AI Manufacturing: Merck Leads the Way – Is the Mittelstand Sleeping?

KI & Automatisierung · 23. Februar 2026 · Manuel Krapf

AI manufacturing is a reality. While chemical giants like Merck KGaA are upgrading with AI robotics, many are hesitant. Learn what this pact truly means for your business.

Do you remember the chemistry sets from your childhood? A little bubbling, a slight discoloration, maybe a strange smell – that was already a huge success. In the laboratories of German big industry, to be honest, it didn't look fundamentally different for decades. Just with more expensive glass flasks and博士-level employees in white coats meticulously mixing formulations. A process characterized by trial, error, and a lot of patience.

But those days of manual mixing, pipetting, and hoping are definitively over. What was sealed in January 2026 between Merck KGaA in Darmstadt and a startup called ChemLex from Singapore is more than just another high-sounding 'Memorandum of Understanding.' It's the burial of the old R&D paradigm. It's the starting gun for a new era in production, whose echo will reverberate far beyond the chemical industry into the workshops of the German Mittelstand. The only question is: Is anyone listening there?

AI Manufacturing is Not Future Music – It's the New Reality

Let's be clear. Merck, one of our German chemical behemoths, is bringing an AI-controlled robot armada into its facilities with ChemLex. The thing is: they are no longer primarily having humans mix substances, but algorithms and high-precision automated systems. ChemLex provides a platform that combines a fully automated laboratory with Artificial Intelligence. The goal? To accelerate research and development (R&D) so radically that competitors will only see their taillights. It's about developing new materials for the semiconductor or automotive industry not in years, but in months. And with a precision and reproducibility that a human can only dream of.

This is not just a gimmick for a rich corporation. This is a paradigm shift. Merck calls AI a 'critical enabler' for solving 'complex scientific challenges.' Translated, this means: without AI, nothing will work for us anymore. And they are not alone. According to current figures available to me, AI already drives 46% of all simulation workloads in the R&D teams of the European chemical, energy, and semiconductor industries. 42% of these teams already use AI-native platforms. This is no longer a niche; this is the new standard at the top.

The Difference: Speed, Cost, and Precision in Numbers

To make this tangible, we need to move away from the image of the laboratory and look at the cold facts. It's about hard economic advantages. During my last visit to a medium-sized mechanical engineering company near Stuttgart, the managing director complained that a development cycle for a new component easily takes 18 months. 18 months! In that time, a competitor relying on AI simulation has already gone through three iterations and is on the market with a better product.

What the Pioneers Say – and What They Conceal

Of course, Merck's corporate speak is only half the story. 'Prerequisite for solving complex scientific challenges' – sounds good on any PowerPoint slide. But what's behind it? Last week, I spoke with Dr. Martin Graf, an analyst who has been working in the chemical industry for 20 years. He put it succinctly: 'Klaus, what we're seeing here is the AlphaFold moment for materials science. Just as AI revolutionized protein folding, it will reinvent material development. Those who miss the boat will be relegated to suppliers for the innovators.'

We are facing a transformation in chemistry that will be as disruptive as AlphaFold was for biology. Companies that don't act now risk becoming technologically irrelevant in a few years.

— Dr. Martin Graf, Industry Analyst at ChemConsult

Also interesting is a look at the competitor Syensqo (the former Solvay division) in Belgium. They are already a step further and use SYGROW, a generative AI solution that not only helps in R&D but also in sales – for lead generation and analysis of customer requirements. The head there, Ms. Colegrave, openly speaks of the guardrails: 'Human-in-the-Loop' and a clear 'no surveillance' principle. They even coordinated this with the European works councils. That's the smart way: introduce technology, but bring people along. A point that many German Mittelstand companies (unfortunately) often overlook.

Automation in the Mittelstand: Between Aspiration and Reality

This all sounds wonderful, doesn't it? The beautiful new world of AI manufacturing, where robots work for us and algorithms show us the way to the next billion-dollar business. But let's get down to brass tacks. The cart is being put before the horse here again. The reality in most medium-sized manufacturing companies I visit – whether in Ostwestfalen-Lippe or the Black Forest – looks different. They are struggling with supply chain problems, exploding energy costs, and a shortage of skilled workers that brings tears to your eyes.

