SK Hynix, Samsung AI chips & ASML China restrictions

Samantha: Welcome to The State of Tech — The European Edition. I'm Samantha Lawrence.

Bob: And I'm Bob Russell. Today: SK Hynix chases massive memory chip deals with Microsoft and Google. Samsung posts a jaw-dropping profit surge driven by AI chips. Broadcom locks in major AI hardware partnerships with Google and Anthropic. OpenAI publishes a policy blueprint for the age of superintelligence. US states race ahead with a patchwork of AI laws. And new US legislation threatens ASML's China business. Let's start with the memory chip scramble.

The race to lock down AI memory supplies is getting intense — and expensive.

Samantha: So SK Hynix, the world's number two memory chipmaker, is reportedly in final-stage negotiations with both Microsoft and Google for multi-year DRAM supply contracts. We're talking potentially tens of billions of dollars here, Bob.

Bob: These are enormous commitments. Microsoft's deal alone is said to cover DDR5 memory over three years starting this year. And both Microsoft and Google are apparently pursuing similar long-term agreements with Samsung as well. So you've got the two biggest cloud players essentially trying to lock up as much advanced memory as they can get their hands on.

Samantha: And the reason is obvious — AI infrastructure is absolutely devouring memory chips. Every large language model, every training run, every inference workload needs massive amounts of high-speed memory. The supply just cannot keep up.

Bob: What's interesting is the knock-on effect. When Microsoft and Google lock in these multi-year contracts, that tightens supply for everyone else. Smaller cloud providers, enterprise buyers, startups — they're all competing for whatever's left.

Samantha: And that brings us to Europe. European cloud providers and tech companies need the same chips. If the big American players are hoarding supply through long-term deals with Korean manufacturers, European companies could face higher prices and longer wait times. It puts real pressure on procurement strategies.

Bob: It also raises the question of whether European firms need to get more creative — diversifying supply chains, maybe exploring alternative memory architectures. Because right now, the AI memory market is being shaped by a handful of very large buyers, and Europe isn't really at that table.

Samantha: Not in the same way, no. And that's a vulnerability worth watching.

Samsung just posted a profit number that made analysts do a double take.

Bob: Staying with chips — Samsung Electronics has reported a preliminary operating profit of 57.2 trillion won for the March quarter. That's roughly 37.9 billion US dollars. An eight-fold increase year over year. A new record.

Samantha: Seven hundred and fifty-five percent jump. That is staggering. And the driver is overwhelmingly AI memory chips — high-bandwidth memory, HBM, and other components going into data centers.

Bob: Cloud providers are building out AI infrastructure at a pace we haven't really seen before. And Samsung is one of only three companies in the world that can make the most advanced memory chips. So they're in an incredibly strong position right now.

Samantha: The bottom line for European companies is cost. If you're a European AI startup or a research institution trying to build out your compute capabilities, these numbers tell you that demand is outstripping supply. That means the chips you need are expensive and getting more expensive.

Bob: And lead times are stretching. European enterprises planning AI deployments need to be thinking further ahead than they might be used to. You can't just order chips and expect them next month. The queue is long, and the biggest buyers are at the front.

Samantha: There's also a strategic dimension. Europe has been talking a lot about digital sovereignty and building its own tech capabilities. But when it comes to the foundational hardware for AI — the memory chips, the processors — Europe is still overwhelmingly dependent on Asian manufacturers. Samsung's record profits are great news for Samsung. For Europe, it's a reminder of where the real bottlenecks are.

Bob: And those bottlenecks aren't going away anytime soon. If anything, they're tightening.

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.

Broadcom is quietly becoming one of the most important companies in AI — and just got more important.

Bob: Broadcom has secured two significant deals — one with Google for its next-generation Tensor Processing Units, and another with Anthropic to support their growing AI compute needs. This cements Broadcom's role as a go-to partner for custom AI silicon.

Samantha: This is a company that a lot of people outside the industry don't really think about, but Broadcom is increasingly the backbone of AI hardware. They design custom chips that companies like Google use instead of — or alongside — Nvidia GPUs.

Bob: The Google relationship has been deepening for years. Google's TPUs are a core part of their AI strategy, and Broadcom is central to making those chips a reality. The Anthropic deal is newer and signals that even AI-native companies — not just cloud giants — are turning to Broadcom for their infrastructure needs.

Samantha: For European companies, this matters in a couple of ways. First, if you're using Google Cloud's AI services, the performance and availability of those services depends partly on these Broadcom-designed chips. So you're indirectly affected by this partnership.

Bob: Second, it highlights a concentration problem. The number of companies that can actually design and manufacture cutting-edge AI hardware is very small. Nvidia, Broadcom, AMD, a handful of others. Europe doesn't really have a player in this space at scale.

Samantha: And European policymakers have noticed. There's growing concern about relying on so few hardware providers, almost all of them American or Asian. The EU Chips Act is trying to address semiconductor manufacturing, but custom AI chip design? That's a different game entirely, and Europe is largely on the sidelines.

