Meta credit outlook turns negative on AI spending

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

Bob: And I'm Bob Russell. Today: a credit warning for Meta over its AI spending, Microsoft offering tens of thousands of staff a way out, Elon Musk calling Grok's running costs unsustainably high, AI moving into companion dolls, and two things to look up yourself: the Artificial Analysis Intelligence Index and Kimi from Moonshot AI. Let's start with Meta.

S&P Global Ratings turns Meta's credit outlook negative as AI spending balloons.

Samantha: Meta's AI spending is now worrying the people who grade its debt. S&P Global Ratings has moved Meta's credit outlook to negative. That's not a downgrade yet, but it is the formal first step towards one, and S&P says a cut could come within the next two years. The reason is the bill for AI infrastructure. S&P expects Meta's capital spending, the money it puts into long-term assets like data centres and servers, to reach a hundred and forty billion dollars this year.

Bob: And the curve gets steeper. S&P forecasts that figure rising to more than two hundred and twenty-five billion dollars next year. That's the core of the warning: spending is accelerating faster than the balance sheet can comfortably absorb, and the agency points specifically to rising leverage, meaning Meta is leaning more heavily on borrowed money. A credit rating matters because it helps set how expensive that borrowing is. A lower rating usually means paying more interest on the next round of debt.

Samantha: What stands out is who is sounding the alarm. Usually it's shareholders grumbling about returns. This time it's a ratings agency, whose job is to judge whether a company can pay back what it owes. And Meta is one of the biggest names in tech, not a struggling startup. The signal from S&P is that even a giant can stretch its finances keeping up in the AI race. That puts a hard number on a question the whole industry is wrestling with.

Bob: There is a counterpoint. A negative outlook is a warning, not a verdict. Meta can still adjust its spending, or show that the investment is paying off, before S&P decides to act. But the timeline is telling. Two years is not long when you're building infrastructure on this scale. If the forecasts hold, Meta spends considerably more next year than this year, while the agency is already flagging the strain on its finances right now.

Samantha: For Europe, this lands on several desks. European financial markets and regulators will be watching Meta's credit health closely, because its AI infrastructure spending stretches across regions. European investors holding big tech shares or bonds will want to know whether this is a one-off or the start of a wider reassessment of AI spending. The bottom line is that the cost of competing at the very top of AI is now visible in credit ratings, not only in stock market swings.

Microsoft offers around sixty thousand employees worldwide a voluntary way out.

Bob: Microsoft is offering around sixty thousand employees worldwide a way out. The company has launched a voluntary exit program with financial support and career resources for eligible staff, including people in India, the US, the UK and Canada. Voluntary is the key word: nobody is being pushed out under this scheme. But the scale is striking, and it comes as Microsoft, like its rivals, pours billions into AI infrastructure and the applications built on top of it.

Samantha: And Microsoft isn't alone. Amazon, Oracle and Meta are also reassessing how big their workforces should be, which hiring plans still make sense, and which skills they actually need. The thread connecting them is AI. The question now being asked openly is whether AI lets companies produce more with fewer people. The report describes a shift in how these firms think about growth: less about adding headcount, more about output per employee.

Bob: That's a real change of model. For decades, a growing tech company meant a growing payroll. Now the metric getting attention is how much each person produces. A voluntary program is the softer route through that shift. It lets a company slim down without announcing layoffs, and it gives employees some say over timing. But it also signals where the company sees its future, and it raises the question of which roles fit into that picture.

Samantha: The UK is explicitly on the list of eligible countries, so this is already reaching Europe. Employees of Microsoft and other tech giants elsewhere on the continent may face similar restructuring as AI reshapes job roles, although nothing specific has been announced for them. For European policymakers, it feeds a debate that is no longer theoretical: what happens to skilled office work when the biggest employers in tech start measuring success per head rather than by headcount.

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.

Elon Musk calls Grok's running costs unsustainably high and signals new limits.

Bob: Elon Musk says running Grok costs too much. According to the report, Musk has indicated that the operating costs of Grok, xAI's chatbot, are unsustainably high compared with the revenue it brings in, and that new limits are coming. What's missing is detail. xAI hasn't disclosed any figures, what the restrictions will look like, or when they kick in. So for now we know the direction of travel, but not the destination.

Samantha: The admission itself is notable, though. AI chatbots are expensive to run: every question someone types costs computing power on the company's side, and that bill keeps coming as long as people keep asking. When usage grows faster than income, those costs pile up. The report frames this as a strategic reassessment of whether Grok makes economic sense in a crowded chatbot market, where several deep-pocketed rivals compete for the same users.

Bob: And it links straight back to our first story. Meta is being warned about the cost of building AI. Here, the warning is about the cost of running it, day after day, for every user. Two different problems, same pressure: the money going in has to be matched, sooner or later, by money coming out. Limits on usage are one of the bluntest tools a company has to close that gap.

