Nvidia & Supermicro indicted: AI export controls escalate

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

Bob: And I'm Bob Russell. Today: staff at Nvidia and Supermicro indicted over AI server exports, Porsche signs a one-and-a-half-billion-dollar AI deal with Tata Consultancy, Apple unveils a new Mac mini built for on-device AI, a fresh warning that the AI investment boom is starting to look like a bubble, and to close out, two things you can actually try today: the writing tool Ellipsus Sprint, and the open-source photo cleanup app Chafa. Let's start with the export indictments.

Staff at Nvidia and Supermicro reportedly indicted over illegal AI server exports.

Samantha: Staff from Nvidia and Supermicro have been indicted over alleged violations tied to AI server exports. This is a significant legal move that puts individuals, not just corporations, in the crosshairs of export control enforcement. The details are still emerging, but the charges appear to focus on how advanced computing hardware moved through the global supply chain, in defiance of restrictions on where these systems can legally end up.

Bob: And the timing is telling. Export controls on high-end AI chips have been tightening for two years now, especially around destinations where Washington fears the hardware could feed military or surveillance programmes. What is new here is the willingness to prosecute employees directly. Companies can absorb fines and treat them as a cost of doing business. Indicting staff changes the calculation completely, because now you are asking individual sales managers, logistics people, and compliance officers whether they want to risk personal criminal liability.

Samantha: For the companies themselves, the reputational damage lands immediately. Nvidia sits at the centre of the entire global AI build-out, and Supermicro assembles a huge share of the servers hyperscalers rely on. Any disruption to their export practices ripples outward fast. Buyers in allied markets will be watching closely to see whether shipments slow while compliance is retooled.

Bob: For Europe the picture is mixed. On one hand, tighter enforcement could squeeze the already limited supply of advanced AI servers reaching European data centres, pushing prices up and delaying build-outs at a moment when Europe is trying to catch up on sovereign AI infrastructure. On the other hand, it strengthens the argument that Europe cannot keep depending on a single supply chain policed from Washington. Expect renewed calls in Brussels for European chip capacity, European assembly, and clearer rules on who can buy what. This story is also a reminder that the AI race is not just a technology contest, it is a legal and geopolitical one, fought in courtrooms and customs offices as much as in labs.

Porsche signs a one-and-a-half-billion-dollar AI deployment deal with Tata Consultancy Services.

Samantha: Porsche has signed a five-year deal worth one-and-a-half billion dollars with Tata Consultancy Services, India's largest IT services firm. TCS will build a dedicated hub to roll out AI across Porsche's operations, from engineering to customer experience. As part of the agreement, Porsche is also transferring its in-house IT consultancy business to TCS for around three hundred and twenty million euros.

Bob: This is a landmark deal for two reasons. First, the size. One-and-a-half billion dollars over five years is a serious commitment from a luxury carmaker that is already navigating a difficult transition on electric vehicles and Chinese competition. Second, it says a lot about who the winners are in the enterprise AI boom. It is not always the model makers themselves. Often it is the integrators, the firms that can actually plug AI into decades of legacy systems and messy real-world workflows.

Samantha: And that is where Indian services giants have quietly become indispensable. TCS, Infosys, Wipro, they have been building AI delivery centres for European clients for a couple of years now. For Porsche, offloading its IT consultancy arm also means shedding costs at a time when margins in the car industry are under pressure. Analysts point out this fits a broader pattern of European industrial names outsourcing digital transformation rather than trying to build it in-house.

Bob: The European angle here is really about competitiveness. European carmakers know they cannot fall behind on AI-driven manufacturing, predictive maintenance, and personalised in-car experiences. The bottom line is that Porsche is betting a partner can move faster than internal teams. Whether that pays off, or whether Porsche ends up dependent on an external supplier for something strategic, is the question we will be watching for the next five years.

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.

Apple unveils a new Mac mini built around on-device AI processing.

Bob: Apple has announced a new Mac mini, and the pitch is clear: this is a machine built for on-device AI. It uses Apple's latest chips, and the company is claiming up to four times faster AI performance than the previous generation. The new chips include what Apple calls neural accelerators inside each graphics core, which for the first time makes serious local AI workloads viable on a Mac mini.

Samantha: What Apple is describing is what they call always-on agentic computing. In plain language, that means the machine can keep AI tasks running in the background, upscaling photos and videos, running image generation models locally, or acting as a small home AI server. Until now, if you wanted that kind of horsepower, you were sending your data to the cloud, or you were spending far more on a bigger workstation.

Bob: This matters because the direction of travel in AI is shifting. For two years the story was all about massive cloud data centres. Now companies from Apple to Qualcomm to Nvidia are pushing hard on running AI directly on your own device. The advantages are lower latency, better privacy, and no per-query cloud bill. For creative professionals, developers, and small studios in Europe, that is genuinely useful, because it means sensitive work does not have to leave the building.

