← All Briefings
Briefing · Sunday, February 22, 2026

Samsung HBM4 Debuts, Taalas Transistor‑Level AI, Siemens Wins Chiplet Packaging Award

Share

Today’s semiconductor landscape was highlighted by Samsung’s launch of HBM4 DRAM, a memory leap that delivers unprecedented bandwidth for AI and HPC workloads. Meanwhile, AI pioneer Taalas showcased a wafer‑scale technique that embeds large SRAM blocks directly into transistor‑level AI engines, promising massive inference speedups. In the packaging arena, Siemens’ Innovator3D IC software captured the Best of Show award at the Chiplet Summit, underscoring the rapid adoption of 2.5D/3D chiplet integration. Power electronics also advanced with Navitas’ refined GeneSiC platform, cutting losses in SiC devices for EV and renewable energy applications.

Key Stories

Samsung leads with HBM4 DRAM performance

Samsung has begun mass production and commercial shipments of its HBM4 DRAM, offering up to 1.5 Tbps bandwidth per stack and 1.5× higher density than HBM3, enabling faster AI training, HPC, and next‑gen graphics workloads.

EDN

Taalas Etches AI Models Onto Transistors To Rocket Boost Inference

Taalas demonstrates a wafer‑scale approach that embeds large SRAM blocks directly into AI tensor engines, dramatically increasing inference throughput and delivering orders‑of‑magnitude performance gains for edge and data‑center AI workloads.

The Next Platform

Navitas tightens SiC losses with refined TAP

Navitas Semiconductor announced its 5th‑generation GeneSiC platform featuring a refined Thermal‑Aided Passivation (TAP) that reduces conduction losses in silicon carbide devices, boosting efficiency for EV power electronics and renewable energy inverters.

EDN

Editor's Picks

Nvidia: Star Attraction at CES 2026

Nvidia's Vera Rubin platform showcased at CES signals a new era of AI‑driven autonomous systems, hinting at powerful on‑chip solutions for robotics and automotive.

EE Times