Samsung's Unusual AI Investment... Seeking 'Beyond HBM' at Euclid

Samsung Electronics has drawn industry attention by investing in Dutch artificial intelligence (AI) semiconductor startup Euclid, signaling a strategic shift toward integrating computational and memory technologies for improved AI inference performance.

The company participated as a lead investor in Euclid's Series A funding round of over €200 million (approximately KRW 310 billion), alongside Summerhill Capital Partners, Scale Up Europe Fund, and Innovation Industries. Samsung Catalyst Fund, wholly owned by Samsung Electronics, led the investment.

Euclid, founded in 2024 in Eindhoven, Netherlands, by Bernardo Castrop and Atul Shenoy, aims to redesign AI semiconductors from fundamental principles rather than extending existing products. The company's name references the ancient Greek mathematician Euclid, symbolizing a fresh approach to semiconductor design.

Euclid's core strategy involves co-designing processors and memory systems, moving away from traditional approaches of connecting multiple GPUs for AI workloads. Its flagship product, Craftwork—a custom ASIC-based system designed specifically for AI inference—aims to enhance computational, memory, and system efficiency simultaneously by integrating these components at the hardware level.

The company claims its design achieves up to 8000 terabytes per second (TB/s) of memory bandwidth, which is critical for rapidly supplying data to processing units during AI operations. This focus on memory bandwidth addresses a key bottleneck in current AI systems where even high-performance processors struggle without fast data access.

Samsung views Euclid's approach as particularly relevant for the growing AI inference market, where performance depends not only on raw computational power but also on how efficiently data moves between memory and processing units. While NVIDIA dominates AI training through GPU performance and software ecosystems, Samsung sees opportunity in inference markets where diverse use cases require tailored hardware solutions.

Beyond Euclid, Samsung has been expanding its AI semiconductor investment portfolio, including recent investments in Oxmicro (GPU design IP), Normal Computing (AI-based chip design automation), and Mistral AI. The company's Samsung Catalyst Fund also holds stakes in other AI inference-focused companies like Grok and Ellyan.

Euclid CEO Bernardo Castrop stated that Samsung's support goes beyond financial investment, citing the company's expertise in memory engineering, supply chain management, and global network as valuable assets. Samsung Semiconductor Innovation Center Executive Vice President Dede Goldschmidt praised Euclid for offering a differentiated vision to solve data center constraints in AI infrastructure.