19 juli 2026 · 14:24
Oracle $500B Stargate Bet, 30K Jobs Cut & AI Fragmentation
Oracle has placed one of the largest single bets in corporate history, committing $500 billion to the Stargate AI infrastructure partnership while cutting up to 30,000 jobs, making the trade-off between human workforce and compute brutally explicit. In the same news cycle, Microsoft launched Project Perception, a multi-model AI security platform designed to undercut Anthropic's Mythos on price, while Xi Jinping used Anthropic's own market struggles to publicly challenge US export controls. This episode unpacks what all three moves mean for European professionals, enterprises and regulators.
Beluister deze aflevering:
Transcript
Samantha: Welcome to The State of Tech, The European Edition, Sunday July nineteen, 2026. I'm Samantha Lawrence.
Bob: And I'm Bob Russell. Today: Oracle bets half a trillion dollars on AI infrastructure and axes thirty thousand jobs, Anthropic tweaks its flagship model as Xi Jinping fires back at US export controls, Microsoft launches a new AI security platform to challenge the market leaders, Vietnam's Da Nang partners with global cloud giants to build an AI hub, and to close, two things you can actually try today, the free travel app Rome2Rio dot com and the encrypted German cloud storage service Filen. Let's start with Oracle.
Oracle commits five hundred billion dollars to Stargate AI infrastructure while cutting thirty thousand jobs.
Samantha: Oracle just made one of the biggest single bets in corporate history. The company is putting five hundred billion dollars into the Stargate AI infrastructure partnership, and to help pay for it, up to thirty thousand jobs are going. That is a stunning trade-off, human workforce for silicon and servers, spelled out on the balance sheet in a way we rarely see this openly.
Bob: And Stargate itself is worth pausing on. It is the mega project tied up with OpenAI and SoftBank aimed at building the compute backbone for the next generation of AI. Oracle wants to be the plumbing underneath all of that, the data centres, the networking, the raw capacity. Five hundred billion is not a rounded-up marketing figure, it is a commitment on the scale of a small country's annual budget, and it tells you where Oracle thinks the money will be made over the next decade.
Samantha: The layoffs are what will sting. Thirty thousand roles is roughly a fifth of Oracle's workforce, and the company is essentially saying those jobs are less valuable than the machines that will replace them. Analysts have been warning for months that AI would reshape white collar work, but usually the story is about productivity gains. Here it is bluntly about funding.
Bob: For Europe, this matters on two levels. First, Oracle is a major cloud provider for European governments and enterprises, so the shape of its investment affects what services show up here, and at what price. Second, it sets a template. If Oracle can openly finance AI hardware by shrinking headcount, expect SAP, Capgemini, Atos and others to face pressure from investors to do the same. European works councils and unions will be watching this very closely, because the model of trading people for GPUs is now officially on the table.
Samantha: And there is a broader market signal here too. Investors are already nervous about whether all this AI spending will ever pay back. Oracle just doubled down anyway, which either looks visionary in five years or reckless. There is very little middle ground with a bet this size.
Anthropic updates Claude Fable 5 as Xi Jinping slams US export controls squeezing tech firms.
Bob: Anthropic is rolling out changes to its Claude Fable 5 model starting July twentieth, and the timing is loaded. It coincides with Chinese President Xi Jinping publicly criticising Washington for using export restrictions as a security tool. Xi specifically pointed out that these blockades have forced US firms, Anthropic among them, to pull software from various overseas markets.
Samantha: What is striking is that Xi is naming individual American companies as collateral damage in his own argument. He is essentially saying, look, your own tech giants are losing revenue and market share because of your government's policies. It is a rhetorical move designed to drive a wedge between US industry and US policymakers, and it lands at a moment when American AI firms are under real commercial pressure abroad.
Bob: The Claude Fable 5 changes themselves have not been detailed in full, but the context matters. Anthropic has been navigating a patchwork of restrictions on where and how its models can be deployed. Every tweak now has to account for which markets are open, which are closed, and which are actively contested. That is a very different operating environment than even two years ago.
