AI COMPUTE•NVIDIA•GPU CLOUD•DATA CENTERS•SEMICONDUCTORS•ENERGY•FUNDING•M&A•AI COMPUTE•NVIDIA•GPU CLOUD•DATA CENTERS•SEMICONDUCTORS•ENERGY•FUNDING•M&A•
GPU·DATA HUB
Newsletter
Semiconductors

Microsoft Boosts HBM Bandwidth by 2.5x with Maia 300 Accelerator

Microsoft increases memory throughput to 18 TBps on custom Maia 300 inference accelerator

By GPU Data Hub Desk · GPU Data Hub·Editorial standards··United States
Microsoft Boosts HBM Bandwidth by 2.5x with Maia 300 Accelerator

Microsoft has revealed its Maia 300 AI accelerator, increasing high-bandwidth memory throughput to 18 terabytes per second from 7.1 TBps on the earlier Maia 200. The 2.5-fold increase in memory bandwidth is engineered to process AI inference workloads more efficiently whilst decreasing energy expenditure on a per-token basis. Cloud and AI executive vice president Scott Guthrie presented the silicon update as part of the group's ongoing custom infrastructure roadmap.

Original source

This summary was written by the GPU Data Hub desk from reporting published by HPCwire on 12 Oct 2026, 22:46. Read the full article at the original publisher.

Read Original Source →

By subscribing you consent to receive the daily briefing. Unsubscribe at any time. See our privacy policy.

Companies

Microsoft
LinkedInX

More from GPU Data Hub

Latest →

Chip Industry Technical Paper Roundup: Oct. 13

A recent collection of semiconductor research papers highlights advances across lithography, materials science and specialised computing hardware. Key subjects addressed include hyper-NA EUV lithography, neuromorphic computing architectures, and methods to detect bit-flip attacks in AI acceleration hardware.

Semiconductors··Semiconductor Engineering

Carbon-Dot Interface Enables Ultrathin High-κ Dielectrics on 2D Semiconductors (UIUC, NETL, Argonne, Monash)

Researchers from several American and Australian institutions have demonstrated a method to integrate high-κ dielectrics onto two-dimensional semiconductors using a carbon-dot monolayer assembly. This atomically thin structure acts as a van der Waals interfacial layer that facilitates oxide deposition without damaging the underlying material. The technique targets longstanding material integration hurdles in the manufacture of next-generation 2D semiconductor devices.

Semiconductors··Semiconductor Engineering

On-Chip Monitors Detect Bit-Flip Attacks in AI Accelerators (UF, UT Dallas, RPI)

Academic researchers have devised a hardware defence architecture titled BARE-AI to protect artificial intelligence processors from bit-flip attacks. The framework deploys built-in performance monitors directly on the silicon to identify malicious tampering aimed at altering deep neural network parameters. The system addresses critical security vulnerabilities in deep learning accelerators deployed across safety-critical operational environments.

Semiconductors··Semiconductor Engineering

Deterministic Nucleation Enables Controlled Growth of 2D Semiconductors (KAIST, TDS, Hanyang et al.)

Researchers at several South Korean academic institutions and TDS Innovation have introduced a method to direct the nucleation of two-dimensional semiconductors using etching flux. The approach establishes spatial control over where and when crystalline structures develop during the semiconductor deposition phase. This process provides an alternative route to conventional area-selective deposition methods for synthesising advanced 2D electronic materials.

Semiconductors··Semiconductor Engineering