How Enterprises can Calculate ROI from their HPC Spend

January 25, 2022

High Performance Computing (HPC) enables researchers, scientists, and engineers to make data-driven decisions and predict product performance before a functional prototype is built. It is a logical step for both private and government organizations to infuse predictive analytics using specialized HPC platforms in their development standard operating practices. Intelligent data-driven decisions drive sustainable market growth, effective strategies, better products, and an overall healthier enterprise culture. Before this adoption takes place, HPC must prove to be cost-effective. 

This blog focuses on the HPC spend ROI from a product development perspective. As a FEA analyst I will be looking at it from a purely numbers perspective using a real-life example from a project that I completed a few years ago in the healthcare field, but let’s first talk about the basics.

Financial metrics to consider for ROI on HPC spend

  • Return on investment (ROI)
  • Total cost of ownership (TCO)
  • Net present value (NPV)
  • Internal rate of return (IRR)

Starting with the basics: how do we calculate ROI for HPC? 

One way to compute the ROI is to use its basic formula:

That was the easy part, right? Now comes the hard part.

To estimate the ROI, we need to determine two things:

  1. the cost of investment
  2. the gain from the investment

The cost of investment is usually easy to estimate but how do you measure the value of research and development outcome? 

What are the derived gains from HPC on the cloud?

To name a few (the ones with the most significant impact on the bottom line):

  • Enhanced productivity and Agility
    Cloud-based HPC is readily accessible in both interactive and “batch mode” from the network most of the time using “instant start” and no administrative tasks
  • Cost reduction
    Cloud CPUs and GPUs are usually cheaper than HPC but work less (think of 15% utilization vs. 85% utilization)
  • Scalability
    HPC can run large jobs over multiple machines
  • Collaboration
    Easy to share data using the systems that are networked together

While it may be difficult to attach dollars to all these gains to compute ROI, it is not impossible. Most of the time ROI depends on who is asking the question. 

Real-life HPC ROI case study

High performance computing return on investment equationsThe real-life HPC case study below shows how ROI was calculated on a pilot project.

A few years ago, I worked as an FEA analyst for a one-off project that had a supply chain and prototyping challenge. This large medical device company needed to design an access device that had a metal component with an exposed window at its distal end. They had an idea of how the window should look but needed to hone-in the dimensions and the tolerances. 

The traditional approach of prototype-iterate cycle was too long and would not meet the launch date. Thus they looked for an HPC-based solution enabling them to use a simulation-based virtual design of an experiment allowing the company to reduce development time and bypass supply chain constraints.

The HCP-based approach worked great for them. We collaborated on the cloud using the JARVICE™ HPC platform on the Nimbix Cloud, enabling us to make decisions based on benchmarked finite element data, and launch the product on time and on budget.

What was the ROI of this project? Below, you will see side-by-side scenarios (some of the costs and charges were scaled for confidentiality purpose):

hpc roi expense breakdown

Assuming that the gain from investment is the amount the client did not spend on the prototyping and testing, the ROI for this project can be estimated as:
hpc roi gain from investment equation
While the long-term cost of lost revenue caused by project delays due to supply chain disruption is not factored in, one can see that the ROI for the HPC simulation on the cloud approach is significant. As commercial enterprises adopt modeling and simulation using HPC in the cloud and the price of the floating-point operations per second (FLOPs) are getting lower, the ROI on HPC becomes even more attractive and can be estimated both at the macro (enterprise) and micro (project) levels depending on particular needs. 

Subscribe to our quarterly email newsletter to get resources like this in your inbox!