AWS vs Azure vs GCP Cloud Services Comparison

This cloud services comparison has been compiled in order to illustrate the differences and similarities between three of the biggest cloud services providers as of April 2018. The cloud services comparison has been compiled without bias towards one provider over another, but readers are advised to check the services being offered by each – and their respective prices – before making a purchasing decision.

Why an AWS vs Azure vs GCP cloud services comparison? There are dozens of cloud service providers offering similar services, but outside of Asia the market is dominated by Amazon Web Services (AWS), Microsoft Azure (Azure) and Google Cloud Platform (GCP). Despite the rapidly changing landscape of the cloud services market, we do not expect the “Big Three” to lose their positions in the near future.

What is Covered in the Cloud Services Comparison?

In order to best illustrate the differences and similarities between AWS, Azure and GCP, our cloud services comparison starts by looking at the three providers´ timelines – as this provides an insight into the direction they are likely to pursue in the future. The comparison then examines the services currently being offered, their pricing structures, and the discounts available to users.

We conclude our AWS vs Azure vs GCP cloud services comparison by discussing multi-cloud management tools. This section may help users who are considering a multi-cloud or hybrid strategy. If you have any comments about our comparison – or have any questions about multi-cloud management tools – you are invited to share them with us[KS1]

The Timelines of AWS vs Azure vs GCP

Although the concept of the cloud dates back to the 1960s, commercial cloud service providers did not emerge until the dot-com boom of the 1990s. Exodus Communications was the first company to offer colocation services and enterprise-level web hosting via “Internet Data Centers”, but was also one of the first companies to go bust when the dot-com bubble burst.

At the time of the dot-com crash, Amazon was revolutionizing e-commerce and in the process of becoming the world´s largest online marketplace. While working on an e-commerce platform for Target, the company identified an opportunity for “selling access to virtual servers as a service” and the first AWS platform was launched in July 2002 consisting of a handful of developer tools and services.

By comparison, Microsoft was much later to market with its Windows Azure Platform. The platform was first announced in 2008, but not made commercially available until 2010 – when it was unkindly described by some in the industry as a “cloud layer for Windows Server systems” as it failed to support Linux-based applications (support for Linux-based applications has since been added in stages).

The release of Google´s Cloud Platform was done in stages similar to AWS. First an app engine was released in April 2008, followed by Google Cloud Storage in 2010, and it was only in December 2013 the Google Compute Engine was made generally available. Despite its late start, Google is catching up with the product range of its two biggest market rivals.

Significant dates in the timelines of AWS vs Azure vs GCP include:

  • In November 2004, AWS releases its first publicly-accessible service – Simple Queue Service (SQS).
  • In March 2006, EC2 compute instances and S3 cloud storage services are added and the AWS platform relaunched as “an integrated suite of core online services”.
  • In April 2008, Google’s App Engine was released as a Platform-as-a-Service for preview only. It came out of preview in 2011.
  • In May 2009, AWS´ compute services were enhanced with the launch of Elastic Load Balancing, Auto Scaling and Amazon CloudWatch.
  • In March 2009, Microsoft announces its Windows Azure Platform will include an SQL Relational Database Service.
  • In October 2009, AWS adds a Relational Database Service (RDS) to its portfolio of publicly-accessible services.
  • In February 2010, the Windows Azure Platform becomes commercially available. Multiple functions are added throughout the rest of the year.
  • In June 2012, Linux VMs are first made available on the Azure platform. Microsoft claims to gain 7,000 new customers a week over the next twelve months.
  • In November 2013, AWS announces G2 instances – a new EC2 instance type designed for apps that require 3D graphics capabilities.
  • In December 2013, Google Compute Engine becomes generally available, shortly followed by Google Cloud SQL.
  • In March 2014, Google announces price reductions of between 30% and 85% across its entire range of cloud services, sparking tit-for-tat price cuts by other providers.
  • In November 2014, both the AWS Lambda and the EC2 Container Services (ECS) are released – ECS integration with Docker is made available at the time of release.
  • In September 2015, Azure Cloud Switch introduced as a cross-platform Linux distribution service to promote open networking across multiple cloud service providers.
  • In February 2016, Google announces its Cloud Functions service is available in Alpha mode. The Beta release occurred in March 2017.
  • In April 2016, The Azure Functions and Azure Containers Services are made generally available across all regions and availability zones.
  • In May 2017, Dell EMC launches a hybrid cloud platform for Microsoft Azure Stack to bring the cloud model to on-premise IT infrastructures.

