Amazon EC2 P6-B300 instances now generally available

Claims to enable the scaling of AI workloads with a lower TCO

Part of the two companies' recent agreement

Nscale alone contributes $23bn

Yet another massive compute commitment made by the cloud giant

Spanning five years, Microsoft will access Iren's Nvidia GB300 GPUs

A4X Max instance contains 72 Blackwell Ultra GPUs and 36 Nvidia Grace CPUs

Would quadruple AI start-up's valuation

Aims to have more than one million Trainium2 chips by end of 2025

Will be based on the Rafay Platform

Contract worth "tens of billions of dollars"

Big Blue riding the AI "growth vector"

Firms look set to occupy entire campus

Will be publicly available in Q3 2026

Connects hundreds of thousands of Nvidia GPUs across multiple data centers

Company says its the first of many

Investors include Pembroke VCT and Praetura Ventures

That may take a while

Osaka University Compute and sTOrage Platform Urging open Science developed by NEC Corp

Through to April 13, 2032

Tech giant will make "significant investments" in its own clusters

Inference-optimized GPUs now on offer via Vultr's platform

18,000 GPUs to be leased over four years

Fluidstack said to be among those deploying the chips