Samantha: The European angle is straightforward. European users get access to a more affordable on-device AI machine, which fits neatly with the region's stricter data protection culture. For anyone building local AI applications, running language models on their own hardware, or simply worried about where their data ends up, this lowers the barrier. It also nudges the whole industry toward a hybrid future, where some AI lives in your pocket or on your desk, and only the heavy lifting goes to the cloud.

A widely cited MIT study finds ninety-five percent of generative AI pilots show no measurable profit impact.

Bob: There is growing concern that the AI investment boom is starting to look uncomfortably like a bubble. Since 2023, trillions of dollars have flowed into AI infrastructure globally, but the returns are lagging badly. A widely cited MIT study found that ninety-five percent of generative AI pilots at large companies show no measurable profit impact. Analysts also estimate a six-hundred-billion-dollar gap between what companies are spending on AI infrastructure and what the ecosystem is actually earning from it.

Samantha: And Gartner is warning that over forty percent of so-called agentic AI projects could be cancelled by 2027, because the returns simply are not showing up. This does not mean AI is a fraud. It means the market is repeating a classic pattern: massive capital expenditure ahead of proven use cases, hoping demand catches up. We saw the same dynamic during the dot-com boom, and during the fibre-optic buildout of the late nineties. Some of that capacity eventually paid off, but a lot of investors got wiped out along the way.

Bob: What is interesting is who is most exposed. The hyperscalers, the Microsofts and Googles of the world, are pouring hundreds of billions into data centres based on the assumption that enterprise demand will scale. If the MIT figures are right, and most pilots are quietly failing, that assumption starts to look wobbly. Meanwhile, chip suppliers and server builders are booking record revenue today on orders that could be cut tomorrow.

Samantha: For Europe, the lesson is caution. European pension funds, insurers, and sovereign wealth funds have significant exposure to US tech through index positions. If a correction comes, it lands on European savers too. European companies rushing into AI projects should critically evaluate return on investment before committing serious capital. The bottom line: this is the moment to ask hard questions about what AI is actually delivering, not what it promises.

The free European writing tool Ellipsus Sprint helps you finish that draft you have been avoiding.

Bob: To close out, two things you can actually use today. First up, a writing tool called Ellipsus Sprint. Ellipsus is a European platform built for collaborative long-form writing, and Sprint is its focused-writing mode. You set a target, either words or minutes, and the interface strips down to just your text and a countdown. No formatting toolbars, no comments, no distractions.

Samantha: What makes Sprint different from a plain text editor is that it saves versions automatically and lets you compare drafts later. Writers report using it for everything from novel chapters to policy memos to long reports. It runs entirely in the browser, works on Windows, Mac, and Linux, and there is no subscription for the core writing features. Teams can also share documents without handing over rights to a large US cloud, which for European users nervous about data protection is worth noting.

Bob: For anyone staring at a blank page this week, Sprint is a low-friction way to just get words down. Users describe it as calmer than the big office suites, because there is nothing pulling your attention sideways. You can be up and running in about two minutes. Sign up with an email, click the Sprint button, pick a target, and start typing.

Samantha: The catch, and there is always one, is that Sprint is designed for writing, not for heavy editing or layout. If you need footnotes, complex tables, or precise typesetting, you will want to export to something else at the end. But for the drafting stage, which is where most projects get stuck, it is a genuinely useful piece of software from a European team. Free to start, and no download required.

The free open-source app Chafa cleans up and organises photos directly on your own computer.

Bob: And the second pick for today is Chafa, a free open-source tool for viewing and processing images directly from your terminal or file browser. It is a small utility with a specific talent: it can preview huge photo folders quickly, without needing a heavy photo app running in the background.

Samantha: The reason to care about this one is speed. If you have twenty years of holiday photos on an external drive, opening them in a full photo library can take forever. Chafa lets you scroll through them almost instantly, spot duplicates, and pick out the ones worth keeping. It runs on Linux, Mac, and Windows, and it is completely free.

Bob: Users praise it for being lightweight and honest. It does one thing well: showing you your images fast, so you can decide what to do with them. It pairs nicely with any of the photo organisers we have mentioned in the past, because you can use Chafa to triage first, then import only the keepers into your main library. Nothing leaves your device, which for privacy-minded European users is exactly the right default.

Samantha: Installation takes a couple of minutes on any modern system. There is no account, no cloud sync, no subscription. Just a small tool that respects your time and your storage.

Samantha: Today we covered: the Nvidia and Supermicro export indictments, Porsche's one-and-a-half-billion-dollar AI deal with Tata Consultancy, Apple's new AI-focused Mac mini, growing warnings of an AI investment bubble, and two things to try yourself: Ellipsus Sprint and Chafa.

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.