Marvell buys Lumentum for $115B, TSMC builds in Dresden

Samantha: Welcome to The State of Tech, The European Edition, Tuesday November three, 2026. I'm Samantha Lawrence.

Bob: And I'm Bob Russell. Today: Marvell's hundred-and-fifteen-billion-dollar bet on Lumentum, Apple beats expectations but sees AI costs bite, why Tee-Es-Em-See is building chip factories from Arizona to Dresden, an AI-discovered antibiotic wins a major science prize, and two things to check out: the motion-capture system MOVIN TRACIN and the National Cyber Security Centre's annual review. Let's start with that megadeal.

Marvell agrees to buy Lumentum for 115 billion dollars to dominate AI connectivity.

Samantha: Marvell Technology wants to buy Lumentum, and the price tag is enormous: one hundred and fifteen billion dollars, paid entirely in shares. That puts it among the biggest deals tech has ever seen. The logic is all about AI data centres. Marvell makes the chips and switches that move data between processors, plus custom chips built for specific customers. Lumentum makes high-performance lasers and optical transceivers, the parts that turn electrical signals into light and back again. Put them together, and you get one company that can design the entire connection, copper and fibre, from end to end.

Bob: And that connection has become a real pressure point. AI infrastructure means huge numbers of processors exchanging data constantly, and the faster that traffic flows, the more work each expensive chip can actually do. The two companies say combining forces lets them design the electronics and the optics together, which should get optimised products to market faster. They also put a number on the opportunity: a total addressable market, meaning all the sales they could theoretically go after, of five hundred and fifty billion dollars.

Samantha: The timing is notable. Investors have been nervous lately about the sheer scale of AI investment, and the deal lands right in the middle of those jitters. Paying in stock rather than cash means Marvell doesn't have to hand over a hundred and fifteen billion dollars in money. Marvell is also promising shareholders a quick payoff: it says the takeover should lift its adjusted earnings per share within the first year after the deal closes. That's a clear signal it expects demand for AI infrastructure to keep running for a while yet.

Bob: A promise of first-year gains is still just a promise. The deal still has to close, and the bet depends on data centre operators continuing to spend heavily. There's also a competitive angle. When one supplier can offer the full package, from the chip to the laser, customers who used to buy those pieces separately may find themselves negotiating with a much bigger counterpart. For rival makers of optical and electrical components, that changes the competitive landscape considerably.

Samantha: For Europe, this lands right in the supply chain for the data centres being built across the continent. European operators and tech companies buy the same optical and electrical components as everyone else, so a merger of this size could affect prices, availability and who holds the bargaining power. Notable for European buyers: one of the key questions is whether faster, jointly designed products end up helping smaller players keep pace, or whether the combined company's size mainly strengthens its hand in every negotiation.

Apple beats earnings forecasts, yet shares slip on China, slower services and AI costs.

Bob: Apple delivered numbers that beat expectations, and its shares still slipped. For its fiscal fourth quarter, Apple reported around one hundred and fourteen and a half billion dollars in revenue and earnings of two dollars and ten cents per share. iPhone and Mac sales grew strongly. Yet in after-hours trading following the results, the stock fell by about one point eight percent. The reasons have little to do with hardware selling badly. There are three, and they all point at profitability.

Samantha: First, Greater China. Revenue there came in roughly six and a half percent below what analysts expected. Second, services, so the App Store, subscriptions and so on, grew by nine point eight percent. Solid on paper, but it's the slowest growth in three years. Third, and according to the reporting the most significant, management expects its gross margin, the share of revenue left after the direct costs of making its products, to fall next quarter, to between forty-five and forty-six percent.

Bob: Apple pins that squeeze on two things: rising storage costs and higher spending on artificial intelligence. That's the same tension running through the whole industry this autumn. Building AI features means paying for computing power and infrastructure up front, while the revenue from those features is far less certain. What this shows is that even a company as profitable as Apple now feels the price of AI in its margins, and investors are reacting before any extra money from those features comes back in.

Samantha: For European investors, Apple works as a bellwether for big tech. If Apple's margins feel pressure from AI spending, that raises the same question for every company racing to build similar features. The China miss matters too: it's a reminder of how much global device makers depend on one large and politically sensitive market. And notable is that Apple's own explanation names storage and AI, two costs that device makers and cloud companies everywhere face in some form.

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.

TSMC builds factories in Arizona, Japan and Dresden, pushed by geopolitics rather than choice.

Bob: Taiwanese chip giant Tee-Es-Em-See, the world's largest contract chipmaker, is building factories far from home: in Arizona in the United States, in Kumamoto in Japan, and in Dresden in Germany. A feature in the Taipei Times makes a sharp point about why. According to that analysis, this expansion isn't really a choice. It's being driven by geopolitics, by domestic politics, and by what the company's major international customers demand.

Samantha: Those forces all push the same way. Governments now treat chip production as a national security priority. The United States backed that view with money, through incentives under its CHIPS Act. And big clients want reassurance that the chips their products depend on aren't all made in one place. The goal, in short, is spreading the risk across several continents, so that a single disruption can't choke off the entire supply.

Bob: The analysis frames this as a balancing act. Tee-Es-Em-See has to keep governments and customers on side while competition across the chip industry keeps growing and political tensions stay high. Diversifying is about resilience: making sure a shock in one region doesn't ripple through every product that depends on these chips, from phones to servers.

Samantha: For Europe, the key piece is Dresden. The plant there operates as the European Semiconductor Manufacturing Company, and it's seen as a significant boost for European chip sovereignty, meaning the ability to make more of the chips Europe needs on its own soil. The aim is stronger local supply chains and less reliance on a single region for manufacturing.

Bob: There's a flip side to that framing, though. If the expansion happens because the company has to, rather than because it wants to, then factories like the one in Dresden rest on political and customer pressure staying in place. For European industry, the plant offers a closer source of chips. For policymakers, the lesson from the analysis is that this kind of investment follows geopolitics and customer demand, with government incentives like the CHIPS Act as part of the mix.

MIT researchers win the Pritzker Prize for using AI to discover new antibiotics.

Samantha: Two professors at the Massachusetts Institute of Technology, Regina Barzilay and James Collins, have won the 2026 Pritzker Prize for AI in Science Research Excellence. The prize recognises their work using artificial intelligence to find new antibiotics. Their best-known result is halicin, described as the first antibiotic discovered by AI, and it can kill bacteria that have become resistant to existing drugs, the so-called superbugs. The work combined three fields: deep learning, bioengineering and experiments in the lab to check the results.

Bob: That combination is what made it work. The deep learning side predicts which molecules might kill bacteria, and the lab side then tests whether those predictions hold up in real experiments. According to Em-Eye-Tee, the approach can shrink a search that used to take years into a matter of weeks. In medical research, being able to focus on the most promising candidates that much sooner changes how quickly scientists can respond to new threats.

Samantha: The urgency comes from what the prize announcement calls an escalating global crisis of antimicrobial resistance: bacteria that no longer respond to the drugs we rely on. That's why a tool that speeds up the search for new antibiotics carries so much weight. An award like this also marks AI-driven drug discovery as recognised, prize-winning science. And it's a reminder that some of the most consequential uses of AI happen in laboratories, far away from chatbots and office software.

Bob: For Europe, the relevance is direct. European pharmaceutical companies and research institutions face the same resistant bacteria, and they're looking for faster ways to find new drugs. The approach Barzilay and Collins pioneered offers a template European labs can build on. One point of precision: the prize recognises the discovery method and the science behind it. The material published about the award doesn't say when, or whether, halicin itself will reach patients as a medicine.

MOVIN TRACIN captures human motion in 3D without suits or markers, using lidar and AI.

Samantha: To close out, two things you can check out today, starting with a tool for creators. It's called MOVIN TRACIN, from the South Korean startup MOVIN. It does motion capture, recording how a person moves and turning that into 3D data, but without the tight suits covered in little markers you might know from behind-the-scenes film footage. Instead, it pairs laser scanning sensors, known as lidar, with the company's own AI, and it works in real time.

Bob: The pitch is cheaper motion capture with much less setup time. That's relevant for game developers, animators and video makers who want a digital character to move like a real person. And the company has just raised fresh money: a Series A round, led by IMM Investment, according to Dealroom. The plan is to upgrade the technology and push hard into markets outside South Korea.

Samantha: It's also moving beyond entertainment into robotics. The system is already being used to remotely control and train robots: a person moves, and the robot copies that movement or learns from it. Boston Dynamics is named among the firms using the technology. That's the interesting crossover here: the same movement data that brings a game character to life can help teach a machine how to move.

Bob: The practical step is simple: look up MOVIN TRACIN on the company's website and watch how the capture works in practice. A few minutes should give you a sense of whether it fits your project. The reports don't mention pricing or which animation software it works with, so check those details directly with the company before planning around it.

Samantha: For Europe, the new funding is explicitly aimed at overseas expansion, which could make the system easier to get hold of for European studios, independent game makers and robotics labs. Markerless capture could lower development costs for smaller teams. The thing to watch is whether motion data like this becomes an everyday tool rather than something reserved for big-budget productions.

Britain's National Cyber Security Centre annual review warns of AI-powered ransomware on infrastructure by 2028.

Bob: The second thing to check out is a document rather than an app: the annual review from Britain's National Cyber Security Centre. Its headline finding is stark. The centre says it is highly likely that malicious software powered by AI will be used in ransomware attacks on critical national infrastructure by 2028. Ransomware is the kind of attack that locks up an organisation's computer systems until a payment is made. The targets the review names are power stations, airports and hospitals.

Samantha: The review spells out how AI helps attackers. It can speed up scouting for targets, spotting weaknesses in a network, and writing malicious code. The agency also expects Russia and China to routinely use AI in their cyber attacks. That makes the review useful context for anyone working in IT at a hospital, a utility or a transport hub, because it describes the threat in concrete terms and attaches a timeframe to it.

Bob: The practical step: look up the review, which the Guardian reported on this week, and set aside some time for the parts on AI-driven threats. For smaller organisations without a dedicated security team, it's a way to understand what larger organisations are preparing for, and to raise the right questions with whoever manages your systems. It is a threat assessment, so expect a picture of the risks and their likelihood, along with the agency's view on which countries are behind them.

Samantha: For Europe, the message travels well beyond Britain. Critical infrastructure across the continent faces the same kinds of state-backed and criminal attackers. The warning points towards more investment in AI-driven defences and closer international cooperation. What the UK expects by 2028 gives European organisations a concrete reference point to compare their own defences against, and a clear sense of how quickly the threat is expected to evolve.

Samantha: Today we covered: Marvell's hundred-and-fifteen-billion-dollar move for Lumentum, Apple's earnings beat with AI costs biting into margins, Tee-Es-Em-See's push into Arizona, Japan and Dresden, an AI-discovered antibiotic winning the Pritzker Prize, and two things to check out yourself: MOVIN TRACIN and the National Cyber Security Centre's annual review.

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.