22 september 2026 · 14:55
AMD nears trillion-dollar value as AI chip demand soars
AMD is knocking on the door of a trillion-dollar market valuation after raising AI chip prices by ten percent, a clear sign that accelerator demand driven by AI agents is still running hot. In the same episode, Belgian firm Aikido launches Altar, an open-weight AI security model companies can run entirely on their own infrastructure, a direct answer to European GDPR and NIS2 obligations. We also cover the Gates Foundation's $1B AI equity commitment, the new Petal petabit transatlantic subsea cable, and two free tools worth trying today.
Beluister deze aflevering:
Transcript
Samantha: Welcome to The State of Tech, The European Edition, Tuesday September twenty-two, 2026. I'm Samantha Lawrence.
Bob: And I'm Bob Russell. Today: AMD closes in on a trillion-dollar valuation as AI chip demand explodes, the Gates Foundation pledges a billion dollars to stop AI widening the inequality gap, Meta teams up with NEC and Sumitomo Electric on a new transatlantic subsea cable called Petal, a Belgian cybersecurity firm launches a locally-run AI model called Altar, and to close out, two things you can actually try today, the offline maps app Organic Maps and the open-source subtitle editor Subtitle Edit. Let's start with AMD.
AMD closes in on a trillion-dollar market value as AI chip demand keeps accelerating.
Samantha: AMD is on the verge of joining the trillion-dollar club. Shares of the chipmaker jumped on Monday, pushing its market value close to that milestone for the first time. The trigger is a mix of things: renewed investor optimism about the whole AI chip cycle, and reports that AMD has raised prices on its AI accelerator chips and other products by around ten percent. Intel and Arm Holdings also climbed on the same wave. And yes, this is a US session move, so we're talking about Monday's close, not today.
Bob: What's interesting is where the demand is actually coming from. It's no longer just about training ever-bigger models in a handful of hyperscaler data centres. The reports point to a broader shift toward AI agents running on people's behalf, doing tasks, calling other tools, checking things in the background. Meta's Muse AI agent is named specifically as a driver. All that agent activity means more inference workloads, more orchestration, and that pushes demand for both accelerators and ordinary processors. Which is why Intel got dragged along in AMD's slipstream.
Samantha: And a ten percent price hike is not a trivial detail. AMD is essentially telling the market, we have pricing power, customers will pay, and we're comfortable saying so out loud. That is the kind of signal investors love, because it flows straight into margins. Nvidia has done this dance for a while now, but AMD joining in changes the story from one dominant supplier to a genuine two-horse race at the top of the AI chip stack. It also suggests supply is still tight relative to what buyers want.
Bob: For Europe, this matters on two levels. The obvious one is that European companies, from carmakers to banks to research labs, are ultimately paying those higher prices for AI capacity, either directly or through their cloud bills. The less obvious one is strategic. Europe has been talking for years about semiconductor sovereignty, its own Chips Act, funding for local fabs. A world where two American designers can casually raise AI chip prices ten percent because demand is that hot puts real pressure on that ambition. It also makes the argument for European alternatives less abstract and more urgent.
Samantha: The bottom line is that the AI hardware boom is not cooling off, it's broadening out. And every European boardroom planning an AI budget for next year just watched their input costs go up.
The Gates Foundation pledges one billion dollars to stop AI from deepening global inequality.
Bob: The Bill and Melinda Gates Foundation has committed one billion US dollars over the next two years to make sure AI helps address social inequities rather than widening them. The announcement came in the foundation's annual Goalkeepers report. Bill Gates himself framed it in fairly stark terms, saying AI could turn out to be either the greatest equalizer humanity has ever built, or the worst source of injustice. The money will go to AI projects worldwide, including training models on local languages, health tools, education, and support for small-scale farmers.
Samantha: The local languages piece is the part that deserves more attention. Most of today's leading AI systems work brilliantly in English, reasonably well in a handful of major European languages, and then fall off a cliff. For hundreds of millions of people, that means the AI revolution basically speaks a language they don't. Funding models trained on underrepresented languages is not charity in the old-fashioned sense, it's infrastructure. Without it, entire regions get locked out of tools that are quickly becoming as basic as electricity or a mobile signal.
