TrendForce News operates independently from our research team, curating key semiconductor and tech updates to support timely, informed decisions.
Recently, a joint research team from Fudan University and Shaoxin Lab, led by Zhou Peng and Bao Wenzhong, has achieved another globally leading breakthrough in 2D semiconductor integrated circuit — the development of the world’s first field-programmable gate array (FPGA) fabricated using wafer-scale two-dimensional semiconductor materials. The results were published in National Science Review.
As per Yuecheng Release, the chip leverages 2D semiconductor materials to deliver low-power operation, reconfigurability, and high reliability, offering a new hardware solution for next-generation intelligent computing and aerospace electronics.
Integrating around 4,000 transistors, the FPGA marks a historic leap for 2D semiconductors from simple logic circuits to complex reconfigurable functional systems. Unlike the team’s earlier “Infinity” chip — the world’s first 32-bit microprocessor based on 2D semiconductor materials, which used an all-digital design — the new FPGA incorporates 2T memory cells into a 2D digital process, achieving the first single-tape-out integration of different circuit types based on 2D materials.
In terms of reliability, the 2D FPGA’s core logic modules remained fully functional after exposure to a total ionizing dose of 10 Mrad, significantly reducing reliance on heavy shielding in space systems. Using industry-standard design flows, the research team has already validated the chip’s reconfigurability and practical utility on a single device.
Beyond aerospace applications, the chip is also suited for AI inference, edge computing, and IoT devices. Its reconfigurable architecture allows users to rapidly program logic functions via hardware description languages, eliminating the high cost and long development cycles associated with custom ASICs and accelerating time-to-market. In particular, for AI workloads, FPGA reconfigurability supports hardware acceleration of AI models and helps overcome the “memory wall,” improving overall energy efficiency.
(Photo credit: FREEPIK)