of fast memory sitting right next to their processors, and the AI industry is currently dealing with severe memory chip shortages. Nvidia sells the chips that power much of the AI boom, so a memory squeeze hits its business directly. By asking its main manufacturer to step in, Nvidia is trying to widen supply. For TSMC, it's a chance to secure a larger slice of the AI hardware supply chain than processors alone.

Bob: This would be a real strategic shift. TSMC built its business on manufacturing processors designed by other companies, and memory has been a separate market with its own specialists. The reports describe plans being drafted, so nothing is final, and the sources give no timeline. But if it goes ahead, existing memory makers would face a powerful new competitor, and Nvidia would gain more influence over one of the tightest bottlenecks in its whole supply chain.

Samantha: For Europe, it cuts both ways. European AI developers could eventually gain access to more advanced memory options, which matters when shortages slow down projects. At the same time, the move intensifies global competition in the semiconductor supply chain, and that competition plays out largely outside Europe. Key takeaway: memory has become as strategic as the processors themselves, and the biggest players are now rearranging who makes what.

Samantha: Quick interruption. If you listen to The State of Tech regularly, hit that like button and subscribe, that way you'll never miss an episode. Okay, moving on.

Chip equipment makers sell a smaller share to China as export controls and AI demand bite.

Bob: The world's biggest chipmaking equipment companies are selling a smaller share of their machines to China. ASML from the Netherlands, Tokyo Electron from Japan, and the American firms Lam Research and KLA all reported a decline in the share of sales going to China in the third quarter. These companies build the machines that factories need to produce chips, so their sales show where the industry is investing next.

Samantha: Two forces are behind the shift. US export controls keep restricting sales of advanced equipment to China. At the same time, surging demand for AI chips is driving more equipment purchases in South Korea and Taiwan. So the picture isn't only about China buying less. Other regions are buying more, and that changes the balance on its own.

Bob: The third piece is inside China itself. Chinese domestic equipment makers are expanding their market share at home. That's the long-term tension with export controls: they limit what China can buy from abroad, while local suppliers gain room to grow. Notable is that the decline was uniform across all four companies, spread over three different countries.

Samantha: For Europe, ASML is the obvious name. The Dutch company is directly affected by both the US export controls and the changing market. Its share of sales to China is falling, while AI-driven buying in South Korea and Taiwan is rising. That shifts where its customers are, and ties its business more closely to the AI chip build-out in Asia.

Bob: The bottom line: chip equipment sales have become a map of geopolitics. Where the machines go tells you where future chip capacity will be. Right now that map points away from China and toward the AI hubs in South Korea and Taiwan, while Chinese firms try to fill the gap at home. And ASML is one of the clearest examples of how decisions made in Washington show up in a European company's numbers.

IBM, AMD and Oak Ridge launch a platform to tackle quantum computing's error problem.

Samantha: Quantum computers have a fragility problem, and a new collaboration is aiming squarely at it. US computing giant IBM, chipmaker AMD and the Quantum Science Center at Oak Ridge National Laboratory have launched a platform for quantum error correction. The basic units of a quantum computer, called qubits, are extremely sensitive. Tiny disturbances, which researchers call noise, scramble the information they hold. Error correction is the set of techniques that protects that information.

Bob: The interesting part is the combination. The platform links Eye-Bee-Em's Nighthawk quantum computer, which has a hundred and twenty qubits, to Oak Ridge's Umbra supercomputer, which runs on Ay-Em-Dee's graphics processors. So a quantum machine and a classical supercomputer are being integrated into one setup. The aim is to use both kinds of computing together to tackle the noise problem, rather than relying on the quantum machine alone.

Samantha: The stated goal is to explore designs and ways of working that can scale error correction. That word, scale, is the key. Protecting a handful of qubits in a lab is one thing. Protecting the far larger numbers needed for useful work is the hurdle the whole field faces. Without reliable error correction, quantum computers stay experimental machines rather than practical tools.