Bob: One billion dollars sounds enormous, and in philanthropic terms it is. But it's worth putting next to the numbers we just discussed in the previous story. AMD alone is knocking on the door of a trillion-dollar valuation on the back of AI demand. Individual data centre projects run into tens of billions. So the Gates commitment is significant as a signal and as seed capital, but it is clearly not going to counterbalance the commercial AI economy on its own. What it can do is shift priorities, fund pilots, and pressure the big labs to take non-English, non-Western use cases seriously.
Samantha: For European policymakers, this pledge lands in an interesting moment. Brussels is already deep into implementing the AI Act, with its focus on risk, transparency, and rights. Gates is essentially pushing a parallel question: not just is this AI safe, but is it fair, does it reach everyone. European development agencies and universities are likely partners here, especially on health and education pilots in Africa and South Asia. And it gives European tech firms working on multilingual or low-resource-language AI a stronger case for funding.
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.
Meta, NEC and Sumitomo Electric start work on the world's first commercial petabit subsea cable between North America and Europe.
Bob: Meta, NEC Corporation and Sumitomo Electric Industries have announced they're jointly building a new transatlantic subsea cable system called Petal. It's being pitched as the world's first commercial petabit optical submarine cable. The route will connect North America and Europe, linking major digital hubs on both sides of the Atlantic. The three companies say the goal is to meet the surge in communication traffic and to add more resilient, reliable capacity to a route that is already carrying an enormous share of the world's data.
Samantha: A petabit per second is a genuinely large number. To put it in plain terms, we're talking about a single cable capable of carrying far more traffic than the transatlantic links built even a few years ago. That extra headroom matters because AI is changing the shape of internet traffic. Training data moves between continents, AI services in the US answer questions from users in Europe, and increasingly, agents chat with other agents across borders. Old capacity assumptions simply don't hold anymore.
Bob: There's also a resilience angle that European regulators have become very focused on. Subsea cables have been in the news for the wrong reasons over the past couple of years, from suspected sabotage in the Baltic to accidental anchor damage. Adding a new high-capacity route between North America and Europe reduces the risk that a single incident takes out a big chunk of transatlantic traffic. The involvement of NEC and Sumitomo Electric on the engineering side also matters, because Japan has long been one of the few places in the world with deep subsea cable manufacturing know-how.
Samantha: For European users and businesses, Petal is the kind of infrastructure you never see and never think about, until it isn't there. It should mean lower latency for cloud services, more room for video and AI traffic, and a more robust backbone connecting European data centres to the American AI economy. It also means Meta continues to behave less like a social media company and more like a telecom operator, quietly building the physical layer of the internet it depends on.
Belgian cybersecurity firm Aikido launches Altar, an AI security model companies can run entirely on their own machines.
Bob: Belgian cybersecurity company Aikido has released Altar, an open-weight AI model built specifically for cybersecurity work. The launch was announced on Monday. The idea behind Altar is that companies can use advanced AI for threat detection and response without sending their source code or sensitive logs to an external cloud provider. Everything runs locally, on infrastructure the customer controls. That's a fairly pointed pitch aimed at organisations that have been nervous about pasting their crown jewels into someone else's chatbot.
Samantha: The context here is that a lot of AI-powered security tools on the market today are cloud-only. You send code, alerts, or network traffic patterns to a vendor, their model analyses it, and sends back suggestions. That works, but it creates two problems. First, you're handing potentially confidential information to a third party. Second, if that third party gets breached, your data goes with it. An open-weight model you can host yourself sidesteps both issues, at the cost of needing more in-house expertise to run it.
Bob: Open-weight is worth clarifying, because it's not the same as fully open-source. It means the trained model weights are published, so you can download and run the model, and adapt it. The training code and data may or may not be fully open. For a company like Aikido, releasing weights is a way to build trust with security teams who want to inspect what they're deploying, and to build a community of users who improve the model over time. It is also a competitive move against the big US cloud providers.