What the above timeline tells us is that not only was AWS first-to-market, the company is also first-to-innovate. Microsoft Azure realizes that it cannot compete by focusing solely on Microsoft-compatible products and is leading the charge towards multi-cloud and hybrid environments. Google is somewhere in the middle – always seemingly trying to catch up with AWS and Azure, but releasing more versatile products and services.

Cloud Services Comparison of Terminology

One of the biggest obstacles to overcome when compiling a cloud services comparison is the different terminologies in use – especially when some terminologies have been accepted to mean the same thing, whereas they are not. Take “Instances” and “Virtual Machines” for instance. Most people believe AWS Instances and Azure Virtual Machines are the same. To confuse the issue further, Google uses both terms interchangeably and the term “VM instances” as well.

In a Private Cloud environment, Virtual Machines operate on a server, and any server resources not used by Virtual Machine A are used by Virtual Machine B, Virtual Machine C, and so on. In the Public Cloud, resources are provisioned (reserved) for each Instance and any unused server resources reserved for Instance A cannot be used by Instance B. Effectively, a Public Cloud instance is more like a virtualized server than a Virtual Machine.

And while on the subject of Private Clouds, the definition of a Private Cloud also causes confusion. Whereas some in the industry define a Private Cloud as an on-premise IT infrastructure, a Private Cloud can also be an off-premise “Internet Data Center”. Our preferred definition is supplied by Gartner analyst Tom Bittman, who describes Private Cloud computing as being “defined by privacy, not location, ownership or management responsibility”.

Due to so many different terms relating to the same element of cloud services – or the same terms relating to different elements of cloud services – we have compiled a cloud services comparison of terminology to assist readers unfamiliar with what names different cloud service providers assign their services. Please note that, as cloud services evolve, so do the names assigned to them. We anticipate, at some time in the future as the cloud providers continue to innovate, this cloud services comparison of terminology may be out of date.

Service AWS Azure GCP
Compute Elastic Cloud Compute Virtual Machines Compute Engine
App Hosting Elastic Beanstalk Cloud Services App Engine
Serverless Computing AWS Lambda Azure Functions Cloud Functions
Container Support ECS/EKS Containers AKS Container Service Kubernetes Engine
File Storage S3 Storage Service Azure Storage Cloud Storage
Block Storage Elastic Block Storage Azure Blob Persistent Disc
Backup Options AWS Glacier Azure Backup Cloud Storage
Data Orchestration Data Pipeline Data Factory Cloud DataFlow
Data Management AWS Redshift SQL Data Warehouse Google BigQuery
NoSQL Database DynamoDB Cosmos DB Cloud DataStore

Naturally this is not a full list of services offered by the three cloud service providers. AWS offers ninety different services, which is around the same number as GCP (although only 46 services are covered by the GCP Terms and Conditions). Microsoft lists an incredible 166 services on its Azure product page; however a number of these are integrations with other Microsoft products.

The bottom line is that most of the products and services offered by AWS, Azure and GCP perform pretty much the same functions. One of the service providers (usually AWS) might launch a unique product or service from time to time; but you can be confident a comparable product or service will be released by both its competitions soon after. In that respect, our cloud services comparison shows little difference between AWS, Azure and GCP. So what about price?

Cloud Services Pricing Comparison

A cloud services pricing comparison is extremely difficult to compile due to the three cloud service providers not supplying identical services. GCP also gives users the option to customize VM instances – making any AWS vs Azure vs GCP cloud services price comparison for compute services nearly impossible. One cloud services pricing comparison website noted that in the six months between their comparisons:

  • The prices of instances/VMs had fallen in 70% of the examples the website used to compile its cloud services pricing comparison.
  • Both AWS and Azure had released new families or updated existing families in order to offer better performance for less cost.
  • AWS had moved from per-hour billing to per-second billing, while Google had reduced its minimum charge from ten minutes to one minute.
  • Azure had introduced per-second billing for containers (although like-for-like, Azure is still more expensive for container deployment than AWS or GCP).
  • There had been significant changes made by all three cloud service providers with regard to the prices and capacities of attached storage.

Their conclusion was that, although cloud services pricing tends to be extremely fluid, there is little to choose between the three cloud service providers. This matches a similar conclusion we arrived at in an “AWS vs Google Cloud Pricing” whitepaper we produced last year (you can download the whitepaper from this blog post). At the time, we found that – in general – AWS is less expensive on a CPU per hour basis for general workloads, while GCP was less expensive for compute intensive workloads.

