Cloud Pricing Comparison in 2020

Compiling a cloud pricing comparison in 2020 is not a simple task. The three major cloud providers (Amazon AWS, Microsoft Azure, Google GCP) reduce their prices on a regular basis – not to mention they also each offer unique pricing models and differing discount options, making it hard to find an “apples-for-apples” cloud pricing comparison. 

In addition to fluid prices, pricing models and discount options, the three major cloud providers are constantly releasing new availability zones (each with their own price ranges), new products and “New Generation” upgrades to existing products that offer improved performance and their own prices. Add the growth of containerization, adoption of multi-cloud and hybrid cloud, and FaaS services, and it all gets very complicated.

Consequently, for the purposes of this cloud pricing comparison, we are going to focus on just the comparative costs of deploying virtual machines (or “instances”) on AWS, Azure and Google. Instance deployment on IaaS services is still the most popular form of cloud computing, so we believe this is the area of a cloud pricing comparison that will benefit the most readers.

Is AWS the Cheapest Option? 

AWS is often perceived to be the cheapest option, but that is not always the case. Although AWS remains at the head of the cloud computing market in terms of dollar turnover and market share, Microsoft Azure and Google Cloud Platform (GCP) have stayed in the competition by matching the market leader’s price reductions like-for-like in recent years. In fact, Google has initiated many rounds of price reductions by being the first to cut its prices.

The reason why AWS is seen as the cheapest option is because of its alleged “virtuous cycle” – a concept in which the more business a cloud provider attracts, the more servers it can afford to add. The more servers the provider has, the better it can take advantage of economies of scale and offer lower prices. The lower the prices, the more business the provider attracts – and the cycle goes on.

Because AWS was first-to-market, many believe its “virtuous cycle” gives it a one up on the other cloud providers and is seen as the top choice in any cloud pricing comparison. However, as the table below demonstrates, there are remarkably few occasions when AWS can genuinely claim to be the market leader in price (the prices quoted in the table are per instance per hour, based on pay-as-you-go “On Demand” pricing).

AWS vs Azure vs GCP On Demand Cloud Pricing Comparison
Type vCPU Mem. AWS Azure GCP
General Purpose 2 8GB $0.0928 $0.0850 $0.1070
4 16GB $0.1856 $0.1700 $0.2140
8 32GB $0.3712 $0.3390 $0.4280
Compute Optimized 2 4GB $0.0850 $0.0850 $0.0813
4 8GB $0.1700 $0.1690 $0.1626
8 16GB $0.3400 $0.3380 $0.3253
Memory Optimized 2 16GB $0.1330 $0.1330 $0.1348
4 32GB $0.2660 $0.2660 $0.2696
8 64GB $0.5320 $0.5320 $0.5393


Notes: Prices are for Linux VMs/Instances located in East Virginia (US East) and are correct as of January 9, 2020. Memory sizes may vary. Google´s VMs generally have less memory than AWS or Azure VMs.

What´s Wrong with this Cloud Pricing Comparison?

There would be nothing wrong with this cloud pricing comparison if all cloud service providers charged all customers by the hour for like-for-like services without applying discounts, but unfortunately they don´t. AWS and GCP and now Azure, bill by the second, with AWS applying a 60-second minimum charge, GCP a 1-minute minimum charge, Azure VMs are rounded down to the last minute. In order to keep up with the competition, Azure has begun matching AWS pricing for comparable services. Although these charging structures do not make much of a difference for a single VM instance, they do when you have thousands deployed.

GCP also throws a spanner into the calculations by allowing customers to create customized VM instances. These custom VM instances are charged based on the number of memory hours and vCPUs used by the instance. Custom GCP instances are subject to the same 1-minute minimum charge as any other instance – however, sustained use discounts are automatically applied (instances mush be used a minimum of 25% of a month). This is an automatic, built-in discount for compute capacity, giving you a larger percentage off the more you run the instance. With custom machine types, you can define exactly how many vCPUs you need and what amount of system memory for the instance. They’re a great fit if your workloads don’t quite match up with any of the available predefined types, or if you need more compute power or more memory, 

Customers of AWS, GCP and Azure can save money by running their VM instances during periods of low demand and taking advantage of Spot Instances (AWS), Preemptible VMs (GCP) and low-priority VMs (Azure); customers of Microsoft may be eligible for a general discount if they have an existing Enterprise Agreement in place. These options allow users to purchase unused capacity for a discount. 

For customers prepared to commit to a predetermined level of use, all three cloud service providers offer a way for customers to purchase compute capacity in advance to further discounts. AWS and Azure offer “Reserved Instances” programs – saving customers up to 72% and 75% respectively – while GCP customers can save up to 55% via a “Committed Use Discount” program. In each case, the value of the discount depends on how long the customer is prepared to commit to a level of use, and how much of that commitment they are prepared to pay in advance.

Why Prepayment Discounts May Not be so Beneficial

Back in 1965, Intel´s co-founder Gordon Moore identified the number of transistors per square inch on integrated circuits had doubled every year since their invention. He predicted the trend would continue into the future; and, although the rate of multiplication has decreased to around 1.5 each year, his prediction – which came to be known as Moore´s Law – has a relevance in our cloud pricing comparison concerning locking yourself into a committed use program in order to achieve the maximum discounts.

Cloud service providers frequently update their services and call them “Next Generation” VM instances. These have a better performance than their comparative predecessors and are usually available at a lower cost. Cloud service providers also frequently reduce their costs for VM instances and other services as the benefits of the “virtuous circle” are realized. The consequences for customers locked into committed use programs is that they are still paying for the old level of performance at the old price.

