SK Hynix HBM Supply: Tech Giants Finance ASML Lines

Samantha: Welcome to The State of Tech, The European Edition, Saturday May ninth, 2026. I'm Samantha Lawrence.

Bob: And I'm Bob Russell. Today: Samsung and Meta racing into humanoid robots, a new mission-critical AI platform launching in Abu Dhabi, Oracle Cloud adding Grok 4.3 and NVIDIA's Nemotron 3 Nano Omni, a global cyberattack taking down the Canvas learning system, Taiwan's exports hitting near-record levels on AI demand, and the global scramble around SK Hynix memory chips. Let's start with the humanoids.

Samsung and Meta race to commercialise AI-powered humanoid robots, with Samsung unveiling its first model later this year.

Samantha: Samsung is reportedly preparing to unveil a humanoid robot in the second half of this year. The first deployments will be on its own turf, semiconductor and display production lines, and from there the plan is to expand into retail and eventually the home.

Bob: And Meta is moving in the same direction. We're seeing a clear shift, AI is stepping out of the chatbot window and into machines that can pick things up, walk around, and operate in physical spaces. The market forecasts back this up. Industry projections put global humanoid robot shipments at around one million units a year between 2030 and 2035.

Samantha: One million a year is a serious number. And it's not just Samsung and Meta. Tesla, Figure, Agility, the Chinese players, they're all pushing hard. What's interesting about Samsung's approach is starting inside its own factories. That gives them a controlled environment to iterate, before sending these things into shops or living rooms.

Bob: It also solves the data problem. Every shift on a production line generates training data on grasping, navigating, handling fragile components. Once that's good enough, you port it to retail, then to consumer use cases. That's the playbook.

Samantha: For Europe, this is a tricky one. We have strong robotics research, ABB, KUKA, the Fraunhofer institutes, and a deep manufacturing base. But on consumer-facing humanoids backed by huge AI models, the capital is mostly in Asia and the United States.

Bob: There's also a regulatory layer coming. The AI Act already covers high-risk systems, and humanoid robots in workplaces or public spaces will sit right in the middle of that. Brussels will need to work out liability, safety certification, and data protection for machines that are constantly observing their environment.

Samantha: And labour markets. If humanoids can credibly do warehouse work or basic retail tasks within five years, European unions and policymakers will want a seat at the table early.

Bob: One thing to watch, the supply chain. These robots need advanced sensors, actuators, and edge AI chips. European suppliers like Infineon, STMicroelectronics, Bosch, they could carve out solid positions, even if the headline brand sits in Seoul or Menlo Park.

Samantha: So the bottom line, physical AI is moving from demo videos to deployment. And the next two years will tell us whether Samsung's factory-first approach actually scales beyond the cleanroom.

Abu Dhabi startup TACTICA AI launches the first agentic AI platform for mission-critical real-time decisions.

Bob: TACTICA AI has come out of Abu Dhabi with a platform aimed at mission-critical operational decisions. It's built on technology from the Technology Innovation Institute, and the pitch is that it goes beyond dashboards.

Samantha: The key word here is agentic. Instead of presenting a screen full of charts, the system defines a mission outcome and then dynamically pulls in the data sources, models, and workflows it needs. Sensor feeds, intelligence reports, operational data, all fused into one operational picture.

Bob: And critically, they're keeping a human in the loop. That's the right call for anything labelled mission-critical. The platform has been validated in real-world deployments, though TACTICA isn't naming names, which is normal in this space.

Samantha: The use cases are obvious, defence, border security, critical infrastructure, emergency response. Anywhere you have fragmented data and need to act fast.

Bob: For Europe, this lands in a sensitive spot. European defence and critical infrastructure operators have been talking about exactly this kind of decision-support layer for years. Some of it sits inside national programmes, some inside companies like Thales, Leonardo, or Helsing.

Samantha: And there's a sovereignty question. Do European agencies want to run mission-critical decisions on a platform built in the UAE, however capable it is? Or will this push Brussels and member states to accelerate homegrown alternatives?

