Nicolas Sauvage bets on the boring parts of AI

Nicolas Sauvage thinks it takes four years for the best bets to look smart — a thought he shared on stage last week at the StrictlyVC event in San Francisco, which TDK Ventures was co-organized.

It’s a theory he’s been trying to prove since 2019, when he founded the venture capital arm of the Japanese electronics giant, which now manages $500 million across four funds. The AI ​​chip startup Groqvalued at $6.9 billion in its most recent funding round last fall, is the highest-profile example of this thinking.

In 2020, long before the AI ​​boom made infrastructure bets a useful place to funnel capital, Sauvage wrote a check to the company, which was founded by Jonathan Ross — one of the engineers who built Google’s Tensor processing units. Groq focused from the beginning on inference: the computational heavy lifting that happens every time a model answers a query. Ross had designed his chip by building the compiler first, stripping away the architecture until, as Sauvage describes it, “you can’t remove a part and still have it work.”

It may have seemed niche to some, but knowing what he was doing about the limitations of his parent company, Sauvage saw opportunity. Unlike consumer hardware, which has a natural ceiling, the demand for inference continues to worsen with each new application and each new model. Sauvage couldn’t have known at the time that demand for inferences would soar this year, thanks to each AI agent planning and acting on dozens of calls (where a single query was enough).

But somehow, Ross got lucky too. After all, a Japanese electronics group best known for magnetic tape doesn’t seem like the most natural investment partner. In fact, Sauvage describes the very existence of TDK Ventures as highly unlikely. But after two back-to-back lectures at Stanford—one making the case for corporate VC, the other listing all the reasons it fails—Sauvage, who is French and joined TDK in Silicon Valley through an acquisition, pitched the idea to senior executives at TDK headquarters even though he had no clear position to do so. (“I’m not Japanese. I don’t speak Japanese, I don’t live in Tokyo,” he told this editor.)

After refusing to take no for an answer, he finally got the go-ahead to raise a fund whose mandate was to answer one question: What is the next big thing for TDK, and what might kill it?

The portfolio he has amassed since then is dotted with technologies that have become of wider interest to VCs in the last year: solid-state grid transformers, sodium-ion batteries for data centers, alternative battery chemistries that bypass the geopolitical fragility of lithium and cobalt.

The discipline behind all of this is the same: identify the bottleneck four years out, then find the founders who are already working on it.

The question, of course, is what comes next. For his part, Sauvage is keeping a close eye on natural artificial intelligence—not all robotics but robots with a very specific job to do. Agility Roboticsfor example, in his portfolio, he focuses on the unique, mundane task of moving things from one place to another in warehouses facing labor shortages. Another holding company, the Swiss holding company ANYboticsbuilds rugged robots for environments too dangerous for workers – places where the definition of work is essentially going where humans can’t. The through line is clarity of purpose. Robots that Sauvage is betting on, don’t try to do everything. Instead, they do a difficult thing reliably.

Sauvage says he’s watching the computer stack change again. GPUs have dominated training — the massive, parallel computation of teaching a model. Inference chips like Groq’s are reshaping what happens when this model talks: faster, cheaper, at scale. Now, Sauvage argues, CPUs must be reborn. They are not the most powerful chips or the fastest. But they are the most flexible and best adapted to the branching decision-making logic of orchestration. When an AI agent assigns a task, checks its progress, and circles back dozens of steps, something has to manage the entire choreography. This something, more and more, looks like a CPU.

And then there is China. A recent report from Eclipse—a closely watched venture capital firm—documents what Sauvage describes as “vibe manufacturing”—the rapid, AI-assisted iteration of physical hardware prototyping, mirroring what vibe coding did for software. Chinese manufacturers, according to the report, are compressing the design-build-test cycle for physical products in ways that Western supply chains are not yet equipped to match.

For Sauvage, it’s a brand — and one he’s acting on with TDK Ventures’ various investments. One problem that remains unsolved, he says, is dexterity. Models are improving fast enough that natural artificial intelligence seems inevitable. What is still missing is the physical fluency to match. Countries and companies that understand how to iterate people as fast as others iterate code will have a manufacturing advantage. That’s the wave he’s positioning TDK Ventures for today.