6 oktober 2026 · 14:37
Defense Tech Funding Hits $40.8B: Stark + Reflection AI
Defense tech funding reached a record $40.8 billion in the first half of 2026, with AI and drones rewriting modern warfare and Germany's Stark raising €500 million. In this episode we explain what the surge means for European strategic autonomy and how Nvidia-backed Reflection AI plans to challenge DeepSeek and Qwen with a Western open-weight model. We also cover DayOne's Nasdaq IPO, OpenAI's ChatGPT image ads, and two free tools worth trying.
Beluister deze aflevering:
Transcript
Samantha: Welcome to The State of Tech, The European Edition, Tuesday October six, 2026. I'm Samantha Lawrence.
Bob: And I'm Bob Russell. Today: defense startups smash funding records as drones and AI reshape warfare, Singapore's DayOne data centers files for a Nasdaq listing to fuel AI expansion, OpenAI brings visual ads into ChatGPT image generation, Nvidia-backed Reflection AI readies an open-weight model to challenge Chinese rivals, and to close out two things you can actually use today: the free open-source illustration app Vectorpea, and the European journey planner Chronotrains. Let's start with defense.
Defense startups attract a record forty billion dollars as AI and drones rewrite modern warfare.
Samantha: Venture capital poured into defense startups at a pace nobody expected this year. In the first half of 2026 alone, forty point eight billion dollars went into the sector. That is a fifty-six percent jump from last year, and the highest level in a decade. The driver is simple: artificial intelligence and drones are rewriting how modern wars are fought, and investors want in on the companies building that next generation of hardware and software. The headline deal was Anduril Industries, which raised five billion dollars at a sixty-one billion dollar valuation.
Bob: And Europe is sitting right in the middle of this wave. Germany's Stark, a drone developer, secured five hundred million euros, which is one of the largest defense tech rounds the continent has seen. Japan is in there too, with Sakana AI raising thirty-one billion yen as it moves into defense work, including a pilot with Japan's Defense Ministry on AI-driven intelligence analysis. What stands out is the breadth. This is no longer a niche where a handful of US primes dominate. Capital is flowing to Berlin, Tokyo, London, Kyiv-adjacent suppliers, all at once.
Samantha: The war in Ukraine has clearly been a catalyst. Cheap drones, autonomous targeting, and AI-driven intelligence analysis have proven themselves on the battlefield, and NATO governments are now scrambling to catch up on procurement. For European startups, that means contracts that used to take years are moving in months. It also means a cultural shift: European venture funds that once avoided defense entirely are now writing cheques, often alongside sovereign wealth and government-backed vehicles.
Bob: There is a flipside though. A forty billion dollar half-year raises serious questions about valuations and about which of these companies will actually deliver fielded systems. Analysts point out that defense procurement cycles remain slow and politically sensitive, even when the technology is ready. For Europe specifically, this surge collides with the ongoing debate about strategic autonomy, how much of the continent's defense tech stack should be built at home versus bought from American suppliers. Stark's half-billion-euro round is a signal that Berlin is picking a side.
Singapore's DayOne Data Centers files for a Nasdaq IPO to fund global AI infrastructure build-out.
Samantha: Singapore-headquartered DayOne Data Centers has filed with the US Securities and Exchange Commission for an initial public offering of American Depositary Shares. The company plans to list on the Nasdaq Global Select Market under the ticker D-O-D-C. DayOne runs a digital infrastructure platform across ten markets in Asia Pacific and Europe, and its focus is on building next-generation, high-capacity campuses tuned specifically for artificial intelligence workloads. Morgan Stanley, J.P. Morgan, BofA Securities, and Citigroup are the underwriters.
Bob: The timing tells its own story. Hyperscalers and AI startups are desperate for power-dense capacity that can handle the heat and electricity demands of modern AI training. Traditional data centers were not built for racks pulling tens of kilowatts each. DayOne is positioning itself as one of the operators that can deliver that at scale, across multiple continents. Listing in New York rather than Hong Kong or Singapore also signals where the biggest pool of AI-focused capital still sits.
