Nicolas Sauvage is not pursuing the most glamorous fads in the field of artificial intelligence. Rather he is making calculated bets on what most people would term as the boring parts of the AI stack- the infrastructure, hardware, and systems that silently drive all other parts of the stack. His philosophy is straightforward yet challenging: sometimes, the most effective investments may require approximately four years to demonstrate how worthy they are. It is a long-term outlook in an industry where it is common that people are obsessed with the rapid hype cycles.
The belief of Sauvage is based on a profound knowledge of the way in which the technological revolutions in reality happen. Although headlines are more likely to hail the applications of chatbots, AI assistants, and creative tools, the true leverage is often found behind the scenes. Scalability is decided in infrastructure, costs managed, and performance breakthroughs achieved. This is why one of his most notable early bets was on Groq, a company that focuses not on training AI models, but on inference—the process that occurs every time an AI system generates a response.
Groq was founded by Jonathan Ross and was a radically focused chip design company. Ross started with the compiler, the software layer that converts code into instructions, and designed hardware around it. What emerged was a simplified architecture that was speedy and efficient. Sauvage realized that this strategy would prove to be a significant challenge as the application of AI increased. The larger market, however, was not yet in the swing at the time. As of 2020, prior to the mainstream adoption of generative AI, such a bet could have been viewed as niche or even risky.
What Sauvage realized, though, was that demand in inference would increase exponentially. Each interaction between an AI model and an application necessitates a computation and as the number of applications increases, so must the efficiency of computation. However, today, with the emergence of AI agents that are able to perform multi-step tasks, the computational load has grown exponentially. It is now possible to have a single user request that can invoke dozens of backend operations. This transformation has confirmed the initial hypothesis of Sauvage the actual bottleneck- and opportunity-maker is the boring layer.
The constraints of the conventional consumer hardware markets also shape his line of thought. Smartphones, laptops and various other devices will reach points of saturation later on where growth becomes slow and differentiation becomes incremental. Conversely, there is no definite limit to AI infrastructure. New models, applications, or use cases all add to the demand. By targeting this layer, Sauvage was able to position his fund such that it would receive a structural, long-term trend as opposed to a short-lived wave.
Remarkably enough, the genesis of TDK Ventures itself is indicative of a comparable readiness to break the mould. The parent company, TDK Corporation, has traditionally focused on such products as magnetic tapes and electronic components. The first name that comes to mind when one considers state-of-the-art venture capital is the first one. Even Sauvage himself admitted the possibility of the existence of the venture arm was improbable. An executive in France who operated in Silicon Valley, yet was not fluent in Japanese or had a base in Tokyo, he nevertheless drove the idea internally.
Sauvage pitched himself to the leaders of TDK, after giving two lectures at Stanford, one of which was a plea in favor of corporate venture capital, and the other an outline of why these attempts usually fail. Although he did not have formal power, he did not give up. At some point, he received permission to start a fund with a definite guiding question: what technologies could define the future of TDK, and what technologies could completely disrupt TDK?
This two-lensed—opportunity and existential risk seeking—portfolio has impacted the portfolio in essential ways. In addition to AI chips, TDK Ventures has also invested in other areas that are only now receiving a wider audience. They consist of next-generation energy technology like solid-state grid transformers and sodium-ion batteries aimed at data centers. These technologies might not make any headlines the same as consumer AI tools, but they tackle the important issues of scalability, sustainability, and geopolitical resilience.
As an illustration, alternative battery chemistries that minimize dependence on Lithium and cobalt are gaining even more prominence as the global supply chains become uncertain. Equally, power infrastructure innovations are necessary to sustain ever-increasing energy requirements of AI systems. The very basis of the entire AI ecosystem, data centers, must not only be efficient in terms of their performance, but also have reliable and sustainable energy sources. Investing in these underlying technologies, Sauvage is actually making a bet on the facilitators of the AI revolution instead of its apparent results.
Another key aspect of his strategy is timing. The four-year rule by Sauvage takes into consideration that the transformative technologies usually pass through phases of underestimation before attaining mainstream acceptance. Investments made early may seem insignificant or even incorrect but when the ecosystem in general catches up, they become very valuable. This takes time, faith, a readiness to act before the majority is convinced, and these are not always simple to sustain in venture capital.
Meanwhile, luck is a component of this. Even Sauvage himself acknowledges that time and other external factors can play a part. The sheer pace at which AI has been adopted in recent years has increased the significance of infrastructure in a way that few could have fully anticipated. However, fate has a way of rewarding those whose position has been put in consideration. By considering long-term needs over short-term trends, Sauvage was more likely to have his bets meet future demand.
Ultimately, Nicolas Sauvage’s approach offers a counter-narrative to the hype-driven culture of modern tech investing. He makes an accent on durability rather than on visibility. He does not put his emphasis on the things that are exciting today, but rather on those that will be indispensable tomorrow. His track record with companies such as Groq shows that the most effective innovations are also the ones that are invisible.
