CUD vs On-demand Breakeven Point

What is the CUD vs On-demand Breakeven Point?

Its a good rule of thumb for lowest cost possible when you are faced with the dilemma of whether to purchase CUD or let usage run on-demand. Before purchasing CUD to cover on-demand usage you should ask yourself if the expected cost is higher on CUD or on-demand. If you plan to use the capacity for the foreseeable future (more than 1 or 3 years) then it's a clear case of CUD purchase being favorable and cheaper compared to running on demand.

But, most times we don’t know if we would need the capacity 1or 3 years down the line. Luckily, the reality is that even if we aren’t sure we would need it for the entire commitment period (1 or 3 years) there is a shorter period for which it is still more cost efficient to purchase CUD - the Breakeven Point.

The Breakeven Point is the amount of time (usually stated in months) for which the same usage under on demand is identical from a cost perspective to what customer can expect to pay for CUD for the commitment period. So in simpler terms, the break even point is the number of months running on-demand is just as costly as a 1 Year or a 3 Year CUD depending on which CUD option you are comparing to.

Knowing the discount percentage a customer gets under CUD can actually tell us what the Breakeven point for the customer is. Lets use an example to explain:

A 3 Year CUD under public list price for N1 machines provides a 55% discount compared to on-demand price (as publicly shared by Google here). This means that if you purchase CUD then you would only pay 45% of the on-demand price. But, by purchasing a 3 Year CUD you would commit to pay this 45% portion of the on-demand every month for the next 3 years (or 36 months).

On-demand	1 YR CUD	3 YR CUD
No commitment	1 year commitment	3 year commitment
No Discount (0%)	37% Discount	55% Discount
Pay only when you use	Pay even when you don't use	Pay even when you don't use

The total cost under CUD in terms of on-demand months is then ~16 months (36 months * 45% = ~16, rounding for simplicity). In other words, the cost of 16 months of running on-demand is equivalent to 36 months under CUD - this is the Breakeven Point for 3 YR CUD in our case. We will incur the same total cost whether we let a certain resource amount run on-demand for 16 months or purchase the same amount under the 3 Year CUD option.

The Breakeven number of months mentioned above is specific for the machine type and region (N1 us-central1) and doesn’t necessarily represent the effective discounts customers are entitled to per their individual contract with GCP. There could be cases where a customer receives a higher discount for certain SKU but not for others so customers should use their induvial pricing and discount structure reflected in GCP Billing Console in order to determine the specific Breakeven point.

Another relevant point that only applies to some customers is Sustained Use Discounts ('SUD'). Its automated discounts applied to on-demand usage that is running for a certain part of the month (read more about it here). Google is depreciating it over time but if SUD is something that applies to you then make to account for it in your Breakeven calculation as it lowers your on-demand price point.

So how is this helpful?

if you are face with the decision of whether to purchase CUD or not to cover on-demand usage then you have to ask yourself - will this capacity be needed for at least 16 months? Ignoring any other long term planning considerations for now, if the answer is yes then purchasing CUD is the lower cost option for you.

If you end up needing the capacity for longer than 16 months then you have clearly made the cost efficient decision by purchasing CUD and you are paying the least amount of cost. Any usage running for more than 16 months in this case is savings compared to the alternative of running on-demand since. And on the other side - if you only needed the capacity for a time frame that is shorter then 16 Months than CUD was not the cost efficient option as it would have been cheaper to let this capacity run on-demand.

Similarly, under list prices the 1 Year CUD provides 37% so compared to on-demand the 1 Year CUD Breakeven point is ~8 months (63% * 12 = 8, rounding for simplicity). Below is a more visual representation of the comparison showing how many months of usage you get for the same cost under the different alternatives.

How to Strategically Implement Breakeven Point in Your Planning

When teams are planning to increase capacity, it will be challenging for them to predict how long they would need it for if you just ask them for how long they need it. A better framing using our above example could be whether they need for more than 16 months (3 Year CUD option) or less? If its less then you can also compare to the 1 Year CUD Breakeven point of 8 months as that might be a better option relative to letting the usage run on-demand.

This way you are giving them more real numbers and they can estimate if they would need the capacity within this timeframe. Be sure to explain why you are sharing these specific numbers so teams understand how their decision would impact the overall costs.

Having clear commitment from teams around time also generates accountability as CUD should be tracked and documented for purchase time and reason for purchasing as well as forecast.  

How to leverage Breakeven Point?

Additional way to advance these conversations (and generally good to have) is to create a comparison data sheet that acts as a calculator that can be used for every conversation with a team that is planning to increase capacity.

The team would input the amount and resource type and the 'calculator' would show the team what the cost alternatives are under on-demand, 1YR CUD, and 3 YR CUD (let me know in the comments if there is interest and I'm happy to share a prebuilt calculator). You can use it to show teams their planned increase cost under the various options and drive more accountability since vCPU tends to be an expensive resource ultimately costing companies many dollars.

It is also helpful if your company has an approval mechanism that requires you to show the cost under the various options so your leadership can decide on the next steps.