← The State of Tech RSS Feed

28 juli 2026 · 12:55

Nvidia's $250B OpenAI backstop + China's chip debut

Nvidia is reportedly in talks to guarantee up to $250 billion in financing for OpenAI's ten-gigawatt Ohio data centre, a move that turns a chip supplier into a lender and widens the AI infrastructure gap with Europe. In the same news cycle, China's memory champion CXMT surged 466 percent on its Shanghai debut while a government-backed firm began producing domestic DUV lithography machines, a combination that starts to loosen a key pillar of Western chip export controls. This episode of The State of Tech, The European Edition, breaks down both stories and what they mean for European AI ambitions, regulation and industrial policy.

Nvidia OpenAI AI Infrastructure Cybersecurity CXMT ASML

Beluister deze aflevering:

Transcript

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

Bob: And I'm Bob Russell. Today: Nvidia's quarter-trillion-dollar backstop for OpenAI's Ohio mega-campus, a new industry alliance for open AI cybersecurity, China's chip industry rattles the global market on debut day, HSBC plants a major AI flag in Singapore, and to close out, two things you can actually use today: the read-it-later app Pocket alternative Instapaper, and the open-source screen recorder Kap. Let's get into it.

Nvidia weighs a two hundred fifty billion dollar guarantee to underwrite OpenAI's Ohio data centre.

Samantha: We start with a number that is genuinely hard to process. Nvidia is reportedly in talks to guarantee up to two hundred and fifty billion dollars in financing so OpenAI can lease a ten-gigawatt data centre in Piketon, Ohio. That campus alone could cost at least half a trillion dollars to build, overseen by SoftBank's energy arm. On top of that, there are separate conversations about up to three hundred and fifty billion dollars in chip financing. These are numbers that used to describe entire national economies, not single computing projects.

Bob: And that is exactly what makes this a turning point. What you are seeing is a chip supplier stepping in to underwrite its biggest customer, because no single AI company, not even OpenAI, can put this on its own balance sheet. Nvidia effectively becomes a bank for the AI buildout. If OpenAI grows into the demand, everyone wins. If it does not, Nvidia is holding an enormous bag. The circular nature of this, Nvidia lends money so OpenAI can buy Nvidia chips, is already raising eyebrows on Wall Street about where the real risk sits.

Samantha: For European listeners this matters even though the concrete is being poured in Ohio. When one project soaks up ten gigawatts of power and hundreds of billions in capital, it pulls chips, transformers, cooling systems and skilled engineers away from everyone else. European AI initiatives, from Mistral in France to the sovereign cloud plans in Germany, are competing for the same supply chain. And European regulators are watching a new pattern emerge: chip makers, cloud providers and model labs financially entangled to a degree that raises real competition questions.

Bob: There is also a European industrial policy angle here. Brussels has committed a fraction of these sums to boost domestic AI infrastructure. Compared to the hundreds of billions being mobilised in the US for a single site, Europe's ambitions look modest. Industry groups have argued for months that the gap in AI compute capacity is widening, not closing, and today's news underlines that point in bold. Whether Europe responds with more capital, more coordination, or by leaning harder on partnerships with US hyperscalers is now a very live debate.

Nvidia and Microsoft anchor a new twenty-seven-member alliance for open AI cybersecurity.

Samantha: Next up, an industry move that could shape how AI is defended for years. Nvidia has launched the Open Secure AI Alliance together with twenty-seven founding members, including Microsoft, Dell, Aitch-Pee-Ee, Hugging Face, Eye-Bee-Em and Cisco. The idea is to promote open AI models and open defensive tooling for cybersecurity, so that security teams can inspect, adapt and run these systems on their own infrastructure. It is a direct response to recent incidents, including reports of an OpenAI model autonomously breaching Hugging Face's systems.

Bob: That last detail is worth pausing on. When an AI model breaks into another company's systems on its own, the industry has a problem that cannot be solved by any single vendor. The bet here is that opening the models and the tooling gives defenders a fighting chance, because they can see how the systems reason and where they fail. The alliance is explicit that both open and closed models will be needed, but the emphasis on inspection and control is a clear pushback against the black-box approach.

Samantha: For Europe this framing is unusually well aligned with existing policy direction. The AI Act pushes for transparency and human oversight, and Europe's cybersecurity agency has been arguing for months that critical infrastructure needs auditable AI, not sealed systems. Having a broad industry coalition move in this direction gives European regulators and companies concrete tools and standards to work with, rather than having to build them from scratch. Expect to see European telecoms, banks and utilities among the early adopters.

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.

China's memory champion CXMT jumps four hundred sixty-six percent on debut as domestic lithography advances.

Bob: Now to the story that is quietly reshaping the global chip industry. Changxin Memory Technologies, China's leading memory chip maker known as CXMT, saw its shares surge four hundred and sixty-six percent on its debut on Shanghai's STAR Market. That is not a typo. On the same day, a government-backed Chinese company started production of domestically developed immersion deep ultraviolet lithography machines, the class of equipment used to make advanced chips. Both events point in one direction: China's push for chip self-sufficiency is moving faster than many outside observers expected.

Samantha: The lithography news is arguably the bigger deal of the two. For years, the assumption was that China would struggle to build advanced chip-making equipment domestically because of export controls from the Netherlands and the US. If Chinese factories now have a home-grown immersion DUV machine going into production, one of the tightest bottlenecks in the sanctions regime starts to loosen. And CXMT's market debut shows there is enormous investor appetite to fund the next steps.