Samantha: For European businesses, the appeal is very concrete. The EU has strict rules on data protection under GDPR, and new obligations coming from the NIS2 directive for critical sectors and from the AI Act. A locally-run cybersecurity model makes many of those compliance questions easier to answer, because sensitive data simply does not leave the organisation. It also fits the broader European push for what officials like to call digital sovereignty, having key technologies developed and hosted inside the bloc rather than borrowed from abroad.
Bob: The bigger picture is that we may be seeing the start of a split in the AI security market: cloud-based tools for companies that want speed and convenience, and locally-run, open-weight tools for those who put data control first. Altar is a clear bet that in Europe, the second group is going to keep growing.
Organic Maps gives you full offline navigation across Europe without tracking, ads or an account.
Samantha: To close out, two things you can actually use today. First up, a free navigation app called Organic Maps. It's an open-source project built on top of OpenStreetMap data, and it's designed to work fully offline. You download the countries or regions you need in advance, and after that the app happily gives you turn-by-turn directions in a rural Alpine valley with no signal, on a Greek island where roaming is expensive, or on a long train ride where you just don't want to burn mobile data.
Bob: The pitch is not just that it works offline. Organic Maps is built around the idea that a maps app shouldn't be spying on you. There are no user accounts, no advertising, and the developers state they don't collect location data. Whether you take that on faith or not, the fact that the app is open-source means independent researchers can and do check the code. That is a very different starting point from the big commercial map apps, which increasingly treat your location as a product.
Samantha: In practical terms, Organic Maps covers driving, cycling, walking and hiking, with detailed contour lines and trails that most mainstream apps simply skip. Users often mention it in the same breath as OsmAnd and similar projects, but Organic Maps tends to get praised for a cleaner, simpler interface. It's available on iPhone, Android, and there are builds for other platforms too. Free to install, no subscription, no premium tier hiding the good features behind a paywall.
Bob: The European angle here is straightforward. If you're planning a weekend trip across a border, or a hiking holiday somewhere with patchy coverage, download the relevant country maps over WiFi before you go. It costs you nothing, works without a data connection, and keeps your movements off someone's advertising dashboard.
Subtitle Edit is the free open-source tool that finally makes cleaning up video subtitles painless.
Samantha: Second find, and this one is for anyone who works with video, whether that's a small business making explainer clips, a teacher recording lessons, or a hobbyist editing holiday footage. It's called Subtitle Edit. It's a free, open-source subtitle editor for Windows, with community-maintained builds for other platforms. And it does one thing extremely well: it takes the misery out of creating, translating and syncing subtitles for a video file.
Bob: The reason this matters is that subtitles are no longer a nice-to-have. Social platforms play videos muted by default, so captions decide whether people actually watch. Accessibility rules across the European Union increasingly require captions on public sector and business content. And translated subtitles are one of the cheapest ways to reach viewers in another language. Subtitle Edit gives you a proper timeline view, waveform display, spell-check, and support for dozens of subtitle formats.
Samantha: It also includes built-in tools for automatic transcription and translation, and users can plug in various speech recognition engines to generate a first draft of captions, then clean them up manually. That workflow, machine-generated first pass followed by a human polish, is where the real time savings live. It's the kind of tool that used to cost real money in the professional video world and is now genuinely free to download and use.
Bob: Two practical picks, then. Organic Maps for anyone travelling around Europe who wants navigation without the tracking, and Subtitle Edit for anyone who has ever tried to add captions to a video and given up halfway. Both free, both worth twenty minutes of your afternoon.
Samantha: Today we covered: AMD closing in on a trillion-dollar valuation as AI chip demand surges, the Gates Foundation's one billion dollar pledge to keep AI from widening inequality, the new Petal subsea cable linking North America and Europe, Belgian firm Aikido launching its locally-run Altar security model, and two things to try yourself: Organic Maps and Subtitle Edit.
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.