No, the AI hype is not over

- 19 mins read

In a campaign hall in Florida a young man steps on stage and says one sentence that earns him the loudest applause of the evening: he will not build any new data centers. Markus Lanz observed that, and for him the applause is a warning sign. The big AI euphoria, he says in the new episode of “Lanz & Precht”, is “completely over”.

I listened to the episode and had to put some things in perspective. Not because the two men are stupid, on the contrary, I enjoy the podcast, especially because they do not always agree with each other. But because nearly every argument follows the same reasoning error, and because that error has a long, embarrassing tradition. As someone who works with this technology all day, I laughed out loud at some points. Not from mockery, from recognition. These are the exact same sentences people said thirty years ago about the internet.

Precht partially pushes back, partially goes along. So let us go through it step by step. This will be longer, but the topic deserves it.

What the two actually said

Let us start clean, with real quotes, not an outrage caricature. Otherwise I make the same mistake I am about to accuse them of.

Lanz tells a story from the United States. He gives a name: James Fishback, a young Republican in Florida. Fishback received the loudest applause of the evening when he promised not to build any new data centers in Florida. From that Lanz draws his conclusion: the next American election could come down to whoever promises to protect the population from AI. His key sentence: the big euphoria is completely over, “this has not arrived yet here”. And, quote, it is about “your weird technology that is not really getting us anywhere”.

Two things are fair to note. Precht does not fully carry the hard thesis. At one point he literally says that AI is “definitely not” at its end. He asks more than he judges. And the second point, which I emphasize even more because it is correct: the energy argument that runs through the episode is real. More on that later.

But the big conclusion, the end of euphoria, hangs on that single applause. And that is exactly where the problem starts.

The star witness does not hold up

If I declare an entire technology failed, my evidence should carry weight. Let us look at it.

James Fishback, born 1995, is running for the Republican nomination in Florida’s 2026 gubernatorial election. Before that: hedge funds, then his own investment firm, whose ETF was blocked by its own regulators over legal concerns. In summer 2025 he claimed to have been an advisor at the government agency DOGE. That was not true. In the polls he is an outsider.

This is the man whose applause is supposed to serve as a seismograph for the end of AI. I am not saying this to belittle him. I am saying it to show the size of the logical error: from a single campaign scene, a global conclusion.

And even if Fishback were the most serious politician in Florida, the fundamental problem remains. Applause for a sentence against technology is nothing new. It is the rule. It was always there. And the people who clapped were almost always wrong.

A methodological aside: anecdote is not causation

Before I get to the history, a point that really bothers me as an engineer. Lanz argues causally, but his arguments are not reproducible.

“One candidate gets applause, so the hype is over.” “OpenAI and SpaceX go public spectacularly, so the corporations sense the end and want to cash out one last time.” That sounds like analysis, but it is reading coffee grounds. Because the second claim cannot even be falsified. If stock prices go up, it is greed before the crash. If they go down, it is the crash itself. If they stay stable, it is the calm before the storm. A thesis that is confirmed by every possible outcome explains nothing. It is, in Karl Popper’s words, not falsifiable.

A single observation, one applause, one IPO, is not evidence for a trend. It is a data point. And drawing a curve from one data point is not a forecast, it is a Rorschach test. With this method you can prove anything, including the opposite. At this point I became fully alert. Because if the applause itself has to serve as proof, it is worth looking at who is actually clapping, and how many times the same kind of applause has been heard in history.

Precht delivers the counterargument himself

The beautiful thing is: I do not even have to bring up the historical parallel myself. Precht does it in the podcast. He says that whoever is against AI reminds him of the Luddites, the machine breakers, who smashed the mechanical looms in the 19th century when their manual labor was replaced. They tried to stop time. And time moved on without them.

Exactly. That is the point. Precht just does not follow it through.

Because that applauding hall in Florida, cheering a politician who wants to keep data centers away, that is the modern version of the machine breakers. Not the developers who use AI. The ones who clap when someone promises to keep it out. Whoever reads that applause as proof of the technology’s end is confusing the weavers’ feelings with the course of history. The weavers had every right to their anger. They were still wrong about their prediction.

The oldest reasoning error in the history of technology

In 1998 an economist, who would later win the Nobel Prize in Economics, wrote the following sentence: by about 2005 it will become clear that the internet’s impact on the economy has not been greater than that of the fax machine. The man’s name is Paul Krugman. He has had to defend himself on this multiple times since then.