Bob: It also sets a benchmark. If TACTICA can demonstrate genuine agentic orchestration in operational settings, that raises the bar for everyone else. Dashboards alone won't cut it anymore.

Samantha: My takeaway, the Gulf region is no longer just buying AI, it's building it and exporting it. European buyers will need to weigh capability against strategic dependency.

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.

Oracle Cloud adds xAI Grok 4.3 and NVIDIA Nemotron 3 Nano Omni to its enterprise AI platform.

Bob: Oracle Cloud Infrastructure has just made two major models available on its enterprise AI platform. xAI's Grok 4.3, which only launched the day before, and NVIDIA's Nemotron 3 Nano Omni.

Samantha: Grok 4.3 is interesting. It has a one million-token context window and ranks in the global top ten on the Artificial Analysis Intelligence Index. That puts it in serious territory for enterprise reasoning workloads.

Bob: And Nemotron 3 Nano Omni is fully open-source and multimodal. Video, audio, images, and text, all reasoned about in a single model. That matters because today, a lot of teams stitch together three or four models to do what Nemotron does in one.

Samantha: Less plumbing, fewer integration headaches, lower latency. For developers, that's a real productivity gain.

Bob: Oracle is positioning itself as the neutral cloud where you can pick the best model for the job. Microsoft pushes OpenAI, Google pushes Gemini, AWS pushes Anthropic. Oracle is saying, take your pick, run it on our infrastructure.

Samantha: For European businesses, this is genuinely useful. Many large European enterprises run on Oracle databases, banks, telcos, public sector. Having advanced models available on the same platform makes adoption a lot smoother.

Bob: There's a data residency angle too. Oracle has been investing in EU sovereign cloud regions, partnerships with operators in France and Germany. If these models are available within those sovereign boundaries, that solves a real procurement headache.

Samantha: Though I'd note, Grok comes with its own reputational baggage given xAI's positioning. European compliance teams will want to look closely at how it handles bias, content moderation, and the AI Act's transparency requirements.

Bob: Fair point. Capability is one thing, deployability under European regulation is another. But the direction of travel is clear, more model choice, faster availability, and pressure on the hyperscalers to keep up.

Global cyberattack knocks Canvas offline, disrupting nearly nine thousand schools and universities.

Samantha: Tens of thousands of students and educators woke up to a Canvas outage after a cyberattack on the online learning platform. Instructure, the company behind Canvas, took the system offline as a precaution.

Bob: The attackers exploited an issue tied to the Free-For-Teacher accounts and made changes to user pages. A hacking group has reportedly claimed responsibility, saying nearly nine thousand schools were affected and that they accessed billions of private messages and records.

Samantha: Billions is a staggering number, and it needs verification. But even if the real figure is a fraction of that, this is a serious incident. Canvas is used across schools and universities globally, including a lot of European institutions.

Bob: And it lands in a context where ed-tech has become critical infrastructure. Lectures, assignments, grades, communication with parents, it all flows through these platforms. When Canvas goes down, teaching effectively stops.

Samantha: For Europe, there are two immediate questions. One, GDPR. If personal data of European students has been accessed, Instructure has seventy-two hours to notify supervisory authorities, and we should expect regulators in several member states to open inquiries.

Bob: Two, the NIS2 directive. Education isn't always classified as essential under NIS2, but a lot of universities are research institutions tied to critical sectors. So there's a broader question about whether the resilience of educational platforms is being taken seriously enough.

Samantha: And there's the procurement lesson. Many European schools picked Canvas during the pandemic for speed. Now they're locked in, and they're discovering that vendor concentration carries risk.

Bob: The fix isn't simple. Most institutions don't have the budget or expertise to run their own learning management systems. But diversifying providers, demanding stronger security guarantees, and rehearsing incident response, those are all things European education ministries should be pushing on now.

Samantha: Expect this incident to be cited in policy discussions for months.

