Microsoft Unveils Powerful Azure ND H100 v5 AI VM and Expands Azure OpenAI Regions

Microsoft launches Azure ND H100 v5 AI virtual machine and expands Azure OpenAI service

Microsoft officially released the Azure ND H100 v5 VM, a preview‑grade virtual machine designed to accelerate generative AI tasks with up to thousands of NVIDIA H100 GPUs interconnected via Quantum‑2 InfiniBand.

The VM supports on‑demand configurations ranging from 8 to over a thousand GPUs linked by a non‑blocking fat‑tree network, each GPU equipped with 400 Gb/s NVIDIA Quantum‑2 CX7 InfiniBand and 3.2 Tb/s per VM bandwidth.

Eight NVIDIA H100 Tensor Core GPUs connected through next‑generation NVSwitch and NVLink 4.0.

NVSwitch provides 3.8 Tb/s split bandwidth between local GPUs.

Fourth‑generation Intel Xeon Scalable processors.

PCIe 5.0 host‑to‑GPU interconnect delivering 64 Gb/s per GPU.

16‑channel 4800 MHz DDR5 memory.

The new VM, currently available in Azure US East and West regions, uses NVIDIA Tensor Core H100 GPUs and Quantum‑2 InfiniBand networking.

Microsoft also announced the expansion of Azure OpenAI Service to East Canada, East US 2, East Japan, and South UK, noting that the service now serves over 11,000 customers with an average of 100 new customers per day.

This growth demonstrates the value and scalability our service provides to enterprises seeking to harness AI for their unique needs.

The most capable version of the ChatGPT platform, GPT‑4, will be available in the newly added Azure OpenAI regions.

Linux kernel disables AMD fTPM RNG across all platforms

Linus Torvalds expressed disappointment with AMD fTPM RNG issues, prompting a patch by Linux engineer Mario Limón‑Cheiro to disable the RNG on all AMD fTPM implementations.

tpm: disable RNG on all AMD fTPM

The patch, merged into the Linux 6.5 mainline, removes the TPM RNG feature even on CPUs that already support the RDRAND instruction, addressing reported system stalls. It is also marked for back‑porting to stable kernels 6.1 and newer.

This change is expected to appear in the upcoming stable kernel releases, alleviating recent AMD‑related performance complaints.

MDN launches AI Helper beta for web developers

Mozilla Developer Network (MDN) released an AI Helper beta aimed at simplifying interactions between web developers and AI platforms, providing contextual assistance and code explanations.

The tool leverages similarity‑search embeddings and generative AI: user queries are embedded, matched against MDN content, and then a generative model extracts answers from the relevant material.

MDN previously introduced AI Explain for code‑block explanations, but after community reports of inaccurate answers, the feature has been temporarily paused.

"The new AI explanation button on MDN code blocks generates human‑like text that can be occasionally correct but may also contain convincing ‘lies’. For technical references, this is a concerning decision," "I strongly believe AI assistant features could cause more harm than help. The best approach is to provide human‑written documentation reviewed by domain experts,"