Samantha: For European users and developers, the practical concern is fragmentation. If Claude behaves differently in Frankfurt than in Singapore than in San Francisco, that is a headache for anyone building products on top of it. European regulators, meanwhile, are watching to see whether US export policy ends up shaping what Europeans can actually access. The EU has its own AI Act to enforce, but it does not control what Washington restricts.
Bob: The bigger picture is that the AI industry is being pulled apart along geopolitical fault lines. Companies that a few years ago pitched themselves as global platforms are increasingly running regional variants, with different capabilities and different customers. That fragmentation is expensive, and eventually someone pays for it. Usually the customer.
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.
Microsoft unveils Project Perception, a multi-model AI security tool to challenge Anthropic's Mythos.
Bob: Microsoft has launched Project Perception, a new AI security platform designed to detect and respond to cybersecurity threats. What is interesting is the design. Instead of relying on a single model, Project Perception pulls together AI systems from Anthropic, OpenAI, and Microsoft's own labs, then routes threat detection through whichever performs best for a given task. Microsoft is pitching it as cheaper than existing rivals, most notably Anthropic's Mythos, and better placed globally thanks to its own distribution.
Samantha: The multi-model approach is a real shift. For years, the assumption was that each cloud provider would push its own AI as the single answer. Microsoft is now saying, actually, mixing and matching gives you better results, even if it means using a competitor's technology inside your own product. That is a pragmatic admission, and it also lets Microsoft hedge if one supplier stumbles or gets caught up in regulatory trouble.
Bob: Cybersecurity is a natural place for this. Attackers do not care which model detects them, they just want to get in. Defenders increasingly need every tool available, because AI-generated phishing, deepfake voice calls and automated intrusion attempts are all rising sharply. A platform that can throw multiple AI systems at a problem in parallel has a genuine edge.
Samantha: For European companies, particularly mid-sized firms that cannot afford elite security teams, this is potentially significant. If Microsoft delivers on the cost promise and the global availability, Project Perception could put enterprise-grade AI defence within reach of organisations that currently rely on much thinner protection. That said, European buyers will scrutinise where data is processed and stored, because routing threat intelligence through American AI providers still raises compliance questions under EU data rules.
Bob: And there is a competitive angle. Anthropic's Mythos has been the reference product in this space. Microsoft coming in with a bundled, multi-model alternative at a lower price forces Anthropic to respond, either on price or on capability. Either way, customers benefit. This is the kind of competition the AI security market has needed.
Vietnam's Da Nang teams with AWS and Renova Cloud to build a regional AI hub.
Samantha: The central Vietnamese city of Da Nang is positioning itself as an AI infrastructure hub, working with global partners including Amazon's cloud arm and Renova Cloud. The plans were unveiled at the Frontier AI and Local Zones Innovation 2026 conference, and the focus is on cloud capacity, low-latency AI applications, and meeting local data residency requirements.
Bob: The data residency point is the important one. A lot of countries are realising that hosting AI workloads locally is not just a technical preference, it is an economic and political priority. If your citizens' data has to stay inside the country, you need physical infrastructure inside the country. Da Nang is signing up global partners to build that, rather than trying to do it alone.
Samantha: It is also a signal about where the next wave of AI adoption is happening. We spend a lot of time talking about San Francisco, Shenzhen, and increasingly Riyadh, but a mid-sized Vietnamese city partnering directly with Amazon Web Services shows how quickly the geography is spreading. Southeast Asia is becoming a serious market, not just a manufacturing base.
Bob: For European infrastructure providers, this is a wake-up call of sorts. Europe has its own sovereign cloud ambitions, projects like GAIA-X, but progress has been slow and fragmented. Meanwhile a single Vietnamese city is stitching together international partnerships and moving quickly. The lesson is that speed and pragmatism matter as much as principle. Europe risks talking about digital sovereignty while others build it.