Does anyone really believe that a hidden champion with 250 employees can simply manage a partnership with an AI startup in Singapore? The reality is tight budgets, an IT infrastructure that sometimes dates back to the Helmut Kohl era, and above all: unstructured, dirty data. AI is only as good as the data it is fed. And the biggest digital garbage dump, with the best AI, only produces faster, more expensive garbage. I strongly doubt that it's as simple as the glossy brochures promise. The danger is huge that the Mittelstand will lose touch here – not out of unwillingness, but out of sheer overwhelm.

Q&A: What Does This Specifically Mean for My Business?

Here are the questions that managing directors ask me every week – and my brutal answers.

Is an investment in AI manufacturing profitable for an SME?

Not if you try to copy Merck. But absolutely yes, if you start small. A pilot project in optical quality control, AI-supported optimization of your setup times, or intelligent sales automation – these are projects with an ROI within 12 to 18 months. The question is not whether it's profitable, but where you start.

Where do I get the necessary skilled workers?

Nowhere. At least not on the open market. You have to qualify your own people. Train internal 'AI champions' who act as a bridge between departments and external service providers. And look for partners – universities, Fraunhofer Institutes, or specialized consultancies that understand the German Mittelstand. You don't need a data scientist from Google; you need an engineer who understands how your machines tick and what AI can improve.

My data is a mess. Can I still start?

You must. Don't wait for the perfect database. Start a data cleansing project in parallel with your first AI pilot project. Use the pilot project to define what data you REALLY need and in what quality. This creates focus. Otherwise, you'll get bogged down in years of data spring cleaning while the competition overtakes you.

1. Step 1: Brutally honest inventory. Forget AI for a moment. Where exactly are you losing money, time, and nerves in your production or sales? Identify the 2-3 biggest pain points. Don't guess – measure! Talk to the people at the machine and in the field.
2. Step 2: Start a lighthouse project. Select ONE of these pain points and define a clear, measurable goal. Example: 'Reduce the scrap rate in line 3 by 15% through AI-supported image recognition.' This project must deliver initial results within 6 months. It's about a quick win that convinces the organization.
3. Step 3: Data diet instead of data palace. Identify the MINIMUM necessary data for your lighthouse project. ONLY focus on this data. Ensure it is clean, accessible, and structured. The rest comes later. Perfection is the enemy of progress.
4. Step 4: Get external horsepower on the road. You don't have to reinvent the wheel. Find a service provider or research institute that has already solved similar problems in the Mittelstand. References are everything here. Don't let them sell you castles in the air.
5. Step 5: Make your employees accomplices. Communication is everything. This is not an IT project; it's a cultural change. Explain what you're planning, what the benefits are for the individual (less monotonous work, more focus on problem-solving), and take fears seriously. An AI system sabotaged by employees is the most expensive misinvestment of all.

The Foundation: Your Ideal Customer Profile (ICP) Before you invest millions in AI production, you need to know FOR WHOM you are producing. Every AI-powered strategy – whether in R&D or sales – begins with a razor-sharp understanding of your ideal customer. Our ICP Playbook helps you lay precisely this foundation.

Honestly: most Mittelstand managing directors I speak with recently still wave off the topic of AI. 'Too expensive,' 'too complex,' 'not for us,' or my favorite: 'Our customers don't want that.' This is a dangerously misguided assessment. Your customers may not want AI, but they want faster delivery times, better quality, and lower prices. And that's exactly what AI-powered competition will deliver.

Merck and Co. are currently creating facts. They are using their financial strength to build a technological advantage that will be insurmountable in a few years. I bet that by 2030, we will see a two-class society in German industry: the AI-integrated, who are agile, data-driven, and highly efficient – and the left behind, who will become mere workbenches and contract manufacturers for the former. There's no getting around it. The question for you is not whether you jump on the bandwagon, but how quickly you secure a first-class ticket. And that train is leaving the station at high speed.