19 november 2026 · 13:47
Anthropic IPO at $1.5T, Nadella on AI agents and ads
The Anthropic IPO priced at $83 a share on the Nasdaq, valuing the AI company at $1.5 trillion and raising more than $94 billion including private placements. In this episode of The State of Tech, The European Edition, Samantha Lawrence and Bob Russell explain what the listing means for European investors and the EU AI Act. They also examine Satya Nadella's prediction that AI agents will reshape web advertising and the economics of online businesses.
Beluister deze aflevering:
Transcript
Samantha: Welcome to The State of Tech, The European Edition, Thursday November nineteen, 2026. I'm Samantha Lawrence.
Bob: And I'm Bob Russell. Today: Anthropic prices a record-breaking stock market debut, Satya Nadella says AI agents will rewrite the economics of the web, Allianz Commercial counts the rising cost of AI failures, Ford starts building batteries in Michigan with Chinese technology, and two things to try yourself: Jan and VLC. Let's start with Anthropic.
Anthropic prices its IPO at eighty-three dollars a share, valuing the company at one and a half trillion dollars.
Samantha: Anthropic, one of the biggest names in artificial intelligence, has priced its initial public offering at eighty-three dollars a share. That is above the range it first aimed for, and it puts the company's total value at one and a half trillion dollars. The shares are listed on the Nasdaq under the ticker ANT. Together with private placements from major technology partners, including Google, Nvidia, chipmaker AMD and Microsoft, the deal raised more than ninety-four billion dollars.
Bob: The demand side is the striking part. The offering was eight times oversubscribed. Investors asked for eight times more shares than were available, and that demand came from both large institutions and ordinary retail investors. According to the report, that surpasses earlier funding records set by private space companies. And it happened against a backdrop of shaky markets and executive turnover across the tech sector. So this was not a calm moment in which everything floats upward. Investors chose to pile into one specific bet.
Samantha: That bet is a pure-play generative AI company. Anthropic's business is the AI itself: models that write, summarise and reason. It is not AI as one feature inside a larger company that sells phones or advertising. Until now, investors who wanted that kind of exposure mostly had to go through the big platforms. A listing like this changes that. Being listed also usually brings more public scrutiny, so the company now operates under a much brighter spotlight.
Bob: The partner list deserves attention too. Google, Nvidia, Ay-Em-Dee and Microsoft took part through private placements alongside the public offering. That puts some of the biggest names in chips and cloud computing, the hardware and infrastructure AI runs on, behind the same company. This shows how closely the AI industry's biggest players are now tied together. And ninety-four billion dollars in fresh money gives Anthropic room to expand at a scale few companies anywhere can match.
Samantha: For Europe, the first question is valuations. European investors and AI startups will be watching whether a one-and-a-half-trillion-dollar price tag lifts the value of AI companies on this side of the Atlantic too, or widens the gap with American players. The second question is regulation. Anthropic is now a listed company whose products also have to line up with the EU AI Act, Europe's rulebook for artificial intelligence. How a public company of this size handles those rules is something both Brussels and investors will follow.
Satya Nadella warns that AI agents will break the ad-funded business model of the web.
Bob: Microsoft chief executive Satya Nadella says the internet's business model is about to break. Speaking at the DealBook Summit, he argued that autonomous AI agents, software that browses, compares and buys things on your behalf, will fundamentally change the economics of the web. Much of the internet is paid for by advertising aimed at human eyes. An AI agent doing the shopping does not look at banner ads. According to Nadella, ad-supported models built on human traffic will crumble as a result.
Samantha: He also had a warning for websites that try to keep these agents out. Platforms that block AI agents from their services, he said, risk losing their dominant position to open competitors who welcome agent traffic. His prediction for what comes next: money moves away from ad impressions, where advertisers pay each time a person sees an ad, toward transaction fees and pay-per-click models designed for automated systems. It is worth stressing that this is a prediction from a tech chief executive, not a measured trend.
Bob: The tension for website owners is clear. Publishers and online shops live from visitors. If agents become the visitors, a site has two options: let the agents in and accept that people see fewer ads, or block them and risk being skipped entirely. In Nadella's framing, blocking is the losing move, and sites that open up gain market share. That is a shift from competing for human attention to competing to be the source an AI system picks.
Samantha: European digital publishers and e-commerce platforms are directly in the line of fire. Publishers that rely on advertising, and online retailers that rely on shoppers landing on their pages, would see those revenue streams shift if consumers increasingly let AI do their online purchasing. The bottom line for European businesses: the question is no longer only how to attract readers and customers, but how to be found, compared and chosen by software acting on their behalf.
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.
Allianz Commercial reports a 150 percent surge in serious corporate AI incidents, with single claims above a hundred million.
Bob: When AI goes wrong inside a company, the bill can now run past a hundred million dollars for a single incident. That is the warning from insurer Allianz Commercial in a new report. It counts a one hundred and fifty percent rise in serious corporate AI incidents over the past two years.
Samantha: According to the report, the real exposure is larger than the figures show. Insurance statistics understate it, because AI failures are often filed under other headings: cyber attacks, faulty products, or claims against company executives. So an AI mistake can show up in the data as a hacking claim or a product liability case, and never be counted as AI at all.
