7 August 2026

MACHINA Summit – Paris, Station F: Highlights from the First Major European Event Dedicated to Physical AI

On July 7, the DATASOLUTION teams attended the MACHINA Summit at Station F, Europe’s first major event dedicated to physical AI. On paper, it was an event for robotics engineers and autonomous systems. In reality, it was a real-world demonstration of a truth that applies to all AI: without well-managed data, no artificial intelligence can deliver on its promises.

What is Physical AI?

Physical AI refers to artificial intelligence systems capable of acting in the real world—such as industrial robots, autonomous vehicles, and embedded systems—as opposed to digital AI, which processes text, code, or images without direct physical interaction.

Its development depends heavily on the availability of reliable training data, which is still scarce compared to the volumes available for digital AI.

3 Takeaways from the MACHINA Summit on Physical AI and Data

1. “AGI Will Come Twice” and Digital Technology Paves the Way

That was the message of the keynote speech by Eric Landau (CEO of Encord): advanced artificial intelligence is deployed in two phases—first in the digital realm, then in the physical realm. The reason for this lies entirely in data.

Digital AI has benefited from an unprecedented abundance of resources: all the text, code, and knowledge available online—what Landau calls the “Great Data Subsidy”—while physical AI still has to build its own datasets. The availability and quality of data determine the order in which new applications become possible.

2. The real value lies in the layer we tend to overlook: data

The focus tends to be on models and spectacular demonstrations. But what doesn’t become commonplace is the layer that translates technological capability into tangible value: structured, reliable data linked to real-world applications.

The investors in attendance made it clear: they no longer fund demos, but rather systems that deliver a measurable ROI. And behind every system that delivers results, there is a clean data pipeline.

These are the “picks and shovels” of AI—the tools that, without making a big splash, make the rush possible.

3. Closed-loop data distinguishes the demo from the production version

One model is never enough. Performance relies on a continuous cycle: act, measure, learn, and adjust. Whether it’s a robot in a factory, a scoring model, a chatbot, or a recommendation engine in production, the principle is the same.

The speed at which an organization validates, corrects, and re-enters its data is directly proportional to the likelihood of its system’s success.

That is, quite simply, what we do at DATASOLUTION.

Our Book Club at DATASOLUTION

Whether AI is physical or digital, the bottleneck is the same: data. Robotics makes this challenge spectacular, while everyday data and AI projects make it decisive.

Without a clean, structured pipeline that’s grounded in reality, even the most impressive demos will remain… just demos.

It is precisely at this intersection that our expertise lies: structuring, ensuring the reliability of, and leveraging data so that models deliver on their promises, from prototype to deployment.

And to wrap up with a quote from Eric Landau: Happiness = Reality – Expectations.

In AI, the real work is to bring reality—data, reliability, and deployment—to the forefront faster than the hype.

Thank you to MACHINA / RAISE Summit for the insightful discussions.