Samsung Is Betting On An Alternative To Nvidia. Dutch AI Chip Startup Euclyd Just Raised $231 Million.

Samsung has backed Dutch semiconductor startup Euclyd in a funding round of more than €200 million, giving the two-year-old company fresh capital to develop AI chips and computing systems designed as an alternative to the graphics processors that currently dominate artificial intelligence infrastructure.

The Series A round was co-led by Samsung, Somerset Capital Partners, the EQT-managed Scaleup Europe Fund and Innovation Industries, Euclyd said. Denmark's Export and Investment Fund, imec.xpand, Brabant Development Agency and Quadri also participated. CNBC valued the round at about $231 million.

Euclyd, founded in 2024 and based in Eindhoven, Netherlands, is developing infrastructure for running foundation AI models, including custom compute silicon, memory architecture and data-center systems. Its technology is focused on inference, the process through which a trained AI model responds to requests and generates outputs.

The startup says its architecture can reduce the power, memory and infrastructure needed to run increasingly large AI models. Those claims have yet to be demonstrated at the scale of major commercial deployments.

CEO Bernardo Kastrup told the media outlet that Euclyd expects to begin rolling out physical chip systems in 2028 and aims to serve thousands of enterprise customers by 2030. The company plans to make money both by selling hardware and rack systems to customers running AI models on their own infrastructure and by licensing its intellectual property to companies developing their own chips.

Samsung's involvement gives Euclyd a strategic investor with extensive semiconductor manufacturing and memory expertise. Kastrup added that Samsung's value to the startup extends beyond its financial investment because of the company's engineering expertise, supply-chain network and position as one of the world's largest memory manufacturers.

Euclyd said the new funding will be used to expand its engineering workforce, speed development of its silicon and systems and build partnerships as it targets enterprise, government and hyperscale AI customers.

The company is developing a programmable AI processor called Craftwerk alongside the Craftwerk Station CWS, a larger computing system designed around its processor and memory architecture. Euclyd says the combination is intended to address power consumption and memory bottlenecks associated with running large AI models.

Former ASML President and CEO Peter Wennink is also joining Euclyd as non-executive chairman, giving the young company an executive with decades of experience in Europe's semiconductor industry. Wennink led ASML from 2013 until his retirement in 2024.

Euclyd is entering a market still heavily dependent on Nvidia GPUs but increasingly crowded with companies developing processors specifically for AI workloads. The shift has accelerated as the cost and electricity requirements of building and operating AI infrastructure have increased.

Some of the world's biggest technology companies are already designing their own chips rather than relying exclusively on general-purpose AI accelerators.

The growing number of custom processors does not mean those companies have stopped using Nvidia hardware. Google, for example, has said Nvidia GPUs remain a core part of its AI accelerator portfolio alongside its own TPUs. Instead, companies are building a broader mix of hardware designed for different AI workloads.

Euclyd is taking a similar specialized approach but must still prove its architecture outside development and testing environments. Kastrup acknowledged that the company's systems have yet to be demonstrated at scale in commercial deployments.

The startup's focus on inference also puts it in a part of the AI chip market drawing increasing attention as companies move beyond training increasingly capable models to deploying them for millions of users. Inference requires models to repeatedly process requests, making the cost, speed and energy efficiency of the underlying hardware important to companies operating AI services at scale.