At the age of 16, founder Gilles Backhus asked his dream company what he should study. Your answer – and what Backhus is doing today.
“We’re not doing any lessons today. Everyone goes into the IT room and sees what they want to do later,” says the teacher. The then teenager Gilles Backhus found the company “Advanced Micro Devices” (AMD), a leading manufacturer of PC hardware from the USA, exciting.
“I was a classic village LAN party gamer: Counter-Strike 1.6, Warcraft 3, you name it. As a result, I also became a huge AMD fan at some point. With every shooting star I really wished that I would work at AMD at some point,” he tells Gründerszene.
And so, at the age of 16, he wrote an email to AMD in school. “Back then I simply wrote to jobs@amd: ‘I want to develop computer chips later – what do I have to study for that?’ And the cool thing is: someone from HR actually answered me and sent me an information folder home.” Backhus then studied electrical engineering at the Technical University of Munich (TUM).
Racing cars and CDTM
At TUM he was involved in Formula Student Racing. This is a group of students who put together their own racing car and compete against other universities. Backhus says his motivation to be active alongside his studies came from his enthusiasm for technology.
The Center for Digital Technology and Management (CDTM) then further encouraged this enthusiasm. The CDTM is a program that supports students at the interface between technology and management.
Here he learned how to “better assess your strengths and weaknesses. And that people from supposedly elite institutions like Stanford or MIT only cook with water.” This insight gave him the decisive impetus to put aside his own reluctance and start something himself, he says.
Difficult time at Lilium
But first we went to Lilium after studying. There he started as a “Sensor Systems Engineer”. The Munich startup Lilium, which is now insolvent, was working on the development of air taxis at the time. Backhus actually wanted to advance AI development at Lilium, but the reality was different: instead of writing algorithms, his everyday life consisted of phone calls with suppliers.
After four and a half months at Lilium, Backhus’ current co-founder RK Anand approached him to build the Tensordyne team in Europe with him – despite an age difference of 29 years. Backhus agreed and left Lilium.
From automotive partner to Nvidia competitor
Tensordyne initially started developing chips for electric cars, with an office in Munich and one in Silicon Valley. According to Backhus, the chip industry for cars was in “big hype” in 2018, there was high customer demand and a lot of investor money. The only problem: As a startup, they would have had to meet automotive standards that were difficult to achieve for a young company with only 40 employees.
Then, according to Backhus, the “ChatGPT moment” came: Due to the boom in LLMs (= AI language models like ChatGPT), they quickly realized that the focus of chips was shifting “massively towards data centers”. That’s why the Tensordyne founders switched from chips for autonomous driving to AI infrastructure for data centers.
The company’s own approach of “logarithmic mathematics” has proven to be an advantage, as it works excellently not only in autonomous driving, but also in the AI area. “Logarithmic mathematics” means that Tensordyne stores numbers as logarithms. Multiplications would become simple additions, making AI calculations more energy efficient. The nuclear technology could continue to be used. Since many of the team came from the “data center and networking world,” they felt comfortable with the pivot.
Tips for a successful pivot
How a successful pivot can be successful for you too? Backhus has three tips for this:
- Find the right time: Balance patience and market feedback – don’t give up too early, says Backhus, but above all, don’t give up too late. Otherwise you’ll be broke
- Switch off the ego: Listen to the market, not your own baby, says Backhus. Mini-pivots are often smarter than big jumps into the deep end
- Test practically: Build transition phases with partners, check market fit and monitor the competition, advises Backhus – just as they explored automotive vs. data centers in parallel