This is not an isolated case, this is a pattern. The electrification of horse trams was fought in every city by a determined minority. The automobile was held back for decades by legislation, against strong prejudice. The historian of technology Christian Vater puts it on the point: skepticism toward technology appears in the earliest written records at all.

Always the same picture. A loud minority declares the end. A hall applauds. And the technology comes anyway. Applause is not an argument. Applause is a feeling, and feelings have never been reliable as technology forecasts.

Stock bubbles and technology are not the same thing

A second thesis appears in the podcast, introduced via Precht’s son who works in the industry. We are seeing the peak right now, the bubble bursts in one to two years, the spectacular IPOs are just the last cash-out before the crash.

Here two things are confused, and the distinction is critical. A financial bubble and a technology are two different things.

Let us remember the dot-com bubble. The NASDAQ rose between 1995 and its peak in March 2000 by 600 percent. Then it fell by 78 percent and gave back all gains entirely. Pets.com, Webvan, Boo.com, WorldCom, all gone. The bubble burst completely.

And the internet? Kept growing. Uninterrupted. Amazon, Google, eBay survived and dominate the world today. Whoever looked at the bursting NASDAQ in 2000 and concluded the internet was finished missed the biggest shift of their generation.

This is the error in the peak thesis. Even if an AI stock bubble exists and even if it bursts, that says nothing about the technology. Stock markets value expectations. Technology changes how people work. One can collapse while the other continues. That is even the normal case.

“This has not arrived here yet”

This is the sentence that really made me turn around internally. The euphoria is over, not yet arrived with people. This is not an opinion you can argue about. This is measurably false.

ChatGPT has around 900 million weekly users according to official OpenAI figures, and crossed the one billion monthly user mark in June 2026. A year earlier it was 400 million. Over two billion messages per day. One billion people per month, that is roughly one in eight humans on the planet. And on television someone says it has not arrived yet with people.

Among developers, the profession supposedly dying out first: according to Stack Overflow’s 2025 Developer Survey, 84 percent use or plan to use AI tools, up from 76 percent the previous year. 51 percent of professionals daily. And, this is important to me, the same developers are critical. Agreement is declining, 46 percent distrust the accuracy. This is the opposite of blind hype. This is mature, critical mass usage. People using a tool and still knowing where it lies.

In companies, McKinsey’s “State of AI” study from November 2025 shows that 88 percent of organizations use AI in at least one function, up from 78 percent. More on that in a moment, because this is where the most honest counterpoint is hidden.

One billion users, 84 percent of developers, 88 percent of companies. “Not arrived yet” is not a bold thesis. It is just the wrong sentence.

The resistance is real. And they still build.

Now the fair part, because the anti-data-center sentiment does exist. In the United States, projects worth around 64 billion dollars were blocked or delayed by local, cross-party resistance, according to Data Center Watch. Fortune reports 48 projects with 156 billion dollars in 2025 alone. The number of canceled projects quadrupled, from six in 2024 to 25 in 2025. Most common reason: water consumption, then energy.

This is real, and it is exactly the point Lanz is making. Only, right next to that number sit roughly 725 billion dollars that the big cloud providers are investing in AI infrastructure in 2026 alone. Amazon around 200, Google around 185, Meta around 125, Microsoft around 120 billion. A plus of roughly 77 percent over 2025. And in the first quarter of 2026 these companies did not lower their forecasts, they raised them together.

64 billion delayed, 725 billion still being built. If the euphoria were completely over, would the companies with the deepest insight invest 77 percent more? Resistance is not collapse. There was resistance against power lines, against cell towers, against wind turbines. A loud minority brakes individual projects. The applauding hall from Florida appears again here, only this time with numbers behind it. And the numbers say: the wave keeps rolling.

“It catches the ones who thought they were never affected”

The strongest, most human moment in the podcast is when Precht talks about the labor market. This time it is not about the hands, but about the brain, the second industrial revolution. The mental middle class, which always believed it was safe, is affected now. And the question is whether social mobility will remain possible in the future.

I take this fear seriously. It is justified. But the conclusion, the end of upward mobility, is again a short-circuit from a snapshot.

