Eliyan has secured $60 million in funding for its chiplet interconnect technology, aimed at accelerating the processing speed for AI chips. The funding round was led by Samsung Catalyst Fund and Tiger Global Management, with a focus on addressing the challenges of developing generative AI chips. Industry forecasts project significant growth in the high-bandwidth memory (HBM) sector, with a projected 331% increase in 2024 and 124% in 2025, according to market researcher Arete Research.
Eliyan's UCIe-, BoW-, or UMI-compliant PHY (called NuLink PHY) targets memory and IO wall constraints on advanced or standard packaging material, providing a solution for connecting multiple chiplets on the same device. The company claims that its chiplet interconnect technology delivers up to four times the performance and half the power consumption compared to other solutions.
The investment includes participation from existing investors such as Intel Capital, as well as SK Hynix, Cleveland Avenue, and Mesh Ventures, among others. This round of funding follows Eliyan's $40 million Series A round in 2022. The additional investment will support Eliyan in addressing the pressing challenges in the design and manufacturing of advanced AI chips using multi-die architectures.
Eliyan aims to tackle the challenge of memory capacity and bandwidth in AI chips through its Universal Memory Interface (UMI). This bi-directional interconnect method aims to address the "memory wall" issue facing large, multi-die designs.
The company's NuLink PHY has been successfully implemented on advanced process nodes and is designed to address both Die-to-Die and Die-to-Memory interconnect with highly efficient performance metrics. The investment is a testament to the confidence in Eliyan's approach to integrating multi-chip architectures to tackle critical challenges in the semiconductor industry.
Eliyan's CEO, Ramin Farjadrad, expressed gratitude for the support of the investors and emphasized the commercial readiness of NuLink technology, optimized to deliver high bandwidth, low latency, and low power capabilities, in line with the vision of enabling the ultimate chiplet systems for the new AI era.
