Huawei LogicFolding: 1.4nm chips without ASML by 2031

Samantha: Welcome to The State of Tech, The European Edition, Tuesday May twenty-six, 2026. I'm Samantha Lawrence.

Bob: And I'm Bob Russell. Today: Huawei claims a chip breakthrough that sidesteps Western restrictions, Google rebuilds its search engine entirely around AI, Nvidia bets big on China for the next wave of AI chips, scientists use AI to invent brand new semiconductor materials, a clever way to squeeze cash out of idle AI computers, and a tiny chip that brings smart vision to everyday devices. Let's start with Huawei.

Huawei claims chip breakthrough that could deliver 1.4-nanometer processors without Western lithography by 2031.

Samantha: Huawei has dropped a real bombshell at a chip conference in China. The company says it has developed something called LogicFolding, a new way of building advanced semiconductors that does not need the most cutting-edge European machines from ASML. Their roadmap points to 1.4-nanometer chips by 2031.

Bob: And that number matters because 1.4-nanometer is roughly where TSMC in Taiwan is heading, the absolute frontier of chipmaking. Huawei has been locked out of the most advanced lithography tools for years now under US export controls. So this is essentially Huawei saying, fine, we'll find another road to the same destination.

Samantha: Their semiconductor chief, He Tingbo, says the new Kirin mobile chips arriving this autumn will be the first to use this architecture. She's promising a, quote, big leap ahead in performance.

Bob: Now, a healthy dose of caution is in order. This is a company announcement at a Chinese conference, not an independently verified product on the market. We've seen Huawei claim surprise wins before, like the Mate 60 phone that shocked Washington two years back. But the gap between a marketing claim and a chip you can buy at scale is huge.

Samantha: Still, if even part of this pans out, the strategic picture shifts. Washington's whole China chip policy is built on the idea that without ASML's extreme ultraviolet machines, you simply cannot make truly advanced chips. Huawei is challenging that assumption head-on.

Bob: And that lands directly in Europe's lap. ASML, based in the Netherlands, is the single most important company in the global chip supply chain. Their machines are the chokepoint that gives the West leverage. If Huawei can genuinely route around them, that leverage weakens, and Europe loses some of its strategic weight in this whole geopolitical chess game.

Samantha: For European tech buyers it's more mixed. More competition could eventually mean more choice and lower prices for chips. But it also means a Chinese supply chain that's far less dependent on European tools, which has real consequences for how Brussels thinks about export rules going forward.

Bob: The realistic timeline is the key thing to watch. 2031 is five years out. A lot can happen, including this not working as advertised. But the direction of travel is clear: China is investing whatever it takes to become self-sufficient in chips, and the West's technology blockade is slowly being chipped away at, no pun intended.

Google flips the switch and makes its entire search engine run on Gemini AI by default.

Bob: This is a genuinely big shift for anyone who uses the internet, which is basically everyone. Google has switched its main search engine over to run entirely on its Gemini AI model. No more choosing between classic search and AI search. The whole thing is now conversational by default.

Samantha: So when you type a question, instead of getting a list of blue links, you get an AI-generated answer, with the option to ask follow-up questions like you would in a chat. The familiar search box is becoming a conversation box.

Bob: Which sounds convenient, and often it is. But it fundamentally changes how we encounter information. We move from actively searching, scanning sources, comparing, to passively consuming whatever the AI hands us. And that AI can absolutely be wrong, or subtly biased toward certain framings.

Samantha: There's also a massive knock-on effect for everyone who runs a website. If Google answers the question directly, users don't click through. News sites, recipe blogs, small businesses, they all suddenly see traffic evaporate. The whole search optimization industry is being upended overnight.

Bob: And there's the energy cost. Every AI-generated answer uses far more computing power than a traditional search. Multiply that by billions of queries a day and the environmental footprint of simply googling something goes up sharply.

Samantha: For European users, this raises real questions under the Digital Services Act and the AI Act. Brussels has been watching AI-generated content closely, especially around transparency and source attribution. Whether Google's new search meets those standards is going to be tested quickly.

Bob: My honest advice: cross-check anything important with a second source. The convenience is real, but treating an AI answer as gospel is a habit we should not let ourselves slip into.

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.

Nvidia's Jensen Huang says China remains essential to a 200 billion dollar AI chip market.

Bob: Nvidia chief Jensen Huang spoke in Taipei ahead of Computex 2026 and made it crystal clear: China is not a market Nvidia is willing to walk away from, no matter what Washington says. He pegs the upcoming market for so-called agentic AI chips at around 200 billion dollars, and China is a significant slice of that.

Samantha: Agentic AI is the buzzword of the moment. The idea is AI that doesn't just answer questions but actually takes actions on your behalf, books your travel, manages your inbox, handles tasks autonomously. That requires a different kind of chip, more focused on central processors than the graphics chips Nvidia is famous for.

Bob: Huang confirmed Nvidia has now obtained licenses to ship its H200 AI chips to China, although actual deliveries are still tangled up in regulatory and geopolitical knots. So the door has opened a crack, but goods aren't really flowing through it yet.

Samantha: This is a delicate dance. Nvidia is the most valuable company in the world right now largely because of AI demand, and they cannot afford to lose China as a customer. But every chip they ship there is scrutinised by US politicians worried about military and surveillance applications.

Bob: For Europe, the strategic question is whether we end up squeezed between an American AI hardware monopoly and Chinese alternatives forced into existence by sanctions. European AI builders depend almost entirely on Nvidia chips, and supply allocations have been a constant headache.

