Skip to main content

«  View All Posts

OCI vs AWS vs Azure: Which Cloud Platform Dominates JD Edwards?

August 5th, 2025

17 min read

By Nate Bushfield

 

This session, led by Stuart Peterman, focuses on selecting the right cloud platform for integrating AI within JD Edwards environments. The discussion covers the strengths and considerations of three major cloud providers—OCI, AWS, and Azure—examining their suitability for AI use cases, particularly within enterprise resource planning (ERP) systems. This session emphasizes the importance of aligning AI models with specific business needs, cost efficiency, data integration, and compliance. The session provides insights into how JD Edwards can benefit from AI-driven process improvements, including predictive maintenance and supply chain optimization, and offers guidance on evaluating cloud platforms based on performance, integration, and long-term scalability. 

 Ask ChatGPT

Table of Contents   




  1. Introduction and Background 
  2. JD Edwards and AI Readiness 
  3. Factors to Consider When Moving to the Cloud 
  4. AI Tools and Services from Cloud Providers 
  5. Comparing AWS, OCI, and Azure 
  6. Closing Thoughts 

Transcript

Introduction and Background

Hi, guys. My name is Stuart Peterman. I appreciate you attending the session. This session is going to be primarily about choosing the right cloud for a potential future in AI. So we're going to compare the big three that I typically see with JD Edwards: OCI, AWS, and Azure. And just going through a little bit of my background. Throughout my career, I have been a part of companies, some of the largest cloud-native companies in the world. I spent some time at Salesforce. I was with NetSuite for a while as well. And with those, of course, they were born in the clouds. There really wasn't any need for a cloud migration. But from that point on, I've also worked quite a bit in migrating existing on-premise clients into various different cloud platforms, primarily like I mentioned earlier, OCI, AWS, and Azure. So from that standpoint, I've had to have a holistic view and a broad understanding of what cloud platforms offer, what benefits, what drawbacks, cost analysis.
And as it's become more prevalent in the scene, AI, a lot of companies are looking at AI as a solution for potential inefficiencies within their organizations.

So what I've helped companies do over my tenure is go through a deep analysis of every piece of their organization, different departments, cost analysis, existing contracts that they may have, you know, so that we can analyze CapEx versus OpEx hits for a cloud migration.
And through that I've been able to hone in some skills on understanding really from a high-level perspective what might be the best fit for each individual company. So I'm going to offer my expertise from that standpoint and go through each specific cloud platform, JD Edwards as a whole and how it is ripe for AI and existing process improvements via AI and automation.  And then of course, I've put together a, a little cheat sheet on what different cloud platforms may offer, a cost of those various AI solutions and services, but also some of the analytics platforms and data warehouse data storage capabilities. And so I'm just going to take you guys through that, hopefully offer a little bit of insight from a, from a bird's eye view. This will be a little bit of a, a departure from maybe some of the sessions you've seen today, which are a lot more technical and deep dive. I attended Drew's session a little bit earlier. I also attended Manuel and, and MO's sessions. So this one's going to be a little more on the surface.


JD Edwards and AI Readiness

So like I said, we're going to dive into the state of AI and JD Edwards, what makes JD Edwards ripe for AI innovation, how to evaluate the right platform, cloud platform specifically for a potential use in AI. You know, whether that's tomorrow or five years from now, you want to set yourself up and the decisions that you're going to make today are going to impact your ability to utilize certain AI platforms. And then of course, the strengths and considerations of the various cloud platforms that we have. I'll leave some time for Q&A at the end as well, but we only have 30 minutes. So I'll try to keep this as brief and informative as possible so that at the end, you guys can ask any questions that you may have and I'll endeavor and do my best to answer them. Those that I don't have time to answer or I'm unable to answer, I'll follow up with either myself and, and another resource or I know Anish is going to be on this as well. So Anish, you can chime in if need be.

So let's jump right into the current state of JD Edwards and what makes JD Edwards right for using various AI models. I'll start by giving JD Edwards a ton of credit. You know, I've worked for a few different companies, some ERP systems, some CRM systems, as I mentioned a little bit earlier.
But JD Edwards is an absolute iron horse and I've come to know that in my time working for them, the user base is extremely loyal and they have done quite a bit to modernize JD Edwards using things like, you know, orchestrations, Udos, low code tools for workflow extensions. But all that to say, JD Edwards is still extremely ripe for AI enhancements. There is quite a bit of room and, and a lot of data-rich environments. You know, a lot of companies have been on JD Edwards for decades at a time. So there's, there's a lot of enterprise data that can be used to build really complex models that can really span a lot of different departments within an organization.  But you know, I, I'll say this at the end, but I'll say it early as well. I think the correct approach, and I believe Manuel and MO have reinforced this as well, is to take some quick wins, start small and build on that so you can utilize the data that exists in, you know, those data warehouses. You know, if you're using ADW today, which you know, I'm, I'm a big proponent of, or if you're, it's sitting in a data lake somewhere kind of ripe for use within an AI model. But there are also quite a few manual repetitive tasks that still dominate JD Edwards. Of course, orchestrations and Udos and things like that have, have really paved the way for innovation in that sense. But there are still quite a few organizations out there that are either on releases of JD Edwards that are too early to utilize orchestrations and things like that, or, you know, they just haven't quite figured out how it best suits their use case.

So with that, you know, those data-rich environments and the combination of, of those manual repetitive tasks that dominate JD Edwards, it really opens the door for use of, of really complicated and, and robust AI platforms. There's also pretty heavy growing cloud adoption within JD Edwards and, you know, that's not necessarily full use and full hosting or a full migration into a cloud platform. It can be a hybrid cloud platform. I mean, there are 1,000,000 different reasons why that might be the best fit for a specific organization. But you know, JD Edwards companies are really starting to realize that the cloud isn't just a buzzword or something that you want to move to, to be more modern quote, unquote. It opens the door for a lot of functionality. I mentioned earlier, you know, Oracle's autonomous database, autonomous data warehouse, you know, Oracle Analytics Cloud, those things really open the door for usage of those robust platforms that Oracle and AWS and Azure have to offer as well. You know, Azure with Power BI.  Along with that, there's some labor shortages, there's supply chain disruptions, expectations for faster decision-making. Again, that all play into increasing a company's usage or at least starting to look in the direction of utilizing AI models, whether they be, you know, custom models that are built specific to an organization's use case or pre-built AI models that, you know, you can kind of tailor to fit your use case. Now with JD Edwards, there are a lot of companies out there that could use it for things like predictive maintenance in a manufacturing environment or, you know, if you want to get a little smarter when it comes to supply chain and demand forecasting.
So all these things make JD Edwards extremely ready for AI innovations.


Factors to Consider When Moving to the Cloud 

But if we jump into the specific factors that you should consider when you're looking at potentially moving to a cloud platform for future use of AI or future use of the various analytics or data storage capabilities that that cloud platform may offer, there are a few things that should be top of mind.
And like I mentioned a little bit earlier, we only have 30 minutes. I have a full cloud diagnostic week-long test that I have put customers through. But you know, with, with the shortened amount of time that we have, I'm going to keep it fairly broad and not dive too deeply into each.  Some of these may fit you and