16 november 2026 · 13:49
Nvidia: Jensen Huang eyes return, SpaceX catches Starship
Nvidia leadership is back in the spotlight after Jensen Huang's forced exit, with the Financial Times reporting that he is weighing a return and allies are lobbying the board. In this episode of The State of Tech, The European Edition, we unpack that power struggle, explain what US export scrutiny means for European firms, and look at SpaceX's first spaceship catch and what reusable rockets mean for space infrastructure.
Beluister deze aflevering:
Transcript
Samantha: Welcome to The State of Tech, The European Edition, Monday November sixteen, 2026. I'm Samantha Lawrence.
Bob: And I'm Bob Russell. Today: Jensen Huang is working on a comeback at Nvidia, days after being pushed out. Apple launches a study into how far people trust medical AI. SpaceX catches its spaceship, and what that means for who gets into space. US tech veteran IBM and Microsoft bundle thousands of AI agents for big business. And two things to try yourself: Goblin Tools and Heavens-Above. Let's start with Nvidia.
Days after his forced exit, Jensen Huang is reportedly plotting a return to Nvidia.
Samantha: Four days ago Jensen Huang stepped down as chief executive of Nvidia, after thirty-three years in charge. Now, according to the Financial Times, he wants the job back. Citing several people familiar with his thinking, the paper says Huang is weighing a return and has discussed it with confidants. Reuters and Bloomberg followed the report. He left after reports of extramarital relationships. Bloomberg says the board pushed him out, and that Huang admitted a relationship with a woman suspected of spying for China. A second relationship, with a subordinate at Nvidia, has been reported, not confirmed.
Bob: The comeback push is already loud. An ally told the FT that a war room of Silicon Valley heavyweights is being set up to lobby for his return, and that they want to seek support at the White House. There's a public campaign too, Stand with Jensen, with an open letter to the board titled Bring Jensen Back. Elon Musk and Epic Games boss Tim Sweeney are among the backers. Musk warned on X that many of Nvidia's supplier deals were a handshake with Jensen, not a written contract. The petition claims well over a hundred thousand signatures, but it doesn't verify identities.
Samantha: Two big shareholders want the door left open. Brad Gerstner of Altimeter Capital and Gavin Baker of Atreides Management urged the board in an open letter to keep all options open, including a possible return of Jensen Huang, and not to rush into choosing a successor. For now, finance chief Colette Kress is interim boss. The FT reports that Rene Haas, head of chip designer Arm, has been approached as a possible successor. Nvidia's board says there's no formal process. Nvidia and Huang's spokesperson declined to comment.
Bob: Washington is pulling the other way. Treasury Secretary Scott Bessent called on Sunday for a dramatic clean-up at the top, urged the board to bring in an outsider, and gave it ten days to show full compliance with US export controls, citing national security. The board has also ordered an independent investigation into Huang's conduct. And unlike many founders, Huang doesn't hold a decisive majority of the votes, so he can't overrule the board. Everyone is pointing to Sam Altman, fired by OpenAI in November 2023 and back four days later after a revolt by staff and investors.
Samantha: So this is a power question. Who really runs the world's most valuable chip company: the board, the shareholders, the staff, or Washington? Nvidia is worth over four trillion dollars and presents quarterly results tomorrow, so any word on leadership will be scrutinised. For Europe, US export controls tightening around such a critical supplier means more compliance scrutiny for European chipmakers and tech firms too. And underneath it all sits one question: can a company that leans this heavily on one man manage without him?
Apple launches a clinical study into how people understand and trust medical AI answers.
Bob: Apple wants to know whether you actually trust a medical chatbot. It has launched the Apple Health AI Usage Study together with OpenEvidence, a platform that uses AI to answer clinical questions. It's a virtual research study with its own iPhone app. The answers participants get are grounded in peer-reviewed medical literature, from journals like JAMA and The New England Journal of Medicine, with a citation attached to every response.
Samantha: The study isn't mainly testing the AI's medical knowledge. It's testing the people. How well do users understand the answers, how much do they trust them, and what do they actually do with them? That last part is crucial. An accurate answer can still lead someone astray if it's misread, or treated as a diagnosis instead of background information. Apple says the insights should feed into scientific benchmarks and help shape responsible design standards for consumer health AI across the industry.
Bob: That's the context. Lots of people already put health questions to chatbots, usually with no idea where the answer came from. Showing a source for every claim makes answers checkable. But it raises exactly the question the study is designed to answer: do people actually look at those citations, or does a confident tone do the persuading? Notable too is the choice of partner, a specialist that grounds its answers in clinical literature.
Samantha: The announcement doesn't say whether people outside the US can take part. But the questions land squarely on European desks. Health regulators and teams monitoring compliance with the EU AI Act will want to know how Apple validates clinical answers for patient safety. And if this study does end up shaping industry design standards, those standards will sooner or later have to work alongside European rules on AI and health information.
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.
SpaceX's first spaceship catch puts a fully reusable rocket within reach, and Musk in a gatekeeper role.
Bob: In our last episode you heard SpaceX had caught its Starship spaceship for the first time. Today, the bigger picture. On Saturday the upper stage, the actual spaceship, came back after fourteen orbits of the Earth and was grabbed by the two mechanical arms on the launch tower at Starbase in Texas, nicknamed the chopsticks. It's the first time an upper stage back from orbit has returned precisely enough to be reused. Until now, that part was lost on every launch.
