8 juni 2026 · 13:27
Nvidia Korea AI Deals: SK Hynix, NAVER & Doosan
Nvidia's CEO Jensen Huang left Seoul this week with multi-year AI infrastructure deals signed with SK Hynix, NAVER and Doosan, committing to gigawatt-scale AI data centres in South Korea while Europe is still debating who funds its own AI factories. This episode unpacks the Nvidia Korea AI deals and what they signal for European competitiveness, covers the ShinyHunters dump of 2.6 million DentaQuest records, and explores why AI app builders are quietly reshaping software development for businesses that cannot find a developer.
Beluister deze aflevering:
Transcript
Samantha: Welcome to The State of Tech, The European Edition, Monday June eight, 2026. I'm Samantha Lawrence.
Bob: And I'm Bob Russell. Today: Nvidia signs sweeping AI deals across South Korea, a major US healthcare data breach hits two point six million people, Nvidia teams with Unitree on humanoid robots for researchers, the rise of AI app builders for non-coders, and to close out, two things you can try yourself: the AI app builder Bolt, and the open-source robotics simulator NVIDIA Isaac Lab. Let's start with that Korean megadeal.
Nvidia strikes sweeping deals with SK Hynix, NAVER and Doosan to build AI data centres across South Korea.
Samantha: Nvidia's CEO Jensen Huang has been in Seoul, and he's not gone home empty-handed. The chip giant has signed multi-year partnerships with three of South Korea's biggest names: memory maker SK Hynix, internet group NAVER, and industrial conglomerate Doosan. The goal: build out gigawatt-scale AI data centres, the kind of facilities that now get called AI factories.
Bob: And each piece fits together. SK Hynix is co-developing the next generation of memory chips for Nvidia's upcoming systems, including the so-called Vera Rubin supercomputers due in the coming years. NAVER, which is essentially South Korea's Google, is expanding its data centre in Sejong and will use Nvidia's full stack to run its own Korean-language AI model and new agent-style services. Doosan brings the heavy industrial side, the power and the infrastructure.
Samantha: What's notable is the scale. NAVER is talking gigawatt-class capacity, which is the energy footprint of a small city. And SK Hynix is locking itself in as Nvidia's preferred memory partner for years to come. That matters because high-bandwidth memory is currently the single biggest bottleneck in AI hardware. If you can't get the memory, you can't ship the chips.
Bob: For South Korea this is also a sovereign AI play. Seoul wants to make sure its companies, hospitals and government run on Korean infrastructure with Korean-language models, not just rent capacity from American clouds. And Nvidia gets a deeper foothold in a country that already makes most of the world's advanced memory.
Samantha: The European angle here is uncomfortable. While South Korea, the United States, and the Gulf states are locking in gigawatt-scale AI deals, Europe is still debating where its own AI factories should go and who pays for them. The EU has its AI Gigafactory plans, but the contracts being signed in Seoul this week are larger and more concrete than anything announced in Brussels so far.
Bob: And there's a supply chain knock-on. If SK Hynix prioritises Nvidia and Korean customers, European AI startups buying Nvidia systems may end up further down the queue. The bottom line: the geography of AI compute is being decided right now, and Europe is increasingly a customer rather than a host.
ShinyHunters dumps stolen data of two point six million people from US dental benefits firm DentaQuest.
Samantha: Onto something darker. The cybercrime group ShinyHunters has published a 234 gigabyte archive of data they say was stolen from DentaQuest, a US dental benefits administrator. The dump followed failed ransom negotiations, which is becoming the standard playbook: steal first, demand payment, leak if the victim refuses.
Bob: The numbers are significant. Around two point six million individuals may be affected. The exposed information includes names, email addresses, phone numbers, home addresses, and in some cases Medicaid identification numbers. That last category is particularly sensitive because Medicaid IDs can be used for medical identity fraud in the US system.
Samantha: DentaQuest has confirmed there was unauthorised access to part of its network and says it's working with cybersecurity experts and law enforcement. But the data is already out there, which is the brutal reality of these leaks. Once it's published, no statement undoes it.
Bob: ShinyHunters has been busy this year. They're the same group linked to several high-profile breaches, and their tactic of dumping data after ransom failure is putting pressure on every company sitting on large customer databases.
Samantha: The European lens: this is a US healthcare breach, but the lesson travels. Under the EU's NIS2 directive, healthcare providers and their suppliers are classed as essential entities with strict cybersecurity and reporting obligations. A breach of this scale at a European dental insurer would trigger mandatory disclosure within tight deadlines and potentially significant fines under both NIS2 and GDPR.
Bob: And the practical reminder for listeners: if you've ever had dental insurance in the US, or your employer used a US-based benefits provider, your data may be in that archive. Watch for phishing attempts that reference dental claims or Medicaid in the coming weeks. That's often where stolen data ends up first.
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 partners with China's Unitree to put humanoid robots into research labs worldwide.
Bob: Nvidia again, but in a very different corner of tech. The company is teaming up with Chinese firm Unitree Robotics to offer humanoid robot platforms aimed specifically at researchers. Unitree is already well known for its agile four-legged robots and more recently for affordable humanoids, and now those machines come bundled with Nvidia's AI software stack.
Samantha: The pitch is straightforward. Until now, if a university or company lab wanted to do serious humanoid robotics research, they often had to build their own hardware or buy from a handful of expensive Western suppliers. Unitree has driven the price of a research-grade humanoid down dramatically, and pairing that with Nvidia's robotics tools lowers the barrier further.
