Bold claim: Europe could finally win its own AI-age by building a world-class, energy-efficient compute stack from the ground up—despite a noisy, costly reality that challenges that dream every day.
In the AI Innovation Center on Eindhoven’s High Tech Campus, Bernardo Kastrup speaks with the calm of someone who has lived many lives: computer scientist, philosopher, ASML strategist, founder. Now, with a touch of reluctance, he accepts the role of figurehead for Europe’s most ambitious attempt to reinvent AI computing from scratch.
Euclyd, the company he launched just over a year ago, has quickly become a sensation. Supported by industry heavyweights—former ASML CEO Peter Wennink, renowned Intel engineer Federico Faggin, Silicon Hive’s Atul Sinha, and Elastic founder Steven Schuurman—the startup burst onto the European tech scene this year, instantly steering conversations. It wasn’t merely the ambition; it was the tone: confident, technical, unapologetically large—Europe speaking with Silicon Valley audacity.
"Yes, you can call it bragging," Kastrup laughs. "It’s tongue-in-cheek. But confidence matters. Why should Americans be the only ones allowed to dream big?"
Beneath the humor lies a deeper frustration that sparked Euclyd’s creation.
"I looked around and asked: Who allowed this to happen?" he recalls.
When generative AI exploded into public awareness in 2023, Kastrup began surveying Europe’s landscape. He saw top-tier researchers, cutting-edge chip equipment, a sophisticated industrial base—and no serious push to design the next-generation silicon powering future data-center AI.
"I remember looking for someone to blame," he says. "Why is nothing happening in Europe? How did we let this happen?" The harsh truth followed quickly: there was no one else to blame. "Who could do something about it? People like me. People I know. So yes, I felt a responsibility."
That sense of duty propelled him into months of stealth. Euclyd would stay quiet until the idea could prove itself reliable; proof trusted by him and his peers. "When you take money from a VC and fail, that’s the risk of business," he notes. "But losing friends’ money? That’s another matter."
For months, a small team—many veterans from earlier chip ventures—tackled a fundamentally new AI-inference architecture. No GPU heritage. No repurposed gaming hardware. No shortcuts. A system designed from the gate level up. Early design work even happened in Kastrup’s attic, where he set up a simulator and spent long nights sketching microarchitectures the way others sketch ideas in notebooks.
When the test chip finally emerged and Samsung agreed to manufacture it, Euclyd stepped into the light.
A European chip with world-scale ambition
Euclyd’s promise is monumental: ultra-low-energy inference, potentially up to 100 times more efficient than Nvidia’s current data-center chips, achieved not by magic but through architectural rigor.
"Nvidia built for video games," Kastrup argues. "When large language models arrived, they were in the right place at the right time. But treating a neural network as a video game with a global variable space—that’s one of the worst approaches for efficiency."
Euclyd’s architecture flips the script: no reuse of off-the-shelf IP, no reliance on generic bus schemes. Everything is specialized, deeply pipelined, and optimized for neural inference first. The test chip packs 64 processors; the production version is planned to scale to 16,384.
In a landscape of incremental improvements, Euclyd dares to break the mold: a genuine reset rather than a tweak.
The dream: a European Nvidia. The reality: global capital
Yet, behind all the technical swagger lies a topic that makes Kastrup pause and choose words with care: money.
Euclyd is pursuing a major funding round. With it comes the possibility—perhaps the likelihood—that the company may no longer be exclusively European in ownership or outlook.
"The original dream was to do AI in Europe," he says softly. "To let Europe play in that space. But when building a business, and spending other people’s money, fiduciary duties come first. That may pressure us to act in ways that aren’t fully aligned with the dream."
It isn’t hypothetical. It’s unfolding now. "We are raising money," he concedes. "This is a situation we may face—or have already faced. That’s part of the game."
He pauses again. "I can dream, right? I can dream anything. But I cannot, on my own, guarantee that the dream will become—or stay—true. The world is not under my control."
That honesty stands out in a European tech scene that often clings to idealism. Kastrup won’t pretend the tension doesn’t exist. He envisions Euclyd as Europe’s Nvidia—rooted in European engineering and the Brainport ecosystem that shaped him—yet he also needs the company to survive in a global race for compute, one that Europe has joined late and perhaps underprepared for.
If the right investors come from outside Europe, and rejecting them would threaten the company’s competitiveness, what should a founder do?
"I’ll do the best I can within the boundaries of what is allowed or required in this role," he says. "But cannot control everything."
A dream worth fighting for
Whether Euclyd remains European in ownership is uncertain. Whether its technology will be adopted globally seems increasingly likely. And whether Europe finally gets its own AI-compute champion may depend as much on Europe’s willingness to move at global speed as on Kastrup’s technical prowess.
Still, the dream endures. "Europe needs to participate in this field," he insists. "That’s my wish as well."
With that, Bernardo Kastrup—philosopher, engineer, founder—returns to his office, where a tiny test chip, no larger than a fingernail, waits. Its architecture could reshape AI’s global energy footprint. The question remains: will Europe still be the place where that future unfolds, or will the world move faster than its aspirations?