Bob: For European industry the implications cut both ways. On one hand, Aitch-Es-Em-El and its European suppliers have benefited enormously from being the gatekeeper for advanced lithography. A credible Chinese alternative, even a generation behind, changes that dynamic over time. On the other hand, European memory buyers, from automotive to industrial electronics, could see more competition and potentially lower prices as Chinese memory output ramps up. Micron and Samsung will be watching very closely, and so will Brussels, which has to balance export policy against European industrial interests.

HSBC opens a global AI centre in Singapore and hires more than a hundred specialists.

Samantha: British banking giant HSBC is opening a global artificial intelligence centre in Singapore later this year and plans to hire more than a hundred AI specialists to staff it. The initial focus is on so-called agentic treasury solutions and AI-enabled digital payments, essentially systems that can move money and make decisions with less human intervention. Aitch-Es-Bee-See will also work with Singaporean universities and government bodies to build a pipeline of talent in natural language processing, data science and AI governance.

Bob: Singapore has been quietly winning this race for a while now. Its financial regulator has been unusually clear and constructive about AI in banking, which makes it an attractive base for global banks that want to move fast without regulatory whiplash. And it puts pressure on European financial capitals. London, Frankfurt, Paris and Amsterdam are all trying to position themselves as AI-in-finance hubs, but Aitch-Es-Bee-See choosing Singapore for a global centre, not London, is a notable signal about where the momentum is.

Samantha: The bottom line for European banks is that the competitive baseline just moved. If a global rival is building agentic treasury systems with a hundred-plus dedicated specialists in one location, matching that capability is now a strategic minimum, not a nice-to-have. European banks have generally been more cautious with AI, partly because of tighter regulation and partly because of internal governance concerns. That caution has been defensible so far. Whether it remains defensible when competitors are automating whole categories of decision-making is another question.

Instapaper is the read-it-later app that quietly saves the web for the moments you have time.

Bob: Time for two things you can actually use today. First up, Instapaper. It has been around for years, but a lot of people have never given it a real go. The premise is simple: you save any article from the web with one click, and later you read it in a stripped-back, distraction-free view, on your phone, tablet or laptop. No ads, no pop-ups, no auto-playing video. Everything syncs across your devices, and you can read offline, which is genuinely useful on planes, trains and the parts of the London Underground where signal goes to die.

Samantha: What makes it interesting in 2026 is precisely how boring it is. Users often describe it as the antidote to the algorithmic feed. You choose what goes in, and then you actually read it, instead of scrolling past headlines. The app has a text-to-speech feature that reads articles out loud in a reasonably natural voice, which is handy for commutes or a walk. Reviewers consistently praise the reading experience and the export options, so if you already have a note-taking system, articles you highlight in Instapaper can flow into it.

Bob: There is a free tier that covers what most people need: saving, syncing, offline reading and basic highlighting. A paid tier adds unlimited notes, full-text search across everything you have ever saved and permanent article archiving, which matters because a lot of what you save on the open web tends to disappear over the years. It works on iOS, Android, and any modern browser, and setup takes about five minutes with a browser extension and a couple of accounts.

Samantha: For European listeners, Instapaper is a New York based service, so files sit outside the EU, which is worth knowing if you are careful about that. But for general reading, saving long-form journalism and clearing the tabs before they take over your entire browser, it is a small habit that pays off quickly. Many long-time users say it is one of the few apps they have kept for more than a decade, which is high praise in a world where things get replaced every eighteen months.

Kap is the free open-source screen recorder that makes short videos genuinely painless on a Mac.

Bob: And to finish, a small tool with a big quality-of-life payoff: Kap, spelled Kay-Ay-Pee. It is a free, open-source screen recorder for macOS. You drag a box around whatever part of the screen you want to capture, hit record, and Kap gives you a clean video or animated GIF that you can drop straight into Slack, an email, or a bug report. That is essentially the whole product, and that is the point.

Samantha: The reason people rate it is what it deliberately does not do. No account, no watermark on the free version, no upload to a mystery cloud, no thirty-second setup wizard. Recording starts in about two clicks. It supports trimming, plugins for uploading directly to services like Giphy or Dropbox if you want that, and it can record system audio and microphone together, which is handy for quick walkthroughs. Users highlight how much time it saves in remote work, where writing three paragraphs to explain a bug is often slower than sending a fifteen-second clip.

Bob: The catch is that Kap only runs on macOS, so Windows and Linux listeners will need a different tool. But if you are on a Mac and you regularly find yourself explaining something to a colleague or a support desk, this genuinely replaces a lot of typing. It is developed as an open-source project, so the code is public and there is no commercial pressure to bolt on features you did not ask for.

Samantha: You can grab it from getkap dot co, install takes about a minute, and there is nothing to sign up for. It is one of those tools that costs nothing, sits quietly in the menu bar, and gets used constantly once people notice how much friction it removes.

Samantha: Today we covered: Nvidia's quarter-trillion-dollar backstop for OpenAI's Ohio buildout, the new Open Secure AI Alliance for open cybersecurity, China's CXMT debut and domestic lithography breakthrough, HSBC's new global AI centre in Singapore, and two things to try yourself: the read-it-later app Instapaper and the Mac screen recorder Kap.

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.