Sep 4, 2026

Semiconductor and Quantum Computing Advances: September 2026

Researchers at Tsinghua University have built a photonic computing chip that merges sensing, processing, and memory functions entirely on-chip, featuring 7,378 neurons across multiple cores. Demonstrated in autonomous racing-drone navigation, the design seeks to minimize data transfers between compute units and external storage.

At EPFL, a GaN-on-Si transistor employing a polarization superjunction achieved a breakdown voltage above 3.4 kV, while companion diodes surpassed 3.9 kV. A team led by Seoul National University devised an AI approach capable of isolating overlapping semiconductor defect patterns at the pixel level, relying solely on image-level defect annotations.

Gartner reports that AI-driven vulnerability discovery and exploitation ranked as the foremost emerging risk in Q2 2026, based on input from 316 executives and risk managers. AMD researchers demonstrated an optical side-channel attack method targeting AMD 7-Series FPGAs. SEALSQ intends to launch its post-quantum TPM 2.0, incorporating wolfSSL's wolfTPM software to enhance adoption and scalability.

With AI-based cryptanalysis, post-quantum transitions, chiplets, and custom silicon altering the threat environment, the semiconductor sector is grappling with how to establish, validate, and certify security across increasingly intricate chips and systems. Safeguarding the full data center supply chain remains a formidable challenge, especially as the worth of data in hyperscale facilities climbs continuously. Additionally, 800V DC architectures introduce heightened safety and reliability concerns.