Because exactly the same thing was said during the first industrial revolution. The industrial revolution did make weavers unemployed, yes. But it also created an entirely new middle class that did not exist before: engineers, technicians, office workers, skilled laborers. The upheaval was brutal for the people caught in it. It was not the destruction of the middle class, but its restructuring. New roles emerged that nobody could have imagined beforehand.

Precht tells a story in that context: his son reported from Meta that the company had employees hand over their work data, trained AI with it, and after that the people could go home unemployed. You make the machine smart, and it replaces you.

I cannot independently verify this specific case, so I do not present it as proven fact. But even if it is exactly correct, it describes a transition, not an end state. The critical question is not whether roles disappear. Roles always disappear. The question is whether new ones emerge. And the historical answer to that is fairly clear: yes, they do. Every time. The error is counting the jobs that disappear, the ones you can see, and ignoring the new ones you cannot yet imagine.

Are programmers making themselves obsolete?

In the podcast the thought comes up that programmers are making themselves redundant with AI. And Lanz gives an educational recommendation: do not study computer science at all, the not-really-ambitious computer scientists will be replaced by AI anyway.

And now we reach the part where I agree with him in parts. He has a true core, he just draws the wrong conclusion.

Yes, it is true: whoever studies computer science only because the parents said do something with computers, it is secure, and who actually has no affinity for it, will have a hard time. That is the assembly line worker of the industry. The person who types code the way others tighten screws, without passion, without feel. This role is disappearing. And that is not even something bad. Maybe this person would have made an excellent lawyer, a good craftsman, a gifted teacher. Not everyone has to be a programmer just because it paid well.

But that does not mean “do not study computer science”. It means the opposite. Computer science is like competitive sports. To play in the top league you need talent, passion and a bit of nerdiness. That was always the case. The really good developers were never the ones who did it just for the money. AI does not change that, it just makes this old truth more visible. It raises the bar. And the people who clear it become faster and more productive, one person delivers what used to take five.

Matt Garman, the CEO of AWS, said exactly that. In a podcast in August 2025 he called it, quote, “one of the dumbest things I have ever heard”, to stop hiring junior developers now. His reasoning is on point: they are the cheapest employees you have, and the ones with the best connection to AI tools. And his strongest sentence: how is this supposed to work in ten years if nobody is there anymore who actually learned something. You should keep bringing in fresh graduates and teach them how to build software and break down problems, just like always. At AWS itself around 80 percent of developers use AI in their workflow. This is not the voice of someone exiting. This is the head of the world’s largest cloud provider saying: get in now.

Whoever tells young people today to avoid computer science is confusing the disappearance of the unmotivated with the end of the field. The field has never been more powerful than it is now.

And school?

The podcast lands, like every AI conversation, on education. Do the kids still do their homework or do they let AI write it? Are we unlearning how to think?

Again, differentiation instead of panic helps. AI in school is neither salvation nor doom, it is a tool, and it depends on how you use it.

A teacher who generates an individually leveled worksheet in minutes for three performance levels gains time for what matters: the children. A student who can have a topic explained to them as many times as needed, without embarrassment in front of the class, might learn more, not less. Adaptive tutoring that adapts to each individual’s pace was a privilege of the wealthy for decades. It is getting cheap right now.

Yes, learning will change. But change is not getting worse. And yes, it needs to be guided, by parents and schools. A child who lets AI do the task and thinks nothing about it learns nothing. A child who uses AI as a sparring partner learns faster. The difference is not in the technology, it is in the guidance. This responsibility cannot be delegated to a ban.

And the reflex is not new. The calculator was supposed to kill mental arithmetic. Wikipedia was supposed to devalue knowledge. Google was supposed to ruin our memory. Every time the same alarm, every time it turned out differently. We learned to deal with the tools. We will do it again.

The real German error

There is a point where Lanz and Precht touch on something right without following the consequence. Precht says that all of Germany is looking for a new prosperity model, the old one no longer works. And both agree that you cannot opt out of the AI race, that Germany cannot become an AI-free zone.

Exactly. Only, what follows from that? For the two, it follows worry. For me, it follows something else.

Europe created the world’s first comprehensive AI regulation with the AI Act, in force since August 2024. That is not wrong in itself, rules are needed. But it describes a pattern: we regulate what we did not build ourselves, preferably. The big models are created in the US and China. Europe delivers the legal framework. We are world champions in raising concerns and second place in building.