Samantha: For Europe, the European presence in those ten markets matters. More AI-ready capacity on the continent means European companies and research groups do not have to ship every workload to the US or Asia. That has implications for data residency, latency, and compliance with the AI Act and GDPR. The filing does not yet disclose which specific European locations are involved, but the direction is clear: a well-funded Asian operator is planting flags in the European infrastructure market, right as EU governments are talking about sovereign AI capacity.
Bob: The broader context is a global race to pour concrete and lay cable fast enough to keep up with AI demand. Every quarter brings new multi-billion dollar announcements from Amazon, Microsoft, Google, and now specialist operators like DayOne. The risk is familiar: if AI demand plateaus even briefly, the market could end up with significant overcapacity. But for now, the capital markets seem willing to fund the build-out, and DayOne's listing is the latest test of that appetite.
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.
OpenAI starts testing visual ads inside ChatGPT image generation, opening a new revenue chapter.
Bob: OpenAI confirmed it will begin testing visual advertisements inside the image generation flow in ChatGPT later this month, starting in the United States. The company says the ads will be clearly labeled and kept visually separate from the user-generated images themselves. This matters because ChatGPT now has one point two billion weekly users, and OpenAI has been under growing pressure to turn that reach into recurring revenue beyond subscriptions and API fees. Advertising is the obvious next lever.
Samantha: The format itself is interesting. Rather than slotting banner ads around the interface, OpenAI is embedding them in the creative moment, when a user is actively generating an image. That is a far more valuable attention window than a passive feed scroll, and advertisers will pay accordingly. For OpenAI, it is also a careful first step. Push too hard, and the clean, assistant-feel of ChatGPT gets diluted. Done well, it becomes a serious challenger to Google's search advertising business over the next few years.
Bob: For European users, the pilot is US-only for now, but the implications travel. European regulators have been watching how AI platforms handle advertising, personalisation, and children. The Digital Services Act already imposes transparency rules on large online platforms about who paid for an ad and why a user is seeing it. If and when visual ads reach European ChatGPT users, OpenAI will need to show that labeling, data usage, and targeting all meet those standards. Expect a conversation with Brussels before any wide rollout here.
Samantha: There is also a creator and brand-safety angle. Image generation already produces content that touches copyright, likeness, and trademark questions. Mixing paid promotion into that same flow raises obvious issues: what happens if an ad appears next to a generated image that parodies a brand, or features a public figure? OpenAI will need clear policies, and probably automated filters, to keep advertisers comfortable. The test run in the coming weeks will be the first real signal of how that balance plays out.
Nvidia-backed Reflection AI readies an open-weight model to challenge Chinese rivals DeepSeek and Qwen.
Bob: Reflection AI, backed by Nvidia, is preparing to release its first open-weight artificial intelligence model. The pitch is pointed: give Western enterprises a reliable, bank-approved foundation to build their own proprietary AI, instead of relying on leading Chinese open systems like DeepSeek and Qwen. Open-weight means companies can download the model, inspect it, fine-tune it on their own data, and run it on their own infrastructure, which is a very different proposition from calling an API hosted somewhere else.
Samantha: The context matters here. Chinese open-weight models have genuinely closed the gap with closed Western systems on many benchmarks, and they are often cheaper to run. That has created real anxiety in regulated industries, particularly banking and healthcare, where firms want the flexibility of open models but are nervous about compliance, provenance, and the geopolitical implications of building core systems on Chinese-origin AI. Reflection AI is explicitly targeting that gap, with Nvidia's backing giving it both capital and preferred access to compute.
Bob: For European enterprises, this is actually welcome news. Many European banks and insurers have been quietly experimenting with DeepSeek and Qwen because the price-performance is hard to argue with, but their legal teams have been uncomfortable. A Western open-weight alternative, with transparent training data policies and support for European compliance frameworks, could unlock a lot of stalled projects. It also fits with the EU's push for sovereign AI options that are neither American closed systems nor Chinese open ones.
