Socionext announced that it would develop a high-performance compute chiplet using TSMC's A14 process technology, positioning the project as a platform for next-generation custom silicon aimed at AI data center infrastructure.
The Japan-based IC design company said the A14-based device will target rising demand from hyperscale AI data centers and other compute-intensive applications that require higher performance, better power efficiency and scalable system architectures.
Socionext plans to tape out a multi-core device in September 2026. The chip will serve as a technology validation platform for CPU and xPU architectures on advanced process technology, helping customers reduce development risk before moving into production SoCs.
A14 platform targets custom AI silicon
The planned device will integrate compute chiplets designed for high performance, energy efficiency, and system-level scalability. Socionext said the development program is intended to accelerate production SoCs tailored to AI and xPU infrastructure applications.
The initiative extends Socionext's work in advanced custom SoCs for enterprise and data center applications, including multi-core CPU and xPU architectures built on leading-edge process technologies.
The company, which has traditionally supported customer designs up to 2nm-class process technologies, is moving into TSMC's A14 generation as AI infrastructure demand pushes chip developers toward more advanced nodes.
Socionext works with A14 customers
Socionext said it is already engaging with customers planning to adopt TSMC A14 technology for high-performance, power-efficient AI SoCs targeting the data center infrastructure buildout.
"Driven by increasing AI data center demands, this engagement highlights the strategic importance of advanced compute platforms and our commitment to working closely with customers and ecosystem partners," said Hisato Yoshida, president and COO of Socionext.
"By collaborating with TSMC on A14 technology, we are de-risking leading-edge product developments and accelerating time-to-market for differentiated, high-performance custom silicon solutions," Yoshida said.
Socionext said it will continue investing in advanced SoC technologies for AI data center platforms, where customers are seeking higher compute density, stronger energy efficiency, and improved system-level performance.
Article edited by Jack Wu