Taiwan's exports hit near-record sixty-seven billion dollars on surging AI and high-performance computing demand.

Bob: Taiwan's exports last month reached sixty-seven point six two billion US dollars, just shy of an all-time record. That's the thirtieth consecutive month of expansion.

Samantha: The driver is AI and high-performance computing, AI servers, advanced semiconductors, cloud infrastructure equipment. Information, communications and audio-visual products jumped sixty-two point three percent. Electronic components rose nearly thirty-nine percent.

Bob: Those are extraordinary growth rates for an export category that was already huge. It tells you the AI infrastructure build-out is still accelerating, not plateauing.

Samantha: Taiwan's finance ministry expects the momentum to continue, because the major cloud providers, hyperscalers, are still ramping up capital expenditure on AI infrastructure. We're talking hundreds of billions of dollars combined this year.

Bob: And almost every dollar of that flows through Taiwan in some form. TSMC fabricates the chips, the Taiwanese ODMs assemble the servers, the component makers supply everything from power systems to cooling.

Samantha: For Europe, this has two faces. On one hand, robust supply means European cloud customers and AI startups can keep buying. On the other hand, prices for advanced GPUs and AI servers remain high, and lead times remain long.

Bob: There's also the geopolitical dimension that never goes away. The concentration of advanced chip production in Taiwan is the single biggest supply chain risk in global tech. The European Chips Act is trying to build alternatives, but realistically we're years away from meaningful capacity in Europe.

Samantha: And the gap between European AI infrastructure spending and what's happening in the US and Asia keeps widening. European hyperscalers and sovereign cloud projects are real, but they're a fraction of what AWS, Microsoft, Google, and the Chinese giants are buying.

Bob: The takeaway is that Europe is a customer in this boom, not a primary beneficiary. Closing that gap is going to take a decade of consistent industrial policy.

Samantha: Or some very creative partnerships. Which brings us neatly to our last story.

Global tech giants line up with cash and equipment offers to lock in SK Hynix memory chip supply.

Bob: SK Hynix is being courted, and the offers on the table are eye-catching. Major global tech companies are proposing direct investment in dedicated memory production lines, and even financing for ASML's extreme ultraviolet lithography machines, which run into the hundreds of millions per unit.

Samantha: That's how desperate the demand is. Customers aren't just placing orders, they're offering to pay for the factories.

Bob: SK Hynix is the leader in high-bandwidth memory, the HBM stacks that sit alongside NVIDIA's GPUs in every AI server. Without HBM, the AI boom doesn't happen. So clients are doing whatever they can to secure allocation.

Samantha: And SK Hynix is playing it cool. They're cautious about accepting financial commitments that could compromise their pricing flexibility or their independence. Smart move, when you're the bottleneck, you don't give that leverage away cheaply.

Bob: It's a fascinating reversal of the usual dynamic. Normally chipmakers chase customers. Here, the customers are chasing the chipmaker with chequebooks open.

Samantha: For Europe, there's a small bright spot in this story, ASML. The Dutch lithography giant is the only company in the world making EUV machines, and every one of those proposed investments includes ASML kit. Veldhoven is having a very good year.

Bob: But beyond ASML, European memory buyers, automotive companies, industrial groups, AI startups, are at the back of the queue. The hyperscalers are willing to pre-pay for entire production lines. European mid-sized firms can't match that.

Samantha: It does make you smile though, the image of executives flying to Seoul with offers to buy the machines, not the chips. We've come a long way from just-in-time procurement.

Bob: From spreadsheets to suitcases full of cash. That's the AI economy in 2026.

Samantha: Today we covered: Samsung and Meta pushing into humanoid robots, TACTICA AI launching mission-critical decision support from Abu Dhabi, Oracle adding Grok 4.3 and Nemotron to its cloud, the Canvas cyberattack hitting thousands of schools, Taiwan's record exports on AI demand, and the global rush to lock in SK Hynix memory.

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.