Samantha: There's also the rise of agentic AI itself to consider. If Huang is right that this is a 200 billion dollar market, European companies need to figure out fast whether they're building these systems themselves or just buying them from American providers.

Bob: And that's a familiar pattern. Europe regulates AI carefully, which is important, but we keep ending up as consumers of someone else's technology rather than producers of our own.

AI helps scientists invent entirely new semiconductor materials, slashing R&D time dramatically.

Samantha: This one is genuinely exciting and gets less headlines than it deserves. Researchers at Flinders University in Australia, working with Khalifa University in the UAE, have built an AI system that designs brand new semiconductor materials from scratch.

Bob: The traditional way of finding new materials is painfully slow. You either run physical experiments in a lab, which takes months, or run complex simulations that eat up enormous computing time. This AI essentially learns the hidden chemistry rules and then predicts entirely new compositions with the properties you want.

Samantha: They focused on gallium-based semiconductors, which are used in everything from high-speed wireless equipment to power electronics in electric cars. Instead of randomly trying combinations, the AI narrows it down to the most promising candidates almost instantly.

Bob: This is part of a broader trend. AI is starting to genuinely accelerate scientific discovery, not just summarise existing knowledge. We've seen similar breakthroughs in drug discovery and battery chemistry. Materials science is the next frontier.

Samantha: For Europe this matters because we have strong materials science traditions, places like Imec in Belgium, the Fraunhofer institutes in Germany. If they integrate this kind of AI-led discovery into their workflows, it could meaningfully speed up European innovation in chips, sensors, and wearable electronics.

Bob: The catch is always commercialisation. Discovering a promising new material in software is one thing. Getting it manufactured at scale is another, and that's where Europe sometimes drops the ball compared to Asia.

Samantha: Still, the direction is hopeful. The bottleneck in tech increasingly is not ideas, but how fast we can test and iterate them. AI is starting to remove that bottleneck in real ways.

Bob: And it shifts what scientists actually do day to day. Less time on trial-and-error in the lab, more time on judging which AI suggestions are worth pursuing. That changes the skillset universities need to teach.

Samantha: To close out, two things you can actually use today.

A new software platform helps companies turn idle AI computers into paying cloud services.

Bob: At Computex in Taipei today, a company called INFINITIX is showing off software that solves a very specific and increasingly common problem. Lots of businesses bought expensive AI servers in the last two years and are using less than 30 percent of them. INFINITIX's tools let you rent out that spare capacity as a cloud service.

Samantha: The platform is called AI-Stack, and it handles the messy bits, dividing up workloads, deploying AI models, charging customers, all the things that make running a mini cloud business too complicated to bother with otherwise.

Bob: Now, this is clearly aimed at businesses rather than individual users. But for our listeners who work in IT or run small companies with AI hardware sitting around, this is concretely worth investigating. Global AI spending is heading toward 2.52 trillion dollars in 2026, but utilization is poor. There's real money in that gap.

Samantha: The practical idea is interesting even if you don't use this specific product. If your company has GPU servers, you don't need to pick between using them yourself and renting them out, you can do both, slicing the capacity based on demand.

Bob: And for European data centres, this is relevant in a different way. Europe has been slow to build hyperscale AI infrastructure compared to the US. Smaller, distributed players who can monetize underused capacity are part of how Europe might catch up without massive new builds.

Samantha: The software supports both Nvidia chips and the newer dedicated AI chips from other vendors, so you're not locked into one brand. That flexibility is increasingly important as the chip market diversifies.

Bob: Worth a look if you've got AI hardware gathering dust. The basic principle, treat compute like a utility you can resell, is going to become standard practice in the next couple of years.

Samantha: And a useful reminder that not every AI breakthrough is about bigger models. Sometimes it's about using what we already have far more efficiently.

A tiny new edge AI computer brings smart vision and object recognition to everyday devices.

Bob: Japanese firm Contec has launched a small, fanless computer with a built-in AI chip from Hailo, designed for what we call edge AI. That means doing AI work directly on a device, not sending everything to the cloud.

Samantha: So think security cameras that recognise objects instantly, factory equipment that spots defects in real time, or smart traffic systems analysing footage right at the intersection. All without needing a fast internet connection or a big cloud bill.

Bob: And helpfully, Contec is running an evaluation loan program. So if you're a developer or a small company curious about building edge AI applications, you can actually borrow a unit to test before committing.

Samantha: They've also released something called Edge AI Compass, basically a getting-started toolkit. The whole package is aimed at lowering the barrier for smaller teams who want to build smart devices without hiring a specialist AI engineer.

Bob: This matters because edge AI is one of the few areas where Europe has real strengths. Companies like Bosch, Siemens, and a lot of mid-sized industrial firms have been pushing AI into factories, vehicles, and infrastructure for years.

Samantha: For listeners who are more on the hobby side, this kind of hardware is also becoming more affordable. Running real-time object recognition on a device the size of a paperback used to be science fiction. Now you can prototype it at home.

Bob: And there's a privacy upside. When AI runs locally, your camera footage or sensor data does not have to leave your premises. That fits much better with European data protection rules than constantly streaming everything to the cloud.

Samantha: A small but meaningful product, the kind that quietly powers a lot of next-generation smart systems.

Bob: Worth bookmarking if your work touches industrial automation, smart cities, or any kind of camera-based AI.

Samantha: Today we covered: Huawei's chip breakthrough claim, Google's full switch to AI-powered search, Nvidia's China bet on agentic AI chips, AI inventing new semiconductor materials, a clever way to monetize idle AI compute, and a compact edge AI computer for smart devices.

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.