Bob: This isn't an AI story, but it does use the same kind of hardware that powers the AI boom. The graphics processors from Ay-Em-Dee are doing classical number-crunching next to a quantum machine. That hints at how quantum computing could arrive in practice: not as a standalone replacement for today's computers, but as one part of a larger system inside a supercomputing centre.

Samantha: For Europe, this is a signal of where quantum research is heading. Progress on error correction is a precondition for quantum computing to become commercially useful, and that could speed up development for European companies and researchers working on quantum technology as well. The set-up itself, a national lab teaming up with two big tech firms, also shows how much quantum work now leans on partnerships between public research and industry.

NASA unveils the X-70 Fractal, a test plane with rapidly reconfigurable flight controls.

Bob: To close out, two things you can look up today, and the first one flies. NASA has announced a new experimental aircraft, the X-70 Fractal. It's a subsonic X-plane, so it flies slower than the speed of sound, and it's built as a testbed: a flying laboratory that researchers can adapt for different experiments, rather than a plane designed for one fixed job.

Samantha: What sets it apart is its flight controls. The X-70 Fractal will have controls that can be reconfigured rapidly, so researchers can change how the aircraft steers and responds from one test to the next. According to NASA's announcement, that lets them explore ways of manoeuvring that have never been demonstrated in piloted flight. In other words, handling that hasn't been tried before with a person in the cockpit.

Bob: Here's what you can do with it. Search for X-70 Fractal and read NASA's announcement. It takes a few minutes, it's free, and it works on any phone or laptop. It gives a clear look at how new aircraft ideas get tested long before they reach a production line. One limitation: the announcement covers the programme, so there's no first flight to watch yet, and the sources give no date for one.

Samantha: The bigger picture is that X-planes are NASA's experimental aircraft, built to gather data for future designs. The X-70 programme is meant to push aerospace engineering forward and collect crucial data on new aircraft designs and control systems. That research could also inform future European work on aircraft design and innovative flight controls.

Bob: For students, aviation fans or anyone curious about engineering, it's a quick and free read. Key takeaway: this isn't a gadget you can buy, but it's a rare chance to follow an experimental aircraft from its very first announcement onward, and to keep an eye out as NASA shares more about the plane and its tests.

Roche brings its AXELIOS DNA sequencing platform to researchers in Australia and South Korea.

Samantha: The second one is for the lab crowd: AXELIOS, from Swiss pharmaceutical company Roche. It's a platform for next-generation sequencing, the technology that reads DNA at large scale for genetic research. Roche launched it globally in June, and it uses a method the company calls sequencing by expansion. Now Roche is expanding it into the Asia Pacific region, following early adoption in South Korea.

Bob: The practical hook is a series of events called AXELIOS Days, in Australia and South Korea, where researchers and laboratory leaders can see the technology up close. Roche presents the platform as offering more flexibility and performance across different kinds of genomic research. This is aimed at professionals, not home users. It's laboratory equipment, and the sources give no price and no details on how to sign up for the events.

Samantha: So what can a listener actually do today? If you work in a research or hospital lab, or you study genetics, look up AXELIOS and Roche's sequencing by expansion method, and compare it with the sequencing tools you already know. If you're based in Australia or South Korea, check whether an AXELIOS Day is planned near you. For everyone else, it's a quick read on where DNA reading technology is heading.

Bob: There's a European angle as well. Roche is a European company, and taking its sequencing platform into a region with growing genomics capabilities strengthens Europe's role in global biotech innovation. The bottom line: a European name is competing in a key part of biotech, and the roll-out in Asia Pacific is the next test of whether labs there take it up.

Samantha: Today we covered: Musk's TeraFab ordering chipmaking machines, TSMC eyeing memory for Nvidia, export controls reshaping chip equipment sales, a new push on quantum error correction, and two things to look up yourself: NASA's X-70 Fractal and Roche's AXELIOS platform.

Bob: Want to know more or react? Visit stateoftech.eu or email us at info@doorzetters.net.

Bob: State of Tech, the tech world in 15 minutes.