For years, the only direction that cloud prices would go was down. But this is already changing.
Google Cloud Platform has introduced significant price increases across various core services around storage. It’s only a matter of time before the price hikes affect compute.
With inflation rates running high around the world and industries suffering from shortages in technology like advanced semiconductors, it’s unlikely for the public cloud to continue getting cheaper as it used to.
You didn’t expect cloud service providers to lower their prices forever, did you?
We all took price drops for granted because the cloud price wars generally inspired providers to do that.
AWS has reduced prices a total of 107 times since it was launched in 2006. Such regular price cuts were a standard way for cloud providers to compete and benefit from economies of scale.
What does that say about the gross margins cloud providers enjoy? They tend to be incredibly high.
In Q1 2021, Alphabet (parent company of Google) shared an impressive 29.7% operating margin. Microsoft recorded an even higher one of 40.9%.
Does this mean that AWS, Azure, and Google Cloud will keep lowering their prices? Considering the current inflation rate and supply chain troubles, that’s not likely to happen.
The next wave of cost savings might never come. In the beginning, providers achieved optimized costs using a mix of Moore’s law, architectural improvements, and economies of scale. Today, these levels are close to saturation in terms of delivering further cost benefits.
Let’s take the economies of scale as an example. As their customer base and usage grew from hundreds to millions of customers, cloud service providers drove optimizations from investments and shared resources (think mega datacenter operations).
But these operational efficiency gains are now slowing down as hyperscalers actually reached that scale.
We’re already seeing the effects of this. When it comes to Google, things don’t look so rosy anymore. In 2011, the provider increased its App Engine pricing by 2x to 10x. And in 2018, Google Cloud users saw a 1400% price hike on Google Maps API.
Meanwhile, AWS has supposedly halved the amount of credit offered free of charge to charities to access IT services, so it might be on the lookout for cost savings too.
So, what can you do to ensure that your company survives the future price hikes?
How to prepare your business for cloud cost increases
Reserve cloud resources while they’re cheap
With a price hike on the horizon, reserving cloud capacity might look like the most sensible thing to do. And it might as well be if everything else in your business gets stuck in time for the 1- to 3-year period that the reservation lasts.
Real-world examples show that even giants with entire teams dedicated to cost optimization fail to make accurate forecasts. For example, Pinterest users once spent so much time on the platform that the company’s cloud bill went flying way over the initial estimates.
Another common scenario is overprovisioning resources. a16z found that committed cloud spend is typically ~20% lower than actual spend. Some companies a16z surveyed reported exceeding their initial estimations by at least 2x!
Optimize your existing resources
Alternatively, you can take advantage of specialized automation tools that optimize cloud costs using different methods:
- Selecting the right types and sizes of cloud resources.
- Scaling resources up and down automatically to avoid overprovisioning and waste.
- Decommissioning resources that aren’t being used.
- Automating the use of spot instances that offer even 90% discounts.
How does an automated cost optimization tool work in practice?
Automated instance type selection
Our team ran an application on a mix of AWS on-demand and spot instances. We used an automation solution to analyze this setup and look for the most cost-effective spot instance alternatives.
The solution suggested an instance called INF1. But wait, isn’t that the supercomputer with the powerful ML-specialized GPU?
Once we checked the cloud prices, it turned out that at that time, INF1 just happened to be cheaper than alternatives. No engineer would bother looking into this type of spot instance on their own.
In this scenario, the initial overprovisioning was pretty high, causing the company to lose money on resources that the application didn’t use.
By implementing automated optimization, the level of overprovisioned resources was reduced significantly, leaving a little headroom to cover the application’s demands.
Automation can take many optimization tasks off your plate, whether it’s selecting the best cloud resources or scaling them for minimum overprovisioning and optimal performance.
Would you like to see what else an automation solution can do to slash cloud costs? Book a demo!