Meta $25B privacy verdict in New Mexico, NVIDIA's EU push

Samantha: Welcome to The State of Tech, The European Edition, Thursday October twenty-two, 2026. I'm Samantha Lawrence.

Bob: And I'm Bob Russell. Today: a twenty-five billion dollar privacy verdict against Meta, OpenAI's frontier models landing inside Google Cloud, NVIDIA pushing open AI families for a European sovereign stack, QuantumScape building solid-state batteries for AI data centres and defence, and two things you can actually use today, the free panorama tool Panorama Tools and the Czech mapping service Mapy dot C-Z. Let's get into it.

A New Mexico jury orders Meta to pay twenty-five billion dollars over misleading privacy practices.

Samantha: We start in the United States, where a New Mexico jury has ordered Meta to pay twenty-five billion dollars after finding the company misled users about how it handled their data. This is not a regulatory fine, this is a civil jury verdict under state consumer protection law. The size alone makes it one of the largest privacy-related penalties ever handed to a tech company, and it lands at a moment when state-level action in the US is becoming the real pressure point on Silicon Valley, not federal legislation.

Bob: And the mechanism matters. New Mexico argued that users were not given a fair picture of what Meta was doing with their information across Facebook, Instagram and WhatsApp. The jury agreed. Meta has already signalled it will appeal, and the number will almost certainly shrink or get restructured on appeal, that is how these cases usually play out. But even a heavily reduced figure sets a benchmark. For context, this dwarfs most GDPR fines Meta has absorbed in Europe over the past five years combined.

Samantha: What is interesting is the shift in who holds the stick. For years the assumption was that Brussels would be the toughest enforcer on data privacy, and Washington would stay quiet. Instead you now have individual US states running aggressive consumer protection cases, often with faster timelines than European procedures. That changes the calculation for every big platform. The legal risk map is no longer just the Irish Data Protection Commission and the European Data Protection Board.

Bob: For European listeners, there are two concrete threads worth watching. First, European regulators and class-action lawyers will study this verdict carefully. The theory of harm, that users were misled rather than merely unaware, maps neatly onto ongoing Digital Services Act and GDPR cases. Second, Meta's appeal strategy will likely argue the penalty is disproportionate, and that argument has a European echo too, where courts have repeatedly trimmed large data protection fines on proportionality grounds.

Samantha: The bottom line: a jury verdict is not a final judgment, but it reframes the conversation. Privacy accountability now has a dollar figure attached that even a company Meta's size cannot shrug off. And it tells European consumers that the gap between how their data is treated and how American users are treated is narrowing, at least in the courtroom.

OpenAI's frontier models, including GPT-6 Astra, arrive natively inside Google Cloud.

Bob: Our second story is a partnership that would have sounded unlikely a year ago. OpenAI is making its latest frontier models, including the flagship it calls GPT-6 Astra, available directly through Google Cloud's model marketplace. Google Cloud customers can now call OpenAI's models the same way they call Google's own Gemini models, with Google's data governance and infrastructure sitting underneath. For enterprises, this removes a lot of plumbing work.

Samantha: And it is a notable shift in posture. OpenAI's compute has historically been tied heavily to Microsoft's Azure cloud. Opening the models up through a direct competitor's cloud suggests OpenAI wants its models to be the default choice regardless of where a customer's data already lives. For Google, it is an admission that enterprise buyers want a menu, not a single-vendor stack. The two companies are rivals at the application layer and now partners at the infrastructure layer.

Bob: For European companies, this matters in a very practical way. Many large European firms, especially in banking, insurance and healthcare, have standardised on Google Cloud partly because of its data residency commitments inside the European Union. Until now, using OpenAI's top models in that environment meant extra legal reviews, custom integrations, or routing data outside the preferred cloud. That friction is largely gone, which should accelerate adoption of agent-based workflows across regulated industries.

Samantha: The caveat is sovereignty. European policymakers have spent two years arguing that depending on American frontier models is a strategic weakness, even when the hosting is European. This deal makes the American models more convenient, not less dominant. So while enterprise adoption gets easier, the political debate about building European alternatives gets harder. Expect this tension to come up again in the next NVIDIA story, where the sovereign question is answered very differently.

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 unveils new open AI model families at GTC Berlin to boost Europe's sovereign AI capabilities.

Bob: NVIDIA used its GTC conference in Berlin to launch a broad expansion of its open AI model families. The headline names are the Nemotron 3.5 Super VL and Nemotron 3.5 Bolt, but the real story is the range: vision, language, retrieval, tabular data, biology, robotics and quantum. These models are being released as open weights, meaning European developers can download them, fine-tune them on local data, and run them on their own hardware without sending anything to an American cloud.

Samantha: The timing is deliberate. Hosting the launch in Berlin and pitching it explicitly as support for European sovereign intelligence is a direct response to the political mood on the continent. European governments have been pushing for AI infrastructure that is not entirely dependent on American hyperscalers, and NVIDIA, which supplies the chips to those very hyperscalers, wants to be seen as a partner in that effort, not an obstacle.

Bob: What makes this practically useful is the breadth. A tabular data model, for example, is unglamorous but hugely important for European banks and insurers who work mostly with structured records, not chatbots. A biology model matters for the continent's strong pharmaceutical sector. And a robotics model lands at a moment when German and Nordic industrial firms are actively looking for foundational tools they can customise without licensing fees stacking up.

Samantha: The honest reading, though, is that open weights do not equal independence. These models are still optimised for NVIDIA hardware, and that hardware is still mostly American-designed and Taiwanese-manufactured. European sovereignty on AI remains a layered question, and this release addresses the model layer while leaving the chip layer untouched. Useful progress, not a complete answer.

