23 juni 2026 · 13:13
ASML, TSMC & imec: 2D Transistors on Production Wafers
ASML, TSMC and imec have jointly demonstrated two-dimensional transistors built on standard 300mm production wafers, marking the most concrete step yet toward a post-silicon chip era with two of the three partners being European. This episode of The State of Tech, The European Edition also covers the White House executive orders on frontier AI security and quantum computing, China's nearly $295 billion domestic AI infrastructure plan built around Huawei Ascend chips, and two open-source tools worth trying today.
Beluister deze aflevering:
Transcript
Samantha: Welcome to The State of Tech, The European Edition, Tuesday June twenty-third, 2026. I'm Samantha Lawrence.
Bob: And I'm Bob Russell. Today: the White House sets new rules of the road for advanced AI, the United States goes all-in on quantum, China commits nearly three hundred billion dollars to its own AI backbone, a major semiconductor breakthrough from ASML, TSMC and imec, and to close out, two things you can actually use today: a clever browser extension called DeArrow that fixes clickbait on YouTube, and a free open-source app called Joplin for keeping your notes truly yours. Let's start with the executive order out of Washington.
The White House sets a new voluntary framework for reviewing frontier AI before it ships.
Samantha: The White House has signed an executive order called Promoting Advanced Artificial Intelligence Innovation and Security. The headline change: a voluntary framework where developers of the most advanced AI models, what the order calls covered frontier models, can have those systems reviewed by the government before public release. Agencies have thirty days to strengthen federal cyber defences, a new AI cybersecurity clearinghouse is being set up, and prosecutors are told to prioritise AI-enabled cyberattacks.
Bob: The framing here is interesting. This is not Brussels-style hard regulation, it's a voluntary handshake between AI labs and national security agencies. The signal is that Washington now treats advanced AI the way it treats nuclear research or advanced weapons: something the state wants visibility on before it goes live. The clearinghouse is essentially a central exchange for threat information, so if one company spots an AI being used to write malware, the others get warned quickly. And the thirty-day window on federal cyber defences suggests the administration thinks the attack side is moving faster than the defence side right now.
Samantha: For Europe the question is whether this complements or competes with the AI Act. Brussels built a rules-based regime with mandatory obligations. Washington is leaning on voluntary cooperation plus national security leverage. European regulators have said they want to align where possible on AI safety testing, and a US clearinghouse could become a de facto standard that European labs are expected to plug into, even without a formal treaty.
Bob: For European companies using American frontier models, this could mean longer lead times between a model being trained and being available in Europe, because the pre-release review adds a step. It also raises a familiar question: will European AI developers get the same access to that clearinghouse, or will it become a club of US labs plus a few close allies? Industry groups in Paris and Berlin will be watching the implementation details closely over the next month.
A second US executive order pours political weight behind quantum computing and a national workforce push.
Samantha: On the same day, a separate executive order targets quantum technology. The order tells the President's science adviser to update the National Quantum Strategy, with a focus on what the White House calls quantum-enabling technologies and tighter industry partnerships. It also launches a national effort to build the first quantum computer capable of real scientific discovery, sets a five-year deadline for deploying quantum sensors and networks, and puts money behind training an American quantum workforce.
Bob: To translate that: quantum computers use the strange behaviour of particles at very small scales to solve certain problems much faster than today's machines. Quantum sensors can measure things like gravity or magnetic fields with extreme precision, useful for navigation without GPS, or for medical imaging. The five-year sensor deadline is striking because that part of quantum is closer to practical use than the full computer.
Samantha: Europe has been quietly competitive here. Germany, the Netherlands and France all run national quantum programmes, and the EU has its own Quantum Flagship initiative. But the US order signals a step-change in political urgency, with the workforce element being the part European officials may envy most. Building quantum engineers takes a decade, and Europe's talent pipeline is thinner than its research output suggests.
Bob: The other angle is industrial. The order explicitly leans on industry partnerships, meaning American firms get pulled closer to government contracts and classified work. For European quantum companies, that could mean either more collaboration through allied programmes, or being locked out of certain US supply chains. Expect ministers in Brussels and Berlin to respond with their own announcements in the coming weeks.
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.
China commits nearly three hundred billion dollars to a domestic AI infrastructure built without NVIDIA.
Bob: China has unveiled a five-year investment plan worth two trillion yuan, roughly two hundred and ninety-five billion dollars, to build a national AI infrastructure. The plan connects data centres across the country and, crucially, mandates domestic technology. Huawei's Ascend chips are designated as the core processors, which means no reliance on NVIDIA. Industry watchers are now focused on the interconnect layer, the high-speed links between data centres, and whether China will go with optical networking or stick with existing approaches.
Samantha: Two hundred and ninety-five billion dollars is a serious number, but the more important detail is the chip choice. Until recently, even Chinese AI labs preferred NVIDIA where they could get it. Designating Huawei Ascend signals confidence that the domestic chip is now good enough at scale. Whether that's true in practice, especially for training the largest models, is the open question. There have been reports of Ascend struggling with reliability under heavy training loads, but Beijing seems willing to absorb that cost for independence.