Bob: That barrier matters. Humanoid robotics is where a lot of the next wave of physical AI research is happening, teaching robots to walk, grip, balance, and increasingly to follow plain-language instructions. Cheaper, more accessible platforms mean more labs can experiment, which usually accelerates progress.
Samantha: There is a geopolitical wrinkle though. Unitree is a Chinese company, and several Western governments have grown nervous about Chinese robotics suppliers, especially anything that ships with cameras, microphones, and connectivity. The US has already restricted some Chinese drones for federal use, and humanoid robots raise similar questions.
Bob: For European universities, this is a real choice. Cheaper hardware and a serious Nvidia software stack on one hand, versus procurement rules and data security reviews on the other. Some German and Dutch research institutes have already started writing China-risk assessments into their robotics purchasing.
Samantha: The wider story is that humanoid robotics is shifting from corporate demos to actual research tools. Whether the robots in European labs end up being Chinese, American, or, eventually, European-built is a question that's going to get louder over the next year.
AI app builders let people ship working software from a single sentence, no coding required.
Bob: Our last news story is about a shift that's quietly reshaping how software gets made. A new generation of AI app builders lets people describe what they want in plain language and get a working application back, with the interface, the database, and the behind-the-scenes logic all generated automatically.
Samantha: This isn't the old no-code promise of dragging boxes around a screen. You type something like, build me a booking system for a small dental practice with email reminders, and the tool produces a deployable app. Tools like Bolt, Lovable and Replit Agent have moved from prototype demos last year to actually shipping software that small businesses run on.
Bob: The market impact is starting to show. Indie developers and small founders are launching products in days rather than months. That's pulling work away from traditional freelance developers for the simpler end of the market, the kind of internal tools and small customer-facing apps that used to be a steady source of contracts.
Samantha: For European small and medium-sized businesses, this is potentially a big deal. The shortage of software developers across the EU has been a persistent drag on digitalisation, especially outside the big tech hubs. A bakery chain in Portugal or a logistics firm in Poland can now build internal tools without hiring a developer they couldn't find anyway.
Bob: There are caveats. Apps generated this way can have security gaps, struggle at scale, and lock you into whatever platform built them. And under the EU AI Act and data protection rules, if your AI-generated app handles personal data, you, the business owner, are still legally responsible, even if you never wrote a line of code.
Samantha: That accountability gap is worth watching. The tools are getting impressive, but the legal and security plumbing around them is still catching up. The takeaway for European businesses: try them for low-risk internal tools first, not for anything touching customer data, until you've had someone competent check the result.
Bolt lets you build and launch a real web app this afternoon, just by describing it.
Bob: Which brings us nicely to our first thing to try yourself. Bolt is one of the AI app builders we just mentioned, and it's genuinely usable today. You go to bolt dot new in your browser, type what you want to build in a sentence or two, and within a minute you've got a running web app you can poke at, change, and publish.
Samantha: The use cases people are actually shipping with it are surprisingly broad. A simple landing page for an event, a small booking form, an internal dashboard that pulls in a spreadsheet, even basic games. It writes the code in the background, so if you do know a bit of programming you can dive in and edit, but you don't have to.
Bob: There's a free tier that gives you enough to experiment, and paid plans if you want to deploy something seriously. It runs entirely in your browser, no installation, works on a laptop running Windows, Mac or Linux. You can have something working before your next coffee.
Samantha: A practical tip: be specific in your prompt. Saying, build me a website, gives you something generic. Saying, build me a one-page site for a yoga studio in Antwerp with a class schedule and a contact form, gets you something you can actually use. Treat it like briefing a junior developer, not a search engine.
Bob: And one honest warning: it's tempting to think you can ship a business on this in a weekend. For internal tools and quick experiments, yes. For anything that handles customer payments or personal data, you still want someone who knows what they're doing to look at it before it goes live.
Samantha: But for tinkering, for that side project you've been putting off, or for replacing a clunky spreadsheet, Bolt is one of the most fun things you can play with this week. Free to try, no credit card needed to get started.
NVIDIA Isaac Lab lets anyone train a virtual robot on a normal computer, for free.
Bob: And to finish, something for the curious and the technically inclined, but no robot required. NVIDIA Isaac Lab is a free, open-source toolkit that lets you train and test robots in a virtual world on your own computer. With all the humanoid robotics news today, it's a way to see what researchers actually do, without spending tens of thousands on hardware.
Samantha: It runs on Windows or Linux, you download it from GitHub, and once it's set up you get a 3D environment where simulated robots, arms, four-legged dogs, even humanoids, learn to walk, balance and pick things up. The same software is used in serious university labs, so this isn't a toy version.
Bob: The honest truth: this is not a five-minute install. You need a reasonably powerful computer, ideally with an Nvidia graphics card, and you'll spend an afternoon getting it running. But there are step-by-step guides on the official NVIDIA developer site and a large community sharing example projects on GitHub.
Samantha: Why bother? Because robotics is moving from research papers into real products fast, and getting hands-on with how robots are actually trained gives you a much better sense of what's hype and what isn't. Teachers, students, hobbyists, anyone curious about where physical AI is heading can run real experiments at home.
Bob: And it's genuinely free. No subscription, no trial period. The whole thing is open source, which means you can also see exactly how the simulations work. If you've ever wondered what it feels like to teach a virtual robot to stand up without falling over, this weekend's project is sorted.
Bob: Today we covered: Nvidia's sweeping AI deals across South Korea, the DentaQuest data breach hitting two point six million people, Nvidia and Unitree bringing humanoid robots to research labs, the rise of AI app builders reshaping software development, and two things to try yourself: the AI app builder Bolt, and the robotics simulator NVIDIA Isaac Lab.
Samantha: 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.