TetraMem Inc., a leader in Analog In‑Memory Computing (A‑IMC), and SK hynix Inc., a global innovator in AI memory and semiconductor technologies, announced the successful completion of their joint technology collaboration—culminating in the publication of their research paper, “A Memristor-based In-Memory Computing SoC with Efficient Depthwise Convolution,” in Advanced Intelligent Systems. The work was selected as the cover feature, underscoring its significance for the future of AI computing.
Advancing AI Beyond the Memory Wall
The partnership combines SK hynix’s deep expertise in advanced memory technologies with TetraMem’s pioneering A‑IMC platform to tackle one of AI’s most urgent challenges: the energy and thermal limits created by massive data movement.
As foundation models scale from billions to trillions of parameters, moving data between processors and memory has become a dominant source of power consumption, latency, and heat. A‑IMC offers a fundamentally different approach byperforming matrix operations directly where model weights reside, dramatically reducing data movement and improving system‑level efficiency.
A Practical Demonstration of Memory‑Centric AI Computing
The published research showcases a memristor‑based AI System‑on‑Chip (SoC) capable of efficient depthwise convolution, a key operation in modern AI inference workloads. More importantly, the project demonstrates the successful integration of:
- emerging memory devices
- circuit and SoC design
- AI architecture
- software and system optimization
This achievement highlights the strength of the engineering collaboration between the SK hynix RTC and TetraMem teams, who combined complementary expertise to push the boundaries of memory‑centric AI computing.
Leadership Perspectives
“This milestone shows what’s possible when companies collaborate across the semiconductor ecosystem,” said Glenn Ge, CEO and Co‑Founder of TetraMem. “As AI evolves, innovation must happen not only in compute, but also in memory and system architecture. We believe memory‑centric computing and Analog In‑Memory Computing will be essential for addressing future AI energy and thermal challenges.”
“We are pleased with the successful outcome of this collaboration and the recognition from Advanced Intelligent Systems,” said Soo Gil Kim, Vice President at SK hynix. “This work demonstrates the value of exploring new memory technologies and computing architectures for future AI systems. We look forward to continued technical exchanges with TetraMem.”
Looking Ahead
The selection of this work as the journal’s cover feature underscores the growing importance of memory‑centric architectures in the AI industry. Both companies recognize that future AI infrastructure will require continued advances in memory technology, compute architecture, and system integration to meet rising demands for performance, efficiency, and sustainability.
Building on the success of this collaboration, TetraMem and SK hynix plan to explore additional opportunities to advance next‑generation AI computing technologies.
The paper, “A Memristor-based In-Memory Computing SoC with Efficient Depthwise Convolution,” is now available online through Advanced Intelligent Systems. https://advanced.onlinelibrary.wiley.com/doi/10.1002/aisy.202501225