And this is not a question of party color. The mindset is the same across all camps: first secure, then maybe do. I do not even want to politically exploit this, that leads into the wrong debate. It is about attitude. We should switch to maker mode more often, build, experiment, learn from mistakes, and then regulate intelligently, instead of regulating what does not even exist yet. A bit more of the courage you need when you shape new things instead of just managing them. Whoever only asks what could go wrong builds nothing in the end that they get to decide about.

There is a second blind spot, and it might be the more expensive one: we lack long-term perspective. Europe thinks in legislative periods, others think in decades. Taiwan patiently built a chip industry over decades, without it throwing off prosperity every quarter, and is today the point the whole world depends on. China has long pursued an agenda to strategically build its manufacturing and economy, across changes of government. The US persists with foundational technologies, even through lean phases, until global companies emerge from them. We, on the other hand, get out the moment a topic does not deliver short-term results, and then wonder why we end up watching instead of shaping. Staying with it, even when it does not pay off yet, is not a waste. It is the prerequisite for even having a choice later.

Where the two are right

I make it too easy for myself if I only push back. Being fact-based also means taking good counterarguments seriously. There are two.

The first is energy. The water and power hunger of data centers is real and growing. It is no coincidence that water consumption is the most common reason for local resistance. Whoever dismisses that is making it too easy for themselves. Interesting that, of all people, the Microsoft CEO cited by Lanz, Satya Nadella, says exactly this himself: if the AI industry does not soon deliver clear, measurable social benefit, it will not have a future. Only, this is not a requiem. This is motivation, spoken by someone who invests billions, not by someone who exits.

The second is the proof of benefit. The same McKinsey study that shows 88 percent adoption also says: only 39 percent see a measurable effect on results. Many companies are stuck in the pilot phase. This is true and important. But it describes an early phase, not an end.

We know this pattern. The economic historian Paul David described it for electrification: between the availability of a base technology and its full productivity effect there are often years. Factories first had to reorganize around the electric motor before the gain arrived. With AI we are right there. Adoption first, productivity later. Whoever reads the gap between them as proof of failure has not read the history of electricity.

A request to the two

And here I want to be honest, even at the risk of sounding presumptuous. Lanz and Precht reach more people with this episode than I ever will. Millions listen, many form their opinion about a technology they have never used themselves at exactly these kinds of conversations. That is enormous reach. And with reach comes responsibility.

Listening to two smart men worry for two hours is stimulating. But it would be so much more valuable if they used that same force to point forward. Not sugarcoat, criticism belongs, the energy hunger should be named, the distribution question too. But whoever has such a stage could use it to ask the constructive questions: how do we build this in Germany, instead of just commenting on it? How do we train people for the new roles, instead of telling them not to study? What does a social system look like that carries through an upheaval? Those are the conversations this reach deserves. Worry is cheap. Orientation is the real work.

To conclude, the contradiction Lanz delivers himself

There is a moment in the conversation that brings the whole thing to a point. Lanz says that AI-generated texts are not very creative, there is a threat of homogenization, of devaluation. And a few sentences later he tells about an AI text that impressed him. Literally: the killer text, a really good one, many journalists could not do that.

So which is it? Boring or the killer text?

The contradiction is not a mistake, it is the core of the confusion. Creativity was never a property of the tool, but of whoever wields it. A boring text comes from boring thinking, not from the instrument. Precht even says it himself, just without realizing it: AI cannot write a book like Arno Schmidt’s “Zettels Traum”, no “Finnegans Wake”. Correct. And that is exactly my point from the computer science section. The top end stays human. The middle ground gets a tool. Whoever has nothing to say just says it faster with AI.

The real lack of creativity in this debate lies elsewhere. It lies in building the end of a technology from a single applauded campaign sentence in Florida, while one billion people use that technology. That is the template. Skepticism anecdote, grand gesture, the hype is over. We read this text in 1998 about the internet, in 1900 about the automobile, and somewhere along the way surely also about the railway.

The AI hype is not over, because it was never just a hype. A hype disappears when people lose interest. A foundational technology remains because people start using it. And they are using it, billions of people, every day, while someone on television explains that it has not arrived yet.

Humanity has adapted to every one of these waves. We will adapt to this one too. Not because a podcast says so, and not because I say so. But because we always have. The applause in Florida will fade. The data centers will be built. And in a few years someone will wonder how anyone could have believed it was already over.

Sources