Samsung has co-led a $231 million Series A funding round for Euclyd, a Dutch AI chip startup designing inference processors that abandon the GPU architecture entirely, according to an exclusive report by CNBC.
The roughly €200 million round was co-led by Samsung alongside Somerset Capital Partners, the Scaleup Europe Fund managed by EQT, and Innovation Industries — a syndicate that pairs one of the world's largest memory and chip manufacturers with a bloc of European growth investors. For more context on this story, see our ongoing breaking AI news.
Euclyd, founded in 2024, is building an AI chip system for inference — the computationally intensive work of running trained models — with what the company describes as a fundamentally different processor and memory architecture than the GPUs that dominate AI data centers today.
Why Inference Is the Target
The choice of inference as a beachhead is strategic rather than incidental. Nvidia became the world's most valuable company after its GPUs, originally designed for gaming, were repurposed for both training and running AI models. The company now holds a near-monopoly over the market for the highest-end chips.
But inference workloads are more predictable than frontier-model training, power efficiency per query dominates the economics, and the software surface a challenger must support is far narrower. It is also where the industry's volume is heading as AI moves from laboratories into everyday products.
"AI is becoming a foundation of economic growth, scientific discovery and national competitiveness, but its potential will remain constrained unless we fundamentally change the infrastructure beneath it," Euclyd CEO Bernardo Kastrup told CNBC.
What Samsung Brings Beyond Money
Kastrup was explicit that Samsung's value goes far beyond its checkbook. "Samsung can help us in more ways than money," he told CNBC. "They are one of the biggest memory manufacturers in the world. They do a lot of engineering, they know a lot about systems, they know the supply chain, they have a huge network."
That matters because the binding constraint on Nvidia's challengers has rarely been chip design itself. It has been securing advanced packaging capacity and high-bandwidth memory allocation in a supply chain where the largest buyers book capacity years in advance. An investor embedded in that supply chain can, in principle, unblock the two hardest problems a GPU-alternative startup faces.
The Business Model: Racks and Royalties
Euclyd is targeting two revenue streams, according to the CNBC report. The first is selling hardware and physical rack systems to enterprise customers that want secure, self-hosted AI inference. The second is licensing its intellectual property to other companies that want to build their own chips on top of Euclyd's technology.
The company says its silicon systems for foundational models will reduce the energy needs and costs of AI data center infrastructure — a claim that has attracted funding across the sector but remains unproven in Euclyd's case. The company's systems have yet to be demonstrated at scale in commercial deployments.
Kastrup told CNBC that Euclyd aims to begin rolling out its physical chip systems in 2028, with the goal of serving thousands of enterprise customers by 2030.
A Crowded, Well-Funded Field
Euclyd is arriving into a market where nearly every large player is already trying to reduce its dependence on Nvidia. OpenAI announced in August that its first in-house AI chip, the Jalapeño, had achieved "industry-leading speed and efficiency." Google, Amazon Web Services and Meta all run their own custom AI chip programs. And a wave of venture-backed startups is chasing inference efficiency with novel architectures and memory designs, in a funding environment where Nvidia alternatives have attracted record investment all year.
The round itself landed on a bruising day for the sector: chip stocks slumped on Monday after several AI leaders called for a slowdown in development, underlining how sensitive sentiment around AI hardware has become.
The Skeptic's View: CUDA, History and Design Wins
The obstacles facing any Nvidia challenger are formidable. The incumbent's deepest moat is not silicon specifications but software: the CUDA programming environment, two decades of accumulated libraries, and a generation of machine-learning engineers trained on it. History is littered with well-funded architectures that arrived a generation late or failed to sustain an ecosystem.
Analysts and investors will be watching three markers to judge whether this round was money well spent: whether Samsung's involvement translates into concrete manufacturing and memory arrangements rather than a purely financial stake; whether Euclyd names production customers, and whether those deployments generate revenue rather than serving as research projects; and whether the European backers follow this round with continued support through the long, capital-hungry path to working silicon.
Europe's Bet on Hard Tech
The composition of the syndicate also carries a continental signal. Somerset Capital Partners, the Scaleup Europe Fund and Innovation Industries represent a deliberate European push to fund hard technology at growth stage — keeping strategic semiconductor design teams in Europe rather than watching them sell early to American acquirers. The Netherlands, home to critical semiconductor equipment suppliers and deep research infrastructure, has emerged as a natural base for that effort.
Whether Euclyd becomes the European answer to Nvidia's dominance or another cautionary tale will not be clear until its first systems reach customers. But with $231 million, a strategic anchor investor inside the memory supply chain, and a clear lane in inference, the company has bought itself the one thing chip startups rarely get: time and a realistic path to silicon.
For continued coverage of the chips, data centers and capital shaping the AI buildout, AI Buzz Wire tracks the hardware layer of the boom.
---
Stay Ahead of AIGet the latest AI news, analysis, and breakthroughs — all in one place.
Read more AI news →