Samantha: It was only half a win. The plan was to catch both parts for the first time, but the booster, the lower stage, couldn't be caught on Friday. SpaceX diverted it to the Gulf of Mexico, where it exploded before hitting the water. The flight did release twenty-six new Starlink satellites, joining a network of more than eleven thousand that is SpaceX's profit engine. And this is the rocket NASA wants to use to put astronauts on the Moon.
Bob: Our newsroom's analysis goes a step further. A fully reusable rocket turns Elon Musk into a gatekeeper. Take AI data centres in space, a serious idea: up there you get abundant solar power, no electricity grid, and no need for cooling water. Whoever can lift heavy cargo that cheaply and that often effectively decides who builds those first. Musk's own companies first, then whoever he lets in, at his price. Everyone else is miles behind, other companies, but also China, Russia and the entire European Union. Nobody else has a rocket this size that comes back whole.
Samantha: Which raises the obvious follow-up. If access to space, and with it computing power in space, sits in the hands of one man, what leverage does anyone else have? Governments and rival companies would be negotiating with a supplier who is also a competitor. For Europe the question is sharp: the next layer of computing infrastructure could depend on launch capacity it doesn't control, set by someone with his own commercial interests in space.
Bob: For now, that's a scenario, not reality. The booster still has to be caught, and a recovered spaceship has to fly again to prove the point. But today, only SpaceX has taken this step.
IBM and Microsoft team up to put thousands of ready-made AI agents inside big companies.
Samantha: Eye-Bee-Em's consulting arm and Microsoft have launched Imagine Enterprise, a joint platform to get autonomous AI agents working inside large companies. Agents here means software that carries out steps in a workflow on its own, rather than just answering questions. Eye-Bee-Em brings more than six thousand ready-made agents and related building blocks. Microsoft brings its systems for managing who can log in, its cloud, and its tools for security and data governance.
Bob: The main target is regulated sectors, finance and healthcare first. And the people may matter as much as the software. Both companies are sending what they call forward-deployed engineers, specialists who build, test and manage systems where several agents work together, directly inside the client's own IT environment. Pairing software with on-site engineers suggests that getting agents to work safely in a bank or a hospital still takes a lot of hands-on effort.
Samantha: It also shows two big players combining strengths. Eye-Bee-Em has the industry know-how and the catalogue of agents; Microsoft has the plumbing that many corporate IT departments already run on. For customers already using Microsoft's login and security tools, the pitch is that agents slot into controls they have, instead of needing a separate system to police them.
Bob: For European companies, that control question is central. Under strict EU data protection law, a bank in Frankfurt or an insurer in Milan needs to know exactly what an autonomous agent touches and where the data goes. The platform is pitched as a way to deploy agents while staying compliant. Whether that holds in practice will depend on individual deployments, and on how European regulators judge software that acts on its own.
Goblin Tools turns overwhelming tasks into small, concrete steps with simple AI helpers.
Samantha: To close out, two things you can actually use today. First, Goblin Tools. It's a collection of small, single-purpose AI helpers, made by a developer in Belgium, originally built for neurodivergent people who find everyday tasks overwhelming. The best-known one is Magic ToDo: you type in a task like clean the kitchen or sort out my taxes, and it breaks it into small, concrete steps. Click on any step and it breaks that down further.
Bob: There's more in the box. The Formalizer rewrites a message in a different tone, more polite, more direct, or less blunt. The Judge reads your text and tells you how it's likely to come across, handy before sending a tricky email. The Estimator guesses how long a task will take, and the Compiler turns a messy brain-dump into a tidy list. There's even a Chef that suggests a recipe from ingredients you list.
Samantha: The practicalities. The website is free and works in any browser, no account needed. There are also apps for iPhone and Android, which cost a small one-off fee. Reviewers often point to how stripped-down it is: no long chat conversation, just one box and one button per job.
Bob: Two caveats. The steps are generated by AI, so they can be generic or occasionally off; treat them as a starting point. And don't paste in anything sensitive, like medical details or confidential work documents, because the text is processed by an online AI service. A quick test this afternoon: take the one task you keep postponing, drop it into Magic ToDo, and see whether the first step suddenly looks doable.
Samantha: Because each helper does just one thing, there's almost nothing to learn before it's useful. That simplicity is a big part of why it has found a loyal following.
Heavens-Above shows when the space station and Starlink satellites pass over your home.
Bob: Second, something for after dark, and it ties straight back to those twenty-six Starlink satellites from the Starship flight. Heavens-Above is a long-running free website that tells you when satellites will pass over where you live, and whether you'll see them with the naked eye. You set your location once, on a map or by searching for your town, and it lists upcoming passes with times and directions.
Samantha: The classic target is the International Space Station, a bright, steady point gliding across the sky in a few minutes. But the site covers many more satellites, including a page dedicated to Starlink. Freshly launched Starlink batches often travel in a line, like a string of pearls, before spreading out. It also has sky charts and information on planets and comets.
Bob: Using it tonight is simple. The best viewing is usually shortly after sunset or before sunrise, when your sky is dark but the satellite is still lit by the sun. Check the time, the direction it appears from, and the brightness figure: the lower that number, the brighter the object. No telescope needed, and no account for the basics, though you can register to save your settings.
Samantha: The limitations: you need a clear sky, and city light hides fainter satellites, so the space station is the most reliable first try. And with more than eleven thousand Starlink satellites up there, a clear night anywhere in Europe now comes with a steady stream of them.
Samantha: Today we covered: Jensen Huang's push to return to Nvidia, Apple's study on trust in health AI, SpaceX's first spaceship catch and the gatekeeper question, Eye-Bee-Em and Microsoft's AI agent platform, and two things to try yourself: Goblin Tools and Heavens-Above.
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.