Bob: For Europe this matters in two ways. First, it deepens the split between Western and Chinese AI ecosystems, which makes life harder for European companies that operate in both markets. A model trained on Ascend hardware in China may not behave identically to one trained on NVIDIA in Frankfurt, and compliance teams will need to track that. Second, it raises the obvious question of where Europe's equivalent commitment is. The EU Chips Act mobilised around forty-three billion euros, which is real money but a fraction of what Beijing and Washington are now putting on the table.
Samantha: The interconnect question is technical but consequential. Optical networking, where data moves as light between servers, could let China leapfrog a generation of hardware. If they pull that off at national scale, it changes the cost curve for training huge AI models, which is exactly the area where Europe is most exposed.
ASML, TSMC and imec show transistors built from two-dimensional materials are ready for industrial wafers.
Bob: For the non-AI story of the day, a genuinely important semiconductor breakthrough. ASML from the Netherlands, TSMC from Taiwan and imec from Belgium have together demonstrated a way to build transistors from so-called two-dimensional materials on standard three-hundred-millimetre wafers, the size used in real chip factories. The materials in question, molybdenum disulphide and tungsten compounds, are sheets just a few atoms thick. The point is they can be made far smaller than silicon while still working reliably.
Samantha: For years, two-dimensional transistors have been a lab curiosity. The leap here is the integration on standard factory wafers at industrial scale, with a fifty-nanometre pitch, which is roughly the distance between the wires connecting transistors. That brings these exotic materials into the realm of something fabs can actually build, rather than just publish papers about. It's a credible path beyond the limits of silicon, which the industry has been warning about for a decade.
Bob: The European angle is unusually strong here. ASML is the Dutch company that builds the lithography machines every advanced chip factory depends on. imec is the Belgian research hub that quietly underpins much of global semiconductor R and D. So two of the three partners on this breakthrough are European, working with the world's leading manufacturer in Taiwan. It's a reminder that despite all the talk about chip sovereignty, Europe sits at the heart of the global semiconductor research stack.
Samantha: The timeline is the cautious part. Industrial adoption of a new transistor material usually takes years, not months. But the work shows the path is real, and chipmakers can start planning. It also strengthens Europe's negotiating position in any future discussion about chip supply, export controls, or research partnerships with the US and Asia.
DeArrow is the free open-source extension that strips clickbait titles and thumbnails from YouTube.
Bob: To close out, two things you can actually use today. The first is a browser extension called DeArrow. It does one specific thing: it replaces clickbait video titles and exaggerated thumbnails on YouTube with descriptions and screenshots submitted by other users. So instead of an all-caps title with three fire emojis and a shocked face, you see a plain summary of what the video is actually about.
Samantha: It's made by the same developer behind SponsorBlock, which many people already use to skip sponsor reads in videos. DeArrow is free, open-source, and works on Chrome, Firefox, Edge and Safari, plus on Android through Firefox. There's no account needed, and the project is community-driven, meaning the alternative titles and thumbnails come from volunteers who flag misleading ones.
Bob: The European angle is privacy. Because DeArrow runs locally in the browser and doesn't require sign-up, there's no profile being built around your viewing habits. Reviewers note that on a typical YouTube homepage, somewhere between a third and half of the thumbnails get cleaner alternatives, which makes browsing less manipulative. It takes about a minute to install from your browser's extension store, and the effect is immediate.
Samantha: It's also a small example of the wider open-source movement to make the web less hostile. You can fine-tune which titles it replaces, keep the original where the community hasn't submitted an alternative yet, and even contribute your own honest titles if you want. Users report it noticeably reduces the urge to click on videos that turn out to be a waste of time.
Bob: For anyone who finds YouTube increasingly noisy, this is one of the cleaner fixes available right now, and it doesn't break anything else. The project lives on GitHub if you want to look under the hood.
Joplin is the free open-source notes app that keeps your knowledge base entirely under your control.
Samantha: The second pick is for anyone tired of notes apps that lock you into a subscription and a cloud service. Joplin is a free, open-source note-taking and to-do app that runs on Windows, Mac, Linux, Android and iOS. Notes are stored as plain text files using a format called Markdown, which means they remain readable even if the app disappears tomorrow.
Bob: The key difference from the well-known commercial options is where your data lives. Joplin lets you sync across devices using your own choice of storage: a Nextcloud server, Dropbox, OneDrive, or even just a folder on your own machine. There's also end-to-end encryption built in. Users in regulated industries, lawyers, doctors, journalists, often cite Joplin for exactly that reason: the notes never have to touch a server you don't control.
Samantha: It handles the things people actually need: attachments, to-do lists, tagging, search across thousands of notes, and a web clipper extension to save articles from your browser. There's also a plugin ecosystem if you want to add things like mind maps or calendar views. It's not as polished out of the box as some paid alternatives, but reviewers consistently praise its reliability and the fact that you genuinely own your data.
Bob: For European users worried about where notes and personal information end up under various data regimes, the appeal is obvious. Setup takes maybe ten minutes if you already have a cloud account to sync with, and longer if you want to host it yourself. Either way, there's no subscription, no upsell, and your text files stay text files.
Samantha: Today we covered: the White House executive order on advanced AI and cybersecurity, the separate US push on quantum computing, China's nearly three hundred billion dollar plan for a domestic AI infrastructure, the semiconductor breakthrough from ASML, TSMC and imec on two-dimensional transistors, and two things to try yourself: DeArrow to clean up YouTube clickbait, and Joplin to keep your notes truly your own.
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.