Samantha: The release details, size, licence terms, and benchmark numbers, have not been fully disclosed yet, so judgment should wait until the model actually ships. The real test will be whether Reflection can match the raw performance of DeepSeek and Qwen while offering a licence enterprises can actually use commercially. If it can, this becomes a meaningful third pole in the open-weight world. If the licence is restrictive or the performance lags, it will struggle to displace the incumbents that have already built developer communities around them.
Vectorpea brings a full free vector illustration studio right into your browser, no install needed.
Bob: To close out, two things you can actually use today. First one is Vectorpea, from the same independent developer behind the browser-based image editor Photopea. Vectorpea is a free vector illustration tool that runs entirely in your browser. Think logos, icons, infographics, posters, the kind of work most people reach for Adobe Illustrator or Affinity Designer to do. The difference is you open a tab, and you are working. No install, no account required, no subscription.
Samantha: It opens and saves in the standard formats, including the native Illustrator file type, SVG, PDF, and EPS. For small businesses, students, teachers, and anyone who occasionally needs to tweak a logo or build a simple graphic, that compatibility matters. You can take a client file, open it in Vectorpea, make your edits, and export back without owning a paid licence. Reviewers describe the interface as deliberately familiar to anyone who has used Illustrator, which flattens the learning curve considerably.
Bob: There are trade-offs worth being honest about. Running everything in a browser tab means very large or complex files can slow down, and the tool does not match the full feature set of a decade-old desktop package. It is also ad-supported in the free version, with a paid tier available to remove them. But for the vast majority of everyday vector work, it is a genuinely capable option. European freelancers who do not want to commit to a monthly Adobe subscription have been quietly recommending it for a while now.
Samantha: Platform coverage is as broad as it gets, because it is a web app. It works on Windows, Mac, Linux, Chromebook, and even on a tablet in a pinch, as long as you have a modern browser. Files stay local to your device unless you choose to export them somewhere. For a casual user who needs vector editing a few times a month, it removes the entire friction of installing software. Open the tab, do the work, close the tab.
Chronotrains maps how far you can travel across Europe by train in just a few hours.
Bob: The second one is a quieter find, but a lovely one. Chronotrains is a free web map that shows how far you can travel across Europe by train from any given city, in one hour, two hours, three hours, up to eight. You click a city, and the map colours in everywhere you can reach in that time by rail. It is built on open timetable data from European rail operators, and it works instantly in any browser.
Samantha: The practical use is obvious if you are planning a holiday or a work trip and want to see what is realistically reachable without flying. Click Berlin, and you suddenly see how much of central Europe is within four hours by train. Click Paris, and the whole of northern France and large chunks of Belgium, the Netherlands and the UK light up. For anyone trying to cut flights for climate reasons, or just avoid airport misery, it is a far more intuitive way to think about European travel than a flight search engine.
Bob: It is free, requires no account, and works on desktop and mobile. The project is open-source, built by an independent developer who has made the code and the data pipeline public. That matters because it means the tool is not going to disappear behind a paywall overnight, and anyone can see how the travel times are calculated. For journalists, teachers, and urban planners it is also a useful visualisation of how connected, or disconnected, different parts of Europe actually are by rail.
Samantha: One caveat: Chronotrains shows how far you can get, not how much a ticket costs or whether a specific connection runs today. For that you still need a booking site. But as a planning starting point, it is genuinely useful. Open it this afternoon, click your home city, and you will probably discover somewhere reachable in three hours that you had never thought of as a weekend option.
Samantha: Today we covered: a record forty billion dollars pouring into defense startups as AI and drones reshape warfare, Singapore's DayOne Data Centers filing for a Nasdaq IPO to fund AI-ready infrastructure, OpenAI testing visual ads inside ChatGPT image generation, Nvidia-backed Reflection AI preparing an open-weight model to challenge DeepSeek and Qwen, and two things to try yourself: Vectorpea and Chronotrains.
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.