Today, many cloud solutions such as Google GCP, IBM Watson, Amazon AWS, and Microsoft Azure provide platforms and infrastructure offerings for every budget and need. Ten years ago, the platform challenge was assuring technical excellence, and today, the challenge shifted to making a choice. However, how do you make the right AI platform choice?
There are countless articles on the internet about comparing cloud AI platform offerings on technical superiority, usability, integrability as well as price-performance aspects. Nevertheless, buy vs. build vs. rent trade-off is still a difficult decision to make because the cloud market is getting highly commoditized, causing profit margins to shrink and price-wars to start, leaving the consumer with complicated pricing models and thousands of features to compare. Therefore, in this article, we will look from a strategic perspective and correlate the choice to the company's capability need, rather than looking at features.
Taking it one step further, I also want to add "vs. invest" to the mix for those who can simply buy-out a whole company.
One way to simplify the decision-making process is to utilize business capability modeling and capability management techniques. In this approach, we evaluate if the concern is a core, enabling, or supplemental capability and then decide to buy, build, rent, or invest. Acclaimed author and speaker Dr. Dorothy Leonard makes a valuable distinction among these three types of capabilities.
Core capabilities constitute a competitive advantage for a company and should be difficult to imitate. Google's TPU technology, for instance, is a core capability to its AI business. If your AI product or service requires a core capability, renting or even buying the platform or infrastructure wouldn't give you the competitive advantage and imitation difficulty. In this situation building or investing would be the right strategy to pursue.
Supplemental capabilities, on the other hand, are those which are required to support the normal business operations and core capabilities but could be imitated, quickly developed, or readily available to the competitors. Amazon SageMaker or IBM Watson, for example, are capabilities anybody can acquire for a few dollars and suitable to develop ML models is effort is of the essence. These solutions are an excellent choice if your company's core competency is not an AI Platform. In this case, you will need to compare the TCO between buying and renting, which drills down to CapEx and OpEx, respectively.
Likewise, enabling capabilities also support the business operations and are necessary to be in the industry but do not necessarily add value. Regulations like the Payment Card Industry Data Security Standard also known as PCI, for example, don't add monetary value to a business but is mandatory for companies who accept, process, store or transmit credit card information, in order to provide secure credit card operations. Of course, not having PCI compliance would impact the company in the long term, but the point here is that it is not a value-based offering. When you are working on an enabling capability, you can focus on forming partnerships that will come with the industry know-how added benefit.
In my experience, classifying your requirements into one of these three strategies is the first step towards a successful buy vs. rent vs. build vs. invest decision. If you want to learn more about all the aspects of making the right decision, please check out my Stanford Continuing Studies course named "Product Management in the Artificial Intelligence Era" at https://continuingstudies.stanford.edu/courses/professional-and-personal-development/product-management-in-the-artificial-intelligence-era/20191_WSP-359. Although nothing can replace a classroom experience, I'm also planning an online version as well for those who are not in this area.
Thank you for reading!