QuantumScape plans a one hundred megawatt-hour solid-state battery line for AI data centres and defence.

Bob: Our fourth story moves away from AI models and into the hardware that keeps the lights on. QuantumScape, a solid-state lithium-metal battery company, has announced plans for a one hundred megawatt-hour production line. The buyers it is targeting are telling: AI data centres, aerospace, and defence. Solid-state batteries promise more energy in less space, faster charging and better safety than today's lithium-ion cells, and the company is aiming to begin cell production in 2028.

Samantha: The European angle sits in the partnership. QuantumScape is working with the German automation firm Muehlbauer Group to build the production line itself. That is a reminder that even when the battery chemistry is American, the industrial equipment that actually turns a lab result into a factory line is often European. Germany in particular has a deep bench of specialist machine-builders whose names rarely make headlines but who show up in every serious manufacturing announcement.

Samantha: And the demand side is worth unpacking. AI data centres need back-up power that can bridge grid interruptions without catching fire, and today's lithium-ion packs are increasingly seen as a risk at the scale these facilities are reaching. Defence buyers want lighter, denser batteries for drones and portable systems. Both markets are willing to pay a premium, which is why QuantumScape is starting there rather than chasing the brutal economics of electric cars.

Bob: The caution is simple: 2028 is still two years away, and solid-state battery companies have a long history of slipped timelines. One hundred megawatt-hours is also a pilot scale, not a mass-market one. But the direction of travel is clear. The AI boom is pulling energy storage along with it, and European industrial firms are quietly becoming essential suppliers to that build-out, even when the brand on the finished product is American.

Panorama Tools is the free open-source toolkit behind many panorama stitchers, and it is quietly doing the heavy lifting on your own computer.

Samantha: To close out, two things you can actually use today. The first is a project many photographers rely on without ever noticing it. It is called Panorama Tools, and it is a free open-source collection of libraries and command-line programs for stitching and correcting photos. It runs on Windows, Mac and Linux. You give it a handful of overlapping photos, say from a mountaintop or a city square, and it helps turn them into a single wide image.

Bob: What makes it worth mentioning now is that modern phones have made panoramas feel disposable, a quick swipe that often produces ghosting and warped horizons. Panorama Tools takes the opposite approach. It is built around proper geometric alignment and lens correction, which is why several well-known stitching programs use it as their engine. The result can look like something a careful photographer took, not a phone sweep. And because it runs entirely on your own machine, your pictures never leave your laptop.

Samantha: The learning curve is honest: this is not a polished one-click app. Much of the work happens through text project files and commands, so the first time you open it you will see more options than you expected. But that is the trade-off for control. For a first attempt, you can start with a graphical front end that sits on top of it, and explore the deeper controls later, or never.

Bob: Practical details: Panorama Tools is completely free, there is no account, no subscription, no upload step. It is maintained by a community of volunteers and has been around for many years, which is unusual stability for an open-source project. If you have a folder of holiday photos from this summer sitting unused, this is a Saturday afternoon well spent, and you will learn how panoramas actually work along the way.

Samantha: And the broader point: not every useful tool needs AI stapled to it. Panorama Tools solves one problem well and gets out of the way. In a week where we have talked about frontier models and sovereign AI, there is something refreshing about a piece of software that just stitches your pictures together and asks for nothing in return.

Mapillary gives you free, crowdsourced street-level photos of European streets, trails and cycle paths, so you can scout a place before you set off.

Bob: Our second find is a street-level imagery service that quietly does something Google Street View often cannot on the kind of trips Europeans actually take at this time of year. It is called Mapillary, it is free to use in any browser, and there are free apps for iOS and Android. The headline strength is that the photos come from ordinary people: walkers, cyclists and drivers who capture images of the road, the path or the trail with a phone and upload them to a shared map.

Samantha: The reason it works so well for European users is that the coverage goes where the big commercial cars do not. Think forest tracks, riverside cycle paths, mountain lanes and narrow village streets, often photographed far more recently than the official imagery. The photos are also released under an open licence, and the service has long been tied closely to the OpenStreetMap community, which uses the images to check and improve the map.

Bob: The practical bit for today: you can open the map, zoom to the route you plan to take, and click through the photo sequences as if you were already there. You can see whether a path is paved or gravel, whether a junction is confusing, or whether a bridge is really a bridge. It is free to browse, and if you want to contribute, you can record your own weekend walk or ride with your phone and upload it for others.

Samantha: One honest limitation: coverage is patchy, because it depends on volunteers. Some regions are dense with imagery, others have almost nothing, and the age of the photos varies from street to street. Mapillary is not a navigation app and will not route you anywhere. It is a scouting tool, the one you open to look before you leap, which is a different job. If your October weekends involve a trail or a cycle route you have never tried, it is worth having a look before Saturday.

Bob: Today we covered a twenty-five billion dollar privacy verdict against Meta, OpenAI's frontier models arriving on Google Cloud, NVIDIA's open AI families for European sovereignty, QuantumScape's solid-state battery line for AI data centres, and two things to try yourself: the free panorama stitcher Panorama Tools and the Czech mapping service Mapy dot C-Z.

Samantha: Today we covered a twenty-five billion dollar jury verdict against Meta, OpenAI's models landing on Google Cloud, NVIDIA's open model families in Berlin, QuantumScape's solid-state battery plans, and two things to try yourself: Panorama Tools and Mapy dot C-Z.

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.