Cloud Service Discounts (and why they are not as good as they appear to be)

In addition to the obstacles hindering the compilation of a meaningful cloud services pricing comparison, all three cloud service providers offer different price points for each service depending on the location of the server, prepayment, committed use, sustained use or off-peak use. The table below provides a general guide to the savings that can be made by taking advantage of discounted price points, but users are again advised to check the discounts are still available before making a purchasing decision.

Provider AWS Azure GCP
Type of Discount Reserved Instances Reserved Instances Sustained Use (SU)
Committed Use (CU)
Length of Commitment 1 or 3 Years 1 or 3 Years SU – No Commitment
CU – 1 or 3 Years
Discount Available Up to 75% Up to 72% SU – Up to 30%
CU – Up to 55%
Other Discounts Volume RI Discounts
Spot Instances
Hybrid Use Discounts
Enterprise Agreements
Preemptible VMs

However, an issue with taking advantage of the maximum discounts available in return for a three- year prepayment is that, if prices come down during the contracted period, users do not benefit from the price reductions and carry on paying the prices they committed to. AWS has addressed this issue to a degree by introducing Convertible Reserved Instances, but these do not attract the same level of discount and – in order to take advantage of the flexible option – users have to exchange their instances even if they do not want to.

While on the subject of AWS and discounts for Reserved Instances, the following tables show the perceived and actual discounts a user would receive compared to On Demand pricing if Moore´s Law and historical price reduction evidence were applied to an m4.16xlarge instance. The calculations are based on cumulative 25% price reductions over a Standard RI one year term and a Standard RI three year term, and the calculations assume the instance is fully utilized for the duration of each term.

Perceived Discount

Standard One Year Term

Payment Option





Effective Hourly Perceived


No Upfront $0 $1,447.15 $1.982 38%
Partial Upfront $8,269 $689.12 $1.888 41%
All Upfront $16,208 $0 $1.850 42%
Standard Three Year Term

Payment Option





Effective Hourly Perceived Discount
No Upfront $0 $1,009.15 $1.382 57%
Partial Upfront $16,819 $467.20 $1.280 60%
All Upfront $31,620 $0 $1.203 62%

Calculated “Actual” Discount

Standard One Year Term
Payment Option Amount Paid On Demand Cost Saving Actual Discount
No Upfront $17,366 $21,024 $3,658 18%
Partial Upfront $16,538 $21,024 $4,486 21%
All Upfront $16,208 $21,024 $4,816 23%
Standard Three Year Term
Payment Option Amount Paid On Demand Cost Saving Actual Discount
No Upfront $36,329 $48,618 $12,289 25%
Partial Upfront $33,638 $48,618 $14,980 31%
All Upfront $31,620 $48,618 $16,998 35%

Because of the long-term nature of prepayment discounts, some users are locked into commitments that prevent them from adopting a multi-cloud or hybrid cloud strategy. As proven above, it is also the case that actual discounts are not as good as they appear to be. This should be taken into account when reviewing any benefits uncovered in our cloud services comparison, as it may be a wiser option to forgo discounts and implement multi-cloud management tools to achieve the same (or better) cost savings.

Multi-Cloud Management Tools for Cost Savings

Users that spend hours studying a cloud services pricing comparison to try to reduce costs are more than likely looking in the wrong place. Research has shown that most users pay too much for their cloud services due to over-provisioning (or failing to re-provision once demand has fallen), failing to terminate zombie assets, or failing to switch off non-production instances/VMs when they are not required. We estimated in this blog postwasted cloud spend in 2018 will amount to $12.9 billion, solely from resources left running when not being used.

There are tools available to identify over-provisioned resources and unused assets. Unfortunately the ones provided by cloud service providers have been designed exclusively for their own products and services, and they tend to be host-centric rather than service- or role-centric – creating potential cost and oversight issues due to the trend towards containerization and serverless computing. Indeed, our estimation of cloud waste in 2018 might prove to be too conservative.

By comparison, ParkMyCloud not only identifies over-provisioned resources and unused assets by role across AWS, Azure and GCP, it automates the scheduling process for non-production instances and VMs to save users up to 65% of their cloud costs. Once costs have been optimized, our multi-cloud management tools for cost savings use policy-driven automation to maintain optimization, enhance accountability and provide complete governance over users´ cloud accounts.

To find out more about ParkMyCloud – or to request a free trial – do not hesitate to get in touch. We will be happy to answer any questions you have and discuss your current cloud strategies in order to determine how best ParkMyCloud can help. If you feel it would be beneficial to you, we can also organize a demonstration of ParkMyCloud in action so you can see how easy it is to optimize and manage your assets deployed in the cloud. To find out more contact us today.