Bob: The risk is that as AI hardware becomes more specialized and more concentrated among a few key players, Europe's ability to build truly independent AI capabilities gets harder, not easier.

OpenAI is now writing policy papers — and this one tackles the big uncomfortable question about jobs.

Samantha: OpenAI has published what it calls "Industrial Policy for the Intelligence Age." It's a document that lays out proposals for managing the economic and social effects of superintelligence, especially on workers.

Bob: This is a notable shift in tone from OpenAI. They're not just talking about what AI can do — they're talking about what happens to people when AI does it. The document argues that workers should have a voice in how AI gets deployed and calls for significant investment to cushion the impact on employment, wages, and job quality.

Samantha: It's interesting timing, right? Because there's growing public anxiety about AI and jobs. And OpenAI — the company probably most associated with accelerating AI capabilities — is now saying, hey, we need to think seriously about the people side of this.

Bob: Whether you see that as genuine concern or strategic positioning ahead of regulation, the substance is worth engaging with. They're talking about prioritizing AI deployments that improve job quality rather than just replacing jobs outright.

Samantha: And in Europe, this feeds directly into existing policy debates. The EU AI Act is moving toward full implementation, and labor market impacts are a huge concern for European regulators. France, Germany, the Nordics — they all have strong labor protection traditions. This OpenAI document gives them a framework to react to, whether they agree with it or not.

Bob: European unions and worker organizations will be paying close attention too. The idea that workers should have a say in AI deployment aligns well with European co-determination models — think German works councils, for example. So this could actually find fertile ground in Europe.

Samantha: The key question is whether proposals like these translate into actual policy or remain just talking points. But the fact that a leading AI company is putting this on paper? That's significant.

While Europe pursues one big AI rulebook, American states are writing dozens of smaller ones.

Bob: Multiple US states are pushing forward with their own AI legislation, and the range is remarkable. South Carolina is tackling AI in therapy. Missouri is advancing its own AI bill. Oregon and Idaho have signed chatbot laws. Louisiana and New York are looking at AI in pricing and landlord-tenant relationships.

Samantha: It's a real patchwork. And for companies operating across multiple states, it's becoming a compliance headache. Every state has different rules, different definitions, different enforcement mechanisms.

Bob: Oregon's law is particularly notable because it includes a private right of action — meaning individuals can sue over violations. That's a significant enforcement tool that not all states are including.

Samantha: Now, the contrast with Europe is striking. The EU AI Act is one comprehensive framework for twenty-seven countries. Love it or hate it, it provides consistency. Companies know what the rules are across the entire single market.

Bob: In the US, you could theoretically comply with California's AI rules and still violate New York's. That kind of fragmentation creates real costs and uncertainty. And some argue it actually slows down responsible AI adoption because companies get paralyzed by conflicting requirements.

Samantha: European regulators can look at this and feel somewhat validated in their approach. A single, clear rulebook has advantages. But there's also something to learn from the US states — they're targeting very specific use cases that the EU Act might not address in as much detail. AI in therapy, AI in rent pricing — those are granular applications that matter to real people.

Bob: So it's not a simple story of one approach being better. But for European businesses expanding into the US, this fragmented landscape is something they need to take very seriously in their compliance planning.

A European tech champion is caught in the crossfire of US-China tensions — again.

Samantha: ASML shares dropped after reports emerged of new US legislation aimed at further restricting Chinese semiconductor manufacturers from buying ASML's equipment. This is the Dutch company that makes the machines that make the world's most advanced chips.

Bob: ASML is essentially irreplaceable. There is no alternative supplier for extreme ultraviolet lithography machines. If you want to make leading-edge semiconductors, you need ASML's gear. Period. And the US has been pressuring the Netherlands for years to restrict sales to China.

Samantha: What's new here is that we're talking about US legislation — not just diplomatic pressure or executive orders, but actual laws being written to tighten the screws further. That's an escalation.

Bob: For ASML, China has been a significant market. Losing more of that business hurts their revenue and growth projections. And it puts the Dutch government and the EU in an uncomfortable position — a European company's commercial decisions are effectively being dictated by American lawmakers.

Samantha: This is the technological sovereignty question in its most concrete form. ASML is arguably Europe's single most strategically important tech company. And yet its ability to sell to a major market is being constrained by a foreign government's geopolitical strategy.

Bob: European policymakers have been wrestling with this for a while. There are voices calling for the EU to develop its own export control framework rather than simply following Washington's lead. But in practice, pushing back against the US on something like China chip restrictions is politically very difficult.

Samantha: And ASML's employees and shareholders are the ones feeling it. The stock decline is real. The uncertainty around future revenue from China is real. This isn't abstract geopolitics — it's hitting a European company's bottom line.

Bob: It's a situation that isn't resolving anytime soon. If anything, expect more pressure, more restrictions, and more difficult conversations between Washington, The Hague, and Brussels.

Samantha: Today we covered: SK Hynix's massive memory chip supply deals, Samsung's record AI-fueled profits, Broadcom's growing AI hardware dominance, OpenAI's workforce policy blueprint, the patchwork of US state AI laws, and new US threats to ASML's China sales.

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.