What Is The Significance Of Google’s New Tau VMs?

By Patrick Moorhead - July 23, 2021

Google Cloud (GCP) recently released its Tau VM platform, based on AMD’s 3rd Generation EPYC Processor (codenamed Milan). With this release, Google is claiming best absolute performance and price-performance leadership related to general-purpose VMs in the public cloud. These are both very bold claims.  What does this mean for customers? And what does it mean for competitors such as AWS, Azure, and Oracle Cloud (OCI)? I will dig deeper into this in the following few paragraphs. 

First – a little more on Tau

If you haven’t heard of the GCP Tau VM platform, don’t feel bad. Tau is a new offering from Google, focusing on scale-out, virtualized environments. The goal of these environments – much like in an enterprise datacenter – is to deliver the best performance at the lowest cost for enterprise workloads. Media transcoding, Java-based applications, containerized workloads, and web servers are all targets for the Tau platform. 

T2D is the first instance available in Tau. As previously mentioned, this is based on AMD’s EPYC processor and can support up to 60 vCPUs per VM with up to 32GB RAM per vCPU and up to 32Gbps networking support. The complete Tau T2D lineup can be seen below. 

VM Type - Maximum Egress Bandwidth MOOR INSIGHTS & STRATEGY

With T2D, Google is claiming up to 56% better raw performance over key competitors.  The competition? AWS with Graviton2 (also known as Competitor A) and Azure with Intel “Cascade Lake” (Competitor B). 

Relative Performance MOOR INSIGHTS & STRATEGY

Google based its findings on estimates derived from SPECrate2017_int_base benchmark, which measures the integer performance of a CPU. The team ran production VMs from other cloud providers and its preproduction Tau T2D instances. The VM shape utilized 32 vCPUs and 128GB RAM.  This fact is important to note as it shows that the company measured the performance of production servers in measuring the SPECint rate. 

As it relates to price-performance, GCP found similar success. The GCP Tau T2D instance saw 42% better price-performance relative to AWS with Graviton2 and a whopping 67% advantage comparative to Azure with Cascade Lake.


When measuring price-performance, Google overlayed vendor pricing on the same SPECrate measurement to determine relative standing. Again, as you can see, the numbers are impressive. 

A caveat to bear in mind is that GCP compares instances based on AMD's latest EPYC CPU (Milan) against an Azure instance deploying an Intel CPU one generation prior. I suspect this is due to the relative newness and availability of Intel Sky Lake instances. Regardless, this price-performance advantage is quite impressive. 

How Tau translates into customer value

The value prop for the T2D instance on Tau is quite simple – best performance and best price-performance. As an ex-IT executive, these two things are what matter to me. Reliability? It’s a given with public cloud providers like GCP, AWS, Azure, and OCI.  Ditto security. These are both assumed; otherwise, the cloud wouldn’t be an option for my workloads and data. 


A 42% price-performance advantage resonates with actual TCO savings for organizations running workloads on-prem and in the cloud. The real-world number (i.e., the dollar value of that savings) will depend on how much a company chooses to deploy to GCP Tau. For x86 based workloads that can’t run in an Arm-based instance, that number is much larger, and the realized cost savings should be significant to a company of any size.

Looking at Tau in the bigger Google Cloud context, I believe T2D hits a VM instance type that customers want. In the general purpose compute space, there is a workload class that has performance requirements, yet doesn’t rise to the level of compute/memory/accelerator intense focus (or cost). And it’s this space that T2D is quite compelling, given its competitiveness on pricing.

General Purpose and Workload-Optimized GOOGLE

Is Google Cloud poised for gains?

According to Canalys, GCP currently sits at about 7% market share in the public cloud space. AWS leads the market with 32%, and Azure owns 19% market share. It has grown its share as it has deployed offerings that compete head-to-head with the likes of AWS and Azure. While I don’t have access to GCP’s customer list, I suspect its growth comes from those organizations that have used other Google services in its organization (i.e., those with enterprise agreements with Google).

Let’s not ignore the obvious

Google chose to lead its Tau offering with AMD’s 3rd Generation EPYC CPUs.  Let that sink in. AMD is the flagship CPU of this new offering that Google believes can differentiate GCP relative to the competition. AMD is well past the point of having to prove it has re-established itself in the datacenter. But just in case there were any lingering doubts or questions in the minds of enterprise IT folks – let this settle things once and for all. 

What does this mean for Arm in the cloud? 

First – let’s state the obvious. Arm has already won. The fact that it is being used as the baseline for performance means it has gained validation in the cloud world. That should be settled. 

Second, I’d be curious to see the comparison of GCP Tau v Oracle Cloud’s Ampere A1 instances (Oracle claims a significant price advantage over AWS Graviton2). To be clear, I believe GCP would still enjoy a price-performance advantage. However, that advantage would not be as significant. 

Third, Google hits on the value of Tau being an x86 architecture for those workloads requiring (x86) compatibility. Cloud vendors deploying Arm aren’t positioned for these legacy workloads. The value of Arm is in openness. As a result, I don’t believe Google needs to amplify this message. 

Arm will undoubtedly continue to gain traction in the cloud market. While price performance is one of its value props – the real value of Arm goes beyond price and price-performance.  But that’s a discussion for another article. 

Closing thoughts

Every technology trend needs that “killer app” to help drive broad adoption. It is what enables tech challengers to unseat the established leaders. In terms of the cloud, there is a similar dynamic. When discussing support for the general-purpose VM environment supporting enterprise applications, it can be hard to differentiate between cloud vendor “a” and cloud vendor “b.” Some (AWS) had first-mover status and have been able to continually innovate in terms of breadth of offerings and levels of support. Others (Azure) have been able to capitalize on the gravity established in the IT solutions space with companies of all sizes and types.  Likewise, Oracle established a foothold with its current customers before expanding services and capabilities available in OCI. 

Google has a strong portfolio of products and services. And as previously mentioned, GCP ranks with AWS and Azure in terms of availability/reliability and security. However, Tau may very well be the “killer app” due to the price-performance savings it’s passing on to its customers.  

Note: Moor Insights & Strategy writers and editors may have contributed to this article. 

+ posts
Patrick founded the firm based on his real-world world technology experiences with the understanding of what he wasn’t getting from analysts and consultants. Ten years later, Patrick is ranked #1 among technology industry analysts in terms of “power” (ARInsights)  in “press citations” (Apollo Research). Moorhead is a contributor at Forbes and frequently appears on CNBC. He is a broad-based analyst covering a wide variety of topics including the cloud, enterprise SaaS, collaboration, client computing, and semiconductors. He has 30 years of experience including 15 years of executive experience at high tech companies (NCR, AT&T, Compaq, now HP, and AMD) leading strategy, product management, product marketing, and corporate marketing, including three industry board appointments.