Bob: The report names three sources of systemic risk. First, errors made by AI agents embedded in business processes. Second, defective safety systems in automated vehicles. Third, shareholder lawsuits over companies that made misleading claims about their AI. That last one is notable: the risk is not only that AI fails, but that a company promised investors more than the technology delivered.
Samantha: In a related move this week, Workday, which makes human resources and finance software, is buying Tel Aviv startup Honeydew. Honeydew builds a layer that makes every system use the same definitions for figures like headcount or operating costs. The goal is to stop AI agents from giving management conflicting or inaccurate reports. It is one example of the kind of data governance insurers are urging companies to adopt.
Bob: For European companies this lands at a sensitive moment. Insurers are adjusting policy terms ahead of enforcement of the EU AI Act, and that points toward stricter risk disclosures and liability requirements for businesses. Allianz's experts urge firms to put proactive risk management and governance standards in place now. For a European company running AI agents, how well it documents and controls those systems may increasingly shape how insurers look at it.
Ford starts making cheaper electric car batteries in Michigan using technology licensed from China's CATL.
Samantha: Ford has started making electric car batteries at its own factory in Michigan, using technology licensed from Chinese battery giant CATL. The batteries are a type called lithium iron phosphate, built around iron and phosphate, and Ford is using them to bring down the cost of its next generation of electric vehicles. The company also wants this plant to help build a more resilient battery supply chain in North America.
Bob: The politics make this complicated. The factory opens in the middle of a heated debate in Washington about federal subsidies, technology transfers and foreign involvement in critical green infrastructure. Using Chinese know-how inside an American plant touches all three. Ford's answer is ownership: the company owns and operates the facility directly, which it says keeps manufacturing jobs in America while advancing the country's energy transition goals.
Samantha: Licensing is the key word. As Ford presents it, the company is not buying finished batteries from China and not handing the factory to a Chinese partner. It pays to use the technology and builds the batteries with its own workforce. In principle, that model gives a carmaker access to proven, lower-cost battery technology without developing it from scratch. Whether that satisfies critics in Washington remains the open question.
Bob: European carmakers are watching closely. They face similar pressure to make electric cars cheaper, and similar questions about relying on Chinese battery technology. If American production of this battery type scales up, it could affect global pricing for these batteries and the competitive position of European brands. The bottom line: battery chemistry is no longer just an engineering choice. It is a question of geopolitics, jobs and price, on both sides of the Atlantic.
Jan lets you run AI chatbots directly on your own laptop, without the cloud.
Samantha: To close out, two things you can actually use today. First, for anyone curious about AI after all that Anthropic news, but who would rather keep things on their own computer: Jan. It is a free, open-source desktop app that lets you download openly available AI models and chat with them directly on your laptop, without sending your questions to a cloud service.
Bob: It works on Windows, Mac and Linux. You install it, browse its model hub, pick one, and it downloads to your machine. After that, the chat window looks familiar to anyone who has used a chatbot. Because the model runs locally, it keeps working without an internet connection once the download is finished. Users often point to that as the main appeal: private notes or work documents never have to leave the device.
Samantha: There is a real limitation, and it is hardware. These models need memory and processing power. A recent laptop with plenty of memory can run the smaller models reasonably well, while an older machine will struggle or run very slowly. The models are also generally smaller than the biggest commercial chatbots, so answers can be less polished. The app gives an indication of which models are likely to fit your computer before you download.
Bob: A practical approach is to start small. Pick one of the lighter models, often a few gigabytes, and try it on something low-stakes, like summarising a document or rewriting an email. Expect the first download to take a few minutes. And a model on your own laptop can still make things up, just like a cloud chatbot. Running locally changes where your data goes, not whether the answers are right.
Samantha: For European users, the privacy angle is the draw. Keeping data on your own device avoids the question of where a cloud provider stores and processes it, which matters under Europe's data protection rules, especially for work documents. It also offers a hands-on look at what these models can and cannot do, at a moment when investors are valuing AI companies in the trillions.
VLC, the free French media player, opens almost any video or audio file you own.
Bob: The second one has nothing to do with AI: VLC, the free media player from the French non-profit VideoLAN. If you have ever double-clicked a video and got an error about a missing codec, the piece of software needed to decode a particular format, VLC is the usual fix. It plays almost any video or audio file, from old home videos to obscure formats, without extra downloads.
Samantha: It runs on Windows, Mac and Linux, with versions for Android and iPhone as well. It is free and open source. Several features are less well known. You can speed up or slow down playback, which users find handy for lectures and long recordings. You can also load subtitle files alongside a film, and shift their timing if they drift out of sync.
Bob: There is also a converter built in. Through the convert option in the menu you can turn a video into a different format, or pull just the audio out of a clip, useful when a file will not play on a phone or a TV. It can open online radio or video streams from a web address too. One limitation: the menus look dated and some options are buried deep, so expect some clicking around.
Samantha: Getting started takes about two minutes. Download it from the official VideoLAN website rather than a third-party download site, because copies bundled with unwanted extras do circulate. Install it, then right-click any video and open it with VLC. For anyone with a drawer of old memory sticks or a hard drive of family videos in odd formats, it is a quick way to make them playable again.
Samantha: Today we covered: Anthropic's record stock market debut, Satya Nadella on AI agents and the economics of the web, Allianz Commercial on the rising cost of AI failures, Ford's Michigan batteries with Chinese technology, and two things to try yourself: Jan and VLC.
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.