Startups and Systems Engineering: The Record That Misses the Mark

In the first half of 2026, 3,053 startups were founded in Germany—more than in any previous half-year. That’s 52 percent more than in the second half of 2025, and more than in all of 2024. Anyone looking for signs of a revival in German industry might see this figure as proof. Unfortunately, the situation is more complicated. The record is largely driven by a single category (take a guess which one?). And ironically, precisely where Germany needs industrial strength, little new is emerging.

Startups are instrumental to this revitalization. Major breakthroughs rarely come from within an established corporation, but rather from outside—from small teams unburdened by legacy issues. This applies to software, to AI, and it applies to systems engineering, which is what makes complex products manageable in the first place. My thesis: The startups are there, but the structural conditions in Germany are not.

My German-American Perspective on Startups

I draw this comparison from my own experience. From 1995 to 2005, I worked in the U.S. for ten years, almost exclusively for startups. After completing my master’s degree at MIT, I moved to San Francisco in 1998 to join the internet startup Post Communications. I was the tenth employee. Two years later, there were 80 of us; then a corporation bought the company, and a year after that, it went bankrupt. This was followed by other startups on the East Coast: Xpogen, which went bankrupt, and Vitae Pharmaceuticals, which was acquired by Allergan.

That was a roller-coaster ride—nothing like the German startups I know, have worked for, and have founded myself. German startups operate with significantly less venture capital; some operate without any at all. Employee stock options outside the founding team remain the exception. And when venture capital is involved, investors are more cautious.

Start a New Business or Reinvent Yourself

It is possible for a corporate group to reinvent itself. There are successful examples of established companies that restructured their business models before the market forced them to do so. In practice, however, starting a new company is the cleaner approach. A young company doesn’t have to fight against its own profit margins. It doesn’t have to protect an existing product line, reassure a sales organization, or convince a works council. The conflict of interest that fundamentally stifles innovation simply doesn’t exist.

Some corporations have realized this. Bosch, for example, has been known for years for its own incubators known for fostering new ideas outside the core business. This is a major corporation’s honest response to a dilemma: The best ideas need a place free from the constraints of the existing business. That is exactly what a spin-off does by its very nature. That is why the startup landscape is not a side issue of industrial policy, but its driving force.

A record, largely driven by AI

The main driver behind the new figures is artificial intelligence. One in three new companies is involved in AI, 1,038 out of 3,053. However, very few of these companies develop their own AI models. Instead, they build on existing engines—mostly from the U.S.—and create an application on top of them.

This makes good business sense but is strategically risky. The business is built in Germany, but the valuable IP remains elsewhere. Anyone who builds only the top layer is dependent on the layer below it—in terms of price, availability, and roadmap. Aleph Alpha It was once the hope for a German foundation model—the European counterpart to OpenAI. The company has since been acquired by the Canadian firm Cohere in a deal described as a merger. Cohere’s shareholders hold about 90 percent of the combined company, while Aleph Alpha holds only 10 percent. Even the Schwarz Group, previously the most important German investor, is now putting its $600 million into Cohere rather than into the German development. The best-funded German attempt to build its own AI foundation is thus effectively under Canadian control. And it’s not an isolated case: With Tubulis and Contentful, two German unicorns have recently been acquired by U.S. corporations.

There are counterexamples, and tellingly, they aren’t in the AI spotlight. The Munich-based nuclear fusion startup Proxima Fusion has just raised 411 million euros and reached a valuation of over 2.4 billion euros. It is building its own hardware, not a layer on top of someone else’s model. Raiqon is led by the team that once built Codebeamer and successfully sold it to PTC—a company with real substance in the requirements and application lifecycle space. Certivity The Munich-based company has raised 13.3 million euros in a Series A round and transforms complex regulations into structured, machine-readable requirements. This is industrial IP in the truest sense—deeply embedded in the engineering process and not easily replaceable.

The best examples can be found in systems engineering

Anyone in Europe looking for this kind of substance will find it in systems engineering. Sensmetry Based in Vilnius, the company is working on the security and reliability of complex autonomous systems and is launching Syside, a tool for SysML v2, the new standard for model-based engineering. Spicy SE takes a completely new approach to the MBSE tool, rather than continuing to maintain a thirty-year-old toolchain. Together with Certivity and Raiqon A small European community is emerging here that is working on the very tools the industry uses to develop complex products.

These companies share one thing in common that sets them apart from AI application startups. They’re building something that no one else can replicate without deep domain expertise—a requirements tool, a safety verification tool for autonomous systems, a SysML v2 modeler: This isn’t just a thin layer on top of someone else’s model, but proprietary, defensible intellectual property. Germany needs a large number of startups exactly like these. They do exist, but there are too few of them, and many of the best have long since moved elsewhere.

A Departure That Was Foreseeable

After all, the overall conditions remain poor. There is too little venture capital in later rounds, too much red tape when starting and growing a business, employee stock ownership plans that remain unattractive from a tax perspective, and public procurement that effectively excludes young providers. Those who want to scale quickly will find a more favorable environment in the U.S. The result is not a footnote, but a measurable drain on resources.

Dalus is an AI-native MBSE tool and is currently part of the U.S.-based Y Combinator ecosystem. Among its founders is Sebastian Völkl, who previously founded a company in Europe. Flow Engineering is charting its own course beyond SysML and building in the U.S. Planetary Utilities The company was founded by Johannes Gross from Germany and Robert Karban from Austria and is based in California. Three teams with European roots are working in systems engineering—the core of industrial value creation—and are now building their capital in the U.S. This is the price of poor framework conditions, played out in specific companies.

Automotive: The Costliest Oversight

Nowhere is this failure more acute than in the automotive industry. It accounts for 3.5 percent of Germany’s GDP, compared with 0.6 percent in the U.S. For Germany, this is not just one sector among many, but the very core of its industrial base. It is precisely in this sector that too little has been invested in innovation over the decades. Between 2011 and 2018, approximately one billion U.S. dollars was invested in automotive startups in Germany. In the U.S., the figure was 56 billion; in China, it was over 30 billion. The gap is not just a little; it is 30 to 60 times larger.

Over the seven-year period from 2011 to 2018, the U.S. invested US$ 56 billion in the automotive sector, while China invested over US$ 30 billion. Germany, on the other hand, invested just one billion.

Florian Nöll, Chairman of the Board of the Federal Association of German Startups e.V.

The structure of today's startups exacerbates the situation. Hardware accounts for only about 11 percent of the mobility sector; nearly one in five mobility startups is working on automotive software. Among the five best-funded mobility startups is the Swedish battery manufacturer Northvolt (bankrupt) Only one is based in Europe; the rest are spread across the U.S. and Asia. Per capita, mobility investment in the U.S. is three times higher than in Germany. The German industry is thus lagging behind in two ways: historically, due to decades of underinvestment, and currently, because new capital is flowing primarily into software and rarely into hardware—the very area on which Germany’s strength has always depended. Startups could make up for these shortcomings if the industry and lawmakers allow them to.

Conclusion

It’s good to see so many startups, and the record set in the first half of 2026 is more than just a statistic. It shows that the will is there. But the number alone isn’t enough. Germany needs fewer superficial AI applications based on third-party models and more industrial startups that build their own, defensible IP: in systems engineering, in hardware, and in the tools used to create complex products. These companies already exist—in Vilnius, in Munich, and in the requirements and MBSE sectors. But too many of the best are moving to the U.S. because the environment there is easier.

So the ball is now in the legislature’s court. Venture capital for later rounds, attractive employee stock options, less bureaucracy, and public procurement that allows young providers to participate: These aren’t new demands, but their urgency grows with every team that sets up shop in California. There remains a huge amount of ground to make up, despite the high number of startups. Anyone who wants to see a revival of German industry must not be blinded by this record. What matters is not how many companies are founded, but which ones—and where they ultimately raise their capital.