Historically, cloud computing prices have come down by around 25% each year. Therefore, if you were to lock yourself into a committed use program for three years, the actual savings you would achieve could be closer to half of what you expect due to compound price reductions. The following table illustrates this using a four-core AWS instance (t2.xlarge) over the course of a three year standard Reserved Instances contract.

AWS t2.xlarge Standard Three Year Reserved Instances Contract
Payment Option Upfront Monthly Effective Hourly Advertised Discount
No Upfront $0 $58.55 $0.080 57%
Partial Upfront $976 $27.08 $0.074 60%
All Upfront $1,834 $0 $0.070 62%
  On Demand Cost (2018) Adjusted Cost (3x 25% cuts) Amount Paid Actual Discount
No Upfront $4,810.75 $2,781.17 $2,107.80 24%
Partial Upfront $4,810.75 $2,781.17 $1,950.88 30%
All Upfront $4,810.75 $2,781.17 $1,834.00 34%

To address this issue, AWS offers “Convertible” Reserved Instances for 1-year and 3-year contracts at a lower discount rate. Convertible Reserved Instances are attractive to customers because they get additional flexibility like the ability to use different instance families, operating systems, or tenancies over the term. They also offer Scheduled Reserved Instances, which allow you to buy an RI that is only used at certain times each day in a recurring schedule.

Azure allows you to cancel prepaid Reserved Instance contract for an adjusted refund or exchange your unused credits for another service. When you purchase a committed use contract through GCP, you are committing to pay a discounted price for vCPUs, memory, GPUs and local SSDs for either 1 or 3 years. GCP does not have an upfront payment option but will hold you to a monthly committed use payment structure for the duration of the term, regardless if you used the services or not.  Nonetheless, there are plenty of ways to save money on VM instances deployed on AWS, Azure and GCP.

Saving Money on VM Instances on AWS, Azure and GCP

Due to the number of variables involved in calculating a cloud pricing comparison in 2020, no general comparison of cloud pricing is likely to go into sufficient depth to answer every question each customer may have with information relevant to their specific circumstances. Where a cloud pricing comparison can be helpful is in comparing what you are paying compared to what you could be paying if you were to adopt a series of cloud management best practices.

For more accurate results, pull up each cloud provider’s price list. Of course, not all instance types will be as easy to compare across providers – especially once you get outside the core compute offerings into options that are more variable, more configurable, and perhaps even charged differently (in fact, AWS and Google actually charge per second). 

It’s important to note that AWS and Azure list distinct prices for instance types with the Windows OS, while Google Cloud adds a per-core license charge, on top of the base instance cost.

It has been estimated businesses waste on average around 30% of their cloud spend each year because they are being charged for services that they do not use. This is not an oversight by cloud service providers, but attributable to overprovisioned and orphaned resources, assigning resources the wrong pricing structure, and leaving non-production resources running when they are not being used – typically those used for development, staging and testing.

Correcting these issues – and implementing measures to ensure they do not reoccur – can be a time-consuming affair depending on how many resources your business has deployed in the cloud and whether they exist on a single cloud or in a multi-cloud environment. In the latter scenario, it can be advantageous to employ software giving you a single-pane view of your resources across multiple clouds to accelerate the optimization process and enhance governance of your accounts. It would also be beneficial to integrate automated software into your business processes in order to monitor and manage your account with more ease and efficiency. 

ParkMyCloud: A Versatile Way to Cut Cloud Waste

ParkMyCloud is a SaaS platform that identifies and eliminates cloud waste. It is also a versatile solution for scheduling on/off times for non-production resources based on their actual usage patterns. ParkMyCloud solves the problem of scheduling non-production public cloud resources to turn off when they’re not being used. Furthermore, in addition to helping businesses cut cloud waste, ParkMyCloud ensures costs stay optimized with our policy-driven software. 

You can use the ParkMyCloud platform to fully optimize your non-production instances without committing to an AWS EC2 RI term that will go underutilized. The platform does this by scheduling, rightsizing, and identifying idle instances. Recently, we added the ability to view all your existing Reserved Instances in the platform so you can better track what commitments you have already made.

For your non-production Azure VMs you’ll save more by using pay-as-you-go pricing, and scheduling those VMs to turn off when they’re not needed which ParkMyCloud can help with. Our customers who have the most cost-effective use of Google resources often mix Google preemptible VMs with other instance types based on the workloads. Non-production systems, like dev, test, QA, and staging, can use on-demand resources with schedules managed by ParkMyCloud to save 65%

One of the recommendations that ParkMyCloud makes, in addition to schedules for non-production resources and size recommendations based on usage data, is to modernize a VM to a newer instance family so that you can optimize performance with the lowest cost.  If you choose to accept this recommendation to move to the latest family, then you can choose to resize right away, or to pick a time in the future (like during a maintenance window) — ParkMyCloud takes the action for you.

Additional benefits of ParkMyCloud include:

  • A single-pane view of resources across multiple clouds/cloud accounts.
  • Governed user access to groups of resources.
  • Can schedule logical groups of resources to start/stop simultaneously.
  • Tells you in advance how much each parking schedule will save and provides detailed savings reports.
  • Identifies under-provisioned resources to enhance performance.
  • Increases accountability and enforces cloud management policies.
  • Typical payback time within two months with the option for free use.
  • Better savings
  • No commitment or upfront payment
  • Price cut protection

To find out more about how you could save 65% or more on your cloud spend and eliminate cloud waste, you are invited to take advantage of a free trial of ParkMyCloud. Simply click on the “Try it Free” button and you can start saving money on VM instances on AWS, Azure and GCP today. If you have any questions about ParkMyCloud before starting your free trial – or comments about our cloud pricing comparison you would like to share – do not hesitate to contact us and speak with our team.