Samantha: The other thing worth noting is that AWS is the anchor partner. That means American cloud infrastructure is deepening its footprint in Vietnam at exactly the moment US-China tensions are pushing companies to diversify away from China. Da Nang is positioning itself as an alternative location for the sensitive AI workloads companies want kept away from Beijing's reach, and that has geopolitical weight beyond the local economy.
Bob: If this works, expect other Southeast Asian cities to follow the same playbook. Bangkok, Jakarta, Manila, all could pitch themselves as regional AI zones. The map of global compute is being redrawn faster than most people realise.
Rome2Rio dot com maps any journey across Europe in seconds, without an account.
Samantha: To close out, two things you can actually use today. The first is Rome2Rio dot com, a free travel planning website that a lot of European travellers swear by. You type in where you are and where you want to go, and it shows you every realistic way to get there, by train, bus, ferry, flight, or car, with estimated times and costs for each. No account needed, and it works in a browser or as an app.
Bob: What makes it useful is the breadth. Most travel sites specialise, they either do flights, or trains, or driving directions. Rome2Rio combines all of them, which matters in Europe where a trip from say Lyon to Ljubljana might involve a train, a bus, and a short flight, and no single national service will show you the full picture. It also handles the awkward cross-border journeys that Google Maps often stumbles on.
Samantha: For summer travel, that is genuinely helpful. You can quickly compare whether the sleeper train from Brussels to Vienna beats flying once you factor in time to the airport, and see costs side by side. Reviewers particularly praise the way it surfaces options you would not have thought of, like an overnight coach or a regional ferry.
Bob: The company behind it is Australian, but the European coverage is excellent, likely because European travellers are the biggest users. The website is free, with optional booking links that generate revenue if you actually book through them. There is no signup wall, no forced newsletter, you just start planning.
Samantha: One tip from experienced users, use it early in your planning to compare options, then book directly with the operator once you have decided. That way you get the best of both, the comparison view and the direct customer relationship if something goes wrong.
Bob: It also works well as a curiosity tool. Type in two random European cities and see what pops up, sometimes you discover a direct connection you never knew existed. For the price of nothing, it is one of the more quietly useful travel tools out there.
Filen offers encrypted cloud storage from Germany, keeping your files private and inside Europe.
Samantha: The second is Filen, a German cloud storage service that keeps your files fully encrypted and hosted inside Europe. Think Dropbox or Google Drive, but with end-to-end encryption meaning even Filen itself cannot read what you upload. That is a meaningful difference if you care about who might rummage through your data.
Bob: Filen is based in Germany, which puts it under strict European data protection rules, and the servers are located inside the EU. For anyone who has been quietly worried about storing sensitive documents, tax records, family photos, or client files on American services, this is a genuine European alternative. It offers apps for Windows, Mac, Linux, iOS and Android, so it slots into most workflows.
Samantha: The free tier gives you ten gigabytes, which is enough to try it out with your most important documents. Paid plans scale up from there, and users report the pricing is competitive with the big American services. The interface is clean, upload speeds are decent, and file sharing works with encrypted links you can set to expire.
Bob: The encryption model is the key selling point. Even if a court order landed on Filen's desk, the company genuinely cannot hand over readable files, because it does not hold the keys. That is a different trust model than most cloud services, where the provider technically can decrypt your data if asked.
Samantha: A word of caution, though. Because Filen cannot read your files, if you lose your password and recovery key, your data is gone. That is the trade-off with end-to-end encryption everywhere. Users recommend saving the recovery key somewhere safe, not just in the same place as the password.
Bob: For freelancers, small businesses, or just privacy-conscious individuals, it is worth a look. You can start with the free tier this afternoon and see if it fits your workflow.
Samantha: Today we covered: Oracle's five hundred billion dollar AI infrastructure bet with thirty thousand job cuts, Anthropic's Claude Fable 5 changes against the backdrop of Xi Jinping's export control criticism, Microsoft's new multi-model security tool Project Perception, Da Nang emerging as a Southeast Asian AI hub with AWS and Renova Cloud, and two things to try yourself: the free travel planner Rome2Rio dot com and the encrypted German cloud storage service Filen.
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.