Infrastructure & Hardware

Enphase and Infineon say they are deploying the world's first 900V GaN bidirectional switches in Enphase's solid-state transformer modules, claiming lower switching losses and fewer parts. The devices passed testing and are being integrated into IQ SST builds.

NVIDIA released Nemotron 3.5 Transcribe, an open-weight, promptable model combining transcription, diarization, translation and audio Q&A. It covers 40 languages and up to an hour of audio with eight speakers, plus Granary v2 data.

The FT reports TPG is in talks to buy Ark Data Centres from Elliott at a valuation of about $5 billion. No deal is assured, and the companies declined to comment. Ark has 27 data centers across nine UK and Belgian sites, totaling over 560MW.

Trillium Asset Management is backing a proposal asking Microsoft to explain how it will meet its 2030 emissions goals, noting emissions are 62% above the 2020 baseline. The vote is expected December 8, and Microsoft recommends shareholders vote against it.

Gingerah Energy has filed plans near Broome for a 6.7GW renewable hub powering a 1GW data center campus, with a first stage of up to 240MW. Construction could start in 2029 and full build-out is targeted for 2038.

ITmedia reports that Oracle Cloud will adopt Google's TPUs and begin offering them in 2027. It signals cloud providers widening their AI accelerator sources. Service details and pricing are not confirmed yet.

NVIDIA and AMD each announced expanded strategic partnerships with Meta in AI, cooperating on development and operation of the Muse AI agent and on next-generation infrastructure. They also pledged investments in Meta ($3 billion from NVIDIA, $1 billion from AMD), as Meta builds out infrastructure with both in-house chips and external supply.

AI data center construction is accelerating in Japan, with sites ranging from former factories and closed schools to offshore and even space. The article examines the aims behind these unconventional location strategies.

NVIDIA describes memory-enforced GPU sharing in Kubernetes using CUDA 13.4. Many AI workloads need only a fraction of a GPU, and giving each one a whole device leaves capacity idle. Sharing GPUs lets platform teams serve more workloads.

Oracle will be the first cloud outside Google to offer Google's TPUs, available on OCI from 2027. It also announced Nvidia Vera Rubin NVL72 availability, new OpenAI-based products, and a Snowflake partnership.

Oracle will offer Google's TPU AI chips through Oracle Cloud Infrastructure. It would be the first hyperscaler to provide Google's TPU outside of Google itself.

Amazon has settled its legal battle against AWS managers accused of kickbacks in data center deals. The long-running case will continue with Northstar Commercial Partners as the defendant.

NVIDIA presents RTX Spark, a new Windows PC platform for AI development, creation and gaming in slim laptops and compact desktops. Built on NVIDIA Blackwell RTX, it powers neural rendering and AI characters in games.

A hearing on Google's data center in Hermantown, Minnesota, was shut down due to protests. It was pushed back a day, but the project faces lawsuits and other challenges.

Oracle announced it will bring Google's AI-specific TPU chips to Oracle Cloud Infrastructure, with Google TPU services becoming available from 2027. The announcement was made at Oracle AI World 2026.