I Saved 40% on Azure VMs Just by Using This One Setting

TL;DR 

If you’re still buying Azure reservations that match your exact VM size, you’re missing out. There’s an option called Size Flexibility that lets a single reservation apply to multiple VM sizes in the same family. In some cases, reserving the cheapest VM size in the group still fully covers the more expensive ones, saving you up to 40%. 

⚠️ But there’s another kicker: on September 4, 2025, they’re changing the instance size flexibility ratios for several VM families, Redis, and Dedicated Hosts. That means you got to know who to play the game after the change. 

“Wait, That Works?”   

It started with a question from a client: 

“Can we cover our D4s_v3 VMs with a cheaper reservation somehow?” 

The answer is: absolutely, if you know how size flexibility works. 

But most people don’t. 

They simply purchase a reservation for the specific VM they are using.
For one D4s_v3 VM, they make one D4s_v3 reservation.

However, this “safe” move does not maximize potential savings. 
Plus, starting September, this “safe” move is even more of a trap given Microsoft’s ratio updates.

So, I did what any good FinOps nerd would do:  
I opened the Azure Pricing Calculator and got to work.   

Why Exact Matches Are a Trap, Even as Ratios Evolve  

Let’s say you reserve a B4as_v2 instance, and a month later your workload shrinks.

You switch to a B2as_v2 to save compute, but now your reservation no longer applies.

You’re paying full price and wasting the reservation. Double burn. 

You could cancel, but Azure only lets you cancel up to $50,000/year. Not great if you’re managing 7-figure Azure spend. 

You could exchange it, but exchanges are clunky and often mean you’re swapping a big block of reserved capacity, not just a small portion. It’s overkill for what should be a simple tweak.   

The Trick No One Talks About   

Azure reservations can be applied across VM sizes within the same family using ratios.  

For example:
One D4s_v3 reservation = two D2s_v3  

That means if you resize down, your reservation discount still applies.  
The hidden gem: not all sizes in the same family are priced the same, and sometimes the cheapest one still covers the expensive one. 

Real Example: The Basv2 Pricing Hack 

In the Basv2 family in East US PAYG prices  

• Standard_B2as_v2 = $54.9/month 

• Standard_B2ats_v2 = $6.85/month 

Both share a 1:1 ratio. A reservation for the cheaper B2ats_v2 fully covers the more expensive B2as_v2.  

Run the costly VM, pay the cheaper reservation rate.
That’s 50% off on top of the reservation discount, just by knowing how to play the ratios. 

⚠️ But with Microsoft’s September update, ratios may shift. Always validate whether these tricks still apply after the changes. 

The Test

I wasn’t going to settle for theory, this is a real game-changer, and I had to check every detail before sharing it with you. 

 
So, I ran a scenario across 100 VMs:

  

Option	On-Demand Cost	RI Cost	Monthly Savings	Savings %
1-Year RI for 100 × Standard_B16as_v2	$43,946.00	$25,908.00	$18,038.00	41.1%
800 × Standard_B2ats_v2 RIs (ratio 8, total)	$43,946.00	$3,267.00	$40,679.00	92.6%

That’s a substantial discount on top of the usual RI savings. 

This is what I call “invisible optimization.” It’s not in the Azure Advisor. It’s not in your CSP recommendations. You have to know where to look. 

Why So Few Get to Enjoy It?   

Three reasons: 

• People assume exact match = best match. 

• Manually finding flexibility group hacks is a pain. 

• Now add in Microsoft’s ratio updates, and the problem compounds. What worked yesterday may not tomorrow. 

To do this properly, you’d need to: 

• Compare all VM sizes in a family 

• Understand their reservation pricing 

• Calculate their flexibility ratios 

• Check which reservation gives you max coverage for least cost 

• Validate the math with the new ratio changes 

• Hope Microsoft does not change it again   

At scale, it’s unmanageable.

How We Automated It   

This is where most teams give up. Size flexibility in practice means comparing ratios, checking every VM size in a family, watching for pricing quirks, and then keeping track of reservations as workloads shift.  
Doable for one VM. A nightmare at scale. 

We ended up automating it with SavePro.  

SavePro is an Automatic Azure Cost Optimization product, and it looks at every flexibility group, finds the cheapest reservation that still covers your workload, and takes care of the purchases or exchanges when things drift.  

It also breaks big reservations into smaller pieces, so you don’t slam into Azure’s cancellation cap. 

In other words: instead of babysitting reservations, we let the system do the math, and the savings keep rolling in.  

 

Try It Yourself  

HeHere’s your 3-step test: 

• Pick a VM family you use a lot (Dsv3, Esv5, Basv2). 

• Check all the reservation prices for each size in that family. 

• Find the cheapest one that still has a 1:1 ratio to your current VM. 

If that cheaper reservation exists, congratulations – you just found a discount Azure wasn’t going to tell you about.  

⚠️ Just remember: after September 4th, Microsoft’s new ratios might change the outcome.     

Microsoft’s Announcement: Ratio Updates Are Coming 

Microsoft announced changes to instance size flexibility ratios (source). These take effect September 4, 2025 for select VM families, Redis, and Dedicated Hosts. 

No pricing changes, but the coverage math shifts. If ratios go up, your reservation covers fewer VMs. If down, more. Either way, optimization must be re-validated. 

Action plan 

1. Before Sep 4th – Review which SKUs are impacted. 

2. After September 4th Monitor reservation utilization closely. 

3. If coverage drops – Run more VMs, exchange unused reservations, or move workloads to a compute Savings Plan. 

Final Thought 

Azure reservations aren’t a set-it-and-forget-it tool. They’re a puzzle. And the flexibility lever is the piece most teams are leaving in the box. 

If you want help solving it, drop us a line.  

 

Dudi Amir – [email protected] 

https://2bcloud.io/save-pro/