Google Cloud announces a preview of Tau T2A, its first ARM-based virtual machines powered by Ampere Altra, with up to 48 vCPUs, each with up to 4GB of memory
It's been a long time coming, but Google Cloud today announced its first Arm-based VMs, following AWS, with its Graviton instances … Source: Google Cloud Blog and Ampere Computing .
Context & Ripple Effects
Google Cloud had already diversified its infrastructure with second-generation TPU chips for machine learning, while Microsoft had recently previewed Azure VMs using the same Ampere Altra processor family. Tau T2A extends that multi-architecture approach to general-purpose virtual machines and puts Google Cloud into the Arm VM race alongside AWS’s Graviton offering.
The later Axion chip rollout shows why this early reliance on Ampere matters: it marks a stepping stone from offering an external Arm CPU to differentiating Google Cloud’s Arm infrastructure more directly.
First-order effects
- Google Cloud customers gain preview access to Arm-based Tau T2A configurations with up to 48 vCPUs and up to 4GB of memory per vCPU.
- Ampere gains a second major cloud deployment for Altra after Azure’s Ampere Altra VM preview, widening the processor’s route to enterprise workloads.
Second-order effects
- Google Cloud and Microsoft must differentiate Arm VM services through their platform integration and workload support because both begin with Ampere Altra hardware.
- AWS’s Graviton becomes a more immediate competitive reference point as Google Cloud adds an Arm alternative rather than leaving AWS as the only named Arm cloud option in the coverage.
Third-order effects
- The sequence from Ampere-powered VMs to Google’s later Axion chips points to cloud providers treating CPU architecture as a controllable infrastructure layer, not merely a commodity server choice.
- If providers continue pairing Arm CPUs with TPUs and GPUs, enterprise cloud procurement will increasingly involve choosing among specialized compute architectures within a single platform.
The trend: Hyperscalers are broadening cloud infrastructure from single-architecture VM fleets into differentiated portfolios of Arm CPUs, TPUs, and GPUs.