Nicolas Sauvage is betting on the boring part of AI

Nicolas Sauvage believes it takes four years for a good bet to appear – thinking he shared last week at StrictlyVC’s San Francisco event, which. TDK Ventures to work with them.

It is an idea that he has been working to prove since 2019, when he founded the Japanese electronics company, which now manages $500 million in four funds. The introduction of the AI ​​chip Grokprecious $6.9 billion during its most recent funding round last fall, it is the most high-profile example of this concept.

In 2020, before the start of the AI ​​boom made development bets visible, Sauvage wrote a check in the company, which was founded by Jonathan Ross – one of the engineers who created the Google Tensor Processing Units. Groq focused from the beginning on a concept: the heavy lifting that happens every time the model answers a question. Ross had created his chip by building the first component, then removing the components until, as Sauvage explains, “you can’t remove one part and still work.”

It may seem like a good idea to some, but knowing what he did in his parent company’s troubles, Sauvage saw an asymmetry. Unlike consumer devices, which have a natural ceiling, the demand for information continues with each new software and each new model. Sauvage would not have known then that the need for information will explode this year, thanks to each AI assistant who organizes and makes several calls (where one question is enough).

But somehow, Ross got lucky, too. After all, the Japanese electronics group best known for magnetic tape is not, on the face of it, a well-known investment partner. In fact, Sauvage describes the existence of TDK Ventures as impossible. But after two interviews at Stanford – one that makes a VC case for the company, the other lists all the reasons why it fails – Sauvage, who is from France and joined TDK in Silicon Valley by purchase, presented the idea to the top at TDK’s headquarters even though he did not have a clear position to do so. (“I’m not Japanese. I don’t speak Japanese; I don’t live in Tokyo,” he told this editor.)

When he didn’t take no for an answer, he got the green light to start a fund whose mission was to answer one question: What is TDK’s next big thing, and will it kill?

The portfolio he has collected includes technologies that have been of great interest to VCs in the past year: hard grid inverters, sodium-ion batteries for data centers, alternative battery chemistries that prevent the weakness of lithium and cobalt.

The discipline behind all of these is the same: identify a four-year barrier, then get the early adopters up and running.

The question, of course, is the next one. For him, Sauvage is focused on physical AI — not all robotics but robots that have a specific job to do. Agility Roboticsfor example, in its history, it focuses on a single, traditional task of moving goods from one place to another in a warehouse that is facing a shortage of workers. Another portfolio company, Swiss portfolio ANYboticsthey build robust robots in dangerous environments for workers – places where the meaning of work must go where humans cannot. This line is the sound of the goal. Sauvage robots are betting don’t try to do anything; instead, he faithfully does one difficult thing.

Sauvage says he’s also seeing changes in the compute stack. GPUs dominated training – massive, parallel computations for training a model. Small chips like Groq’s are redefining what happens when brand speaks: fast, cheap, scalable. Now, Sauvage argues, CPUs need to reboot. They are not the most powerful or fastest chips. But they are very flexible and suitable for branching, making orchestration decisions. When an AI assistant assigns a task, looks at how it’s being executed, and jumps through a series of steps, something has to manage the entire choreography. That something, more and more, looks like a CPU.

And then there is China. A recent report from Eclipse – a company he closely follows – documented what Sauvage describes as “creating a vibe” – a quick, AI-assisted look at hardware, showing what vibe coding has done for software. Chinese manufacturers, the report found, are embracing the practice-build-test of physical products in ways that Western chains are not ready to match.

For Sauvage, it’s a bottle brand — and he’s already on the move with TDK Ventures’ various ventures. One problem that hasn’t been solved, he says, is talent. Models are getting so good that physical AI feels inevitable; what is still missing is good communication to make it work. Countries and companies that figure out how to move atoms as fast as others do on code will have an opportunity to create. That’s the wave that TDK Ventures is setting today.