Samantha: For European users, the practical effect depends on details we don't have. European Grok users may well see new usage restrictions as xAI tries to bring costs under control. Whether that means fewer questions per day, slower answers or features behind a paywall, xAI hasn't said, so anything beyond the word limits is speculation for now.

Bob: The report describes this as a sign of how hard it is to scale AI services while balancing innovation and profitability. For European businesses that build products on top of any chatbot, the takeaway is that usage terms are not fixed. They can change when the economics change, and this is a public example of exactly that happening.

Sex dolls are gaining conversational AI, voice recognition and cloud connections.

Samantha: AI is moving into a product most tech shows steer clear of: sex dolls. According to an industry website, these are shifting from static objects to devices with built-in AI chat, voice recognition and software that learns from how a user behaves. Many connect to a phone app or the cloud, which lets them respond dynamically and simulate a personality. It's a niche market, but it's a place where conversational AI, robotics and new materials all meet.

Bob: The drivers are the same ones powering the rest of AI. Better conversational software, progress in robotics and materials, and what the industry calls growing demand for customised, consistent companion experiences. Next on the roadmap, according to the source, are more realistic voices, simulated emotions and better learning abilities. Worth keeping in mind: the overview comes from a website within the business itself, so this is how the industry describes its own future.

Samantha: The cloud connection is where it gets sensitive. A device that listens, learns behaviour and sends data over the internet is collecting some of the most intimate information there is. That raises questions about where that data is stored, who can access it, and how well the connection is secured. None of those questions is unique to this product, but the stakes here are unusually personal for the people using it.

Bob: The source expects this to prompt discussion in Europe on ethics, regulation and consumer protection. Devices that simulate emotions and build a picture of a user's behaviour sit in uncomfortable territory for privacy and consumer law. For now it's a debate rather than a rulebook. But as AI companions move from screens into physical objects, the source suggests these questions will get more attention from European societies and legal frameworks.

The Artificial Analysis Intelligence Index shows which AI models lead on quality and speed.

Samantha: To close out, two things you can actually look up today. First, the Artificial Analysis Intelligence Index, a leaderboard that ranks AI models on how capable they are, and also tracks how fast they respond. It just had a shake-up. Meta's Muse Wonder Full and Moonshot AI's Kimi both moved up.

Bob: The index shows something a single headline can't: the trade-off between quality and speed. Speed is measured in tokens per second. A token is roughly a chunk of a word, so that number describes how quickly text appears on your screen. Kimi earns its place on quality rather than speed.

Samantha: Here's a practical way to use it. If you rely on a chatbot for work or study, look up the model you use and see where it lands, on both intelligence and speed. A model near the top that answers slowly may suit careful research. A fast one may suit quick drafts and short questions. It takes a few minutes in your browser, and it turns vague marketing claims into something you can compare side by side.

Bob: For European businesses and developers, the shake-up matters too. More strong models in the top tier means more choice for anyone building AI into their own products, including a prominent entrant from outside the United States. A leaderboard is one lens, not the final word, and rankings move regularly. But as a quick reference before committing to one provider, it gives you a structured starting point instead of a hunch.

Moonshot AI's Kimi becomes the first non-US model in the index's top five.

Samantha: The second thing to look up is Kimi, from Moonshot AI. Its latest Pro model reached the top five of that same index. So if you've only ever tried the big US chatbots, Kimi is a model this benchmark puts in the same league, and it's worth seeing how it handles the kind of questions you actually ask.

Bob: One caveat up front: the report doesn't say what Kimi costs or exactly where you can use it, so the first step is checking Moonshot AI's own channels for access and pricing. What the ranking does tell you is the profile. Kimi is described as favouring quality over speed. Answers may take a little longer to appear, but the index rates their substance highly, which is why it sits ahead of many faster rivals.

Samantha: That profile suits certain jobs. Long questions, summarising dense material, or working through a problem step by step are tasks where a slower, stronger model can pay off. A simple test: give Kimi and your usual chatbot exactly the same question and compare the answers. A few minutes is enough to get a feel for the difference in depth, tone and accuracy.

Bob: For European users, it adds a different option to the mix. Before pasting in anything sensitive, check the privacy terms and where your data is processed, as with any AI service. With that in mind, it's an easy experiment: one question, two models, and you'll see whether the leaderboard matches what you get in practice.

Samantha: Today we covered: Meta's credit outlook turning negative over AI spending, Microsoft's voluntary exit offer to sixty thousand staff, Musk's warning on Grok's costs, AI arriving in companion dolls, and two things to try yourself: the Artificial Analysis Intelligence Index and Kimi from Moonshot AI.

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.