Bosch Makes Chips: Inside Its $1.2 Billion Dresden Wafer Fab

Most people know Bosch for drills, dishwashers and car parts. Fewer know it makes the chips inside many of them. In June 2021, Bosch opened a 300-millimeter wafer fab in Dresden that it called the biggest single investment in its more than 130-year history: roughly one billion euros, or about $1.2 billion at the exchange rate of the time.

The factory is impressive. The more useful question is the one every operator faces after a large technology bet: how to know whether it is paying off.

Bosch has made chips for decades

Dresden was not a sudden move into a new industry. Bosch says it started producing power diodes in 1958, and its Reutlingen plant has made application-specific chips since 1970, components that Bosch says were not commercially available. Reutlingen moved to 150-millimeter wafers in 1995 and 200-millimeter wafers in 2010, and today makes automotive chips, MEMS sensors and silicon carbide power semiconductors.

At the 2021 opening, Bosch said it had invested more than 2.5 billion euros in its Reutlingen and Dresden wafer fabs since 2010. Dresden added scale and a newer wafer size to a capability Bosch already had, rather than creating one from nothing.

What Bosch built in Dresden

Bosch opened the Dresden fab on June 7, 2021, with 72,000 square meters of floor space and 250 people already working there, a number it planned to grow to roughly 700. It said automotive chips would be the plant’s primary focus.

The first customer was Bosch itself. Bosch said chips from Dresden would go into its power tools from July 2021, six months earlier than planned, with production for automotive customers starting in September, three months early. Those were the dates Bosch announced at the opening.

Data, AI and the digital twin

Bosch calls Dresden its first AIoT factory, combining artificial intelligence with connected machines. It said the plant would generate production data equivalent to 500 text pages every second. Bosch says AI looks for tiny error patterns on the wafer surface, called signatures, so their causes can be corrected before they affect the product, and predicts when equipment needs maintenance rather than following a fixed schedule.

The division of labor is spelled out. In Bosch’s words: "The AI algorithm determines when intervention is necessary; humans decide what needs adjusting."

The plant also exists twice. During construction, Bosch recorded the building in a digital twin of roughly half a million 3D objects, from ventilation and cable ducts to production lines, so it can simulate process changes and renovation work without interrupting production. And with camera-equipped data glasses, an equipment specialist in Japan can guide a technician in Dresden through a repair in real time.

All of these are Bosch’s descriptions of how the factory works. Bosch has not published yield, uptime or cost figures that show what each capability delivers.

Was it worth the money?

Bosch does not publish the Dresden fab’s returns, so the honest answer is that nobody outside Bosch can say. Group results do not settle it either. In 2025 the Bosch Group reported sales of 91.0 billion euros and an EBIT margin from operations of 2.0 percent, and spent 7.9 billion euros on research and development. Bosch said it continued to make substantial investments in areas including semiconductors. None of that isolates one factory.

If I were evaluating a bet like this from the inside, I would not start with the group margin. I would ask what the factory was supposed to solve: whether it secured chips for Bosch’s own products when supply was tight, what its chips cost compared with buying them, and how quickly it can bring a new product into full production. Bosch has published an early answer to the last of those, the ramp that started ahead of schedule, and very little on the others.

The lesson: judge the capability against the problem

Large technology investments are easy to admire and hard to evaluate. A cleanroom, a digital twin and AI-assisted monitoring are visible. The return is not, especially when the factory supplies its owner’s own products and no market price shows up anywhere.

It is the same question I asked about Walmart buying Jet.com: the asset is what you can see, but the value sits in the capability it gives you. The way to judge that capability is against the specific problem it was meant to solve, measured from outside the system that produced it.

How would you know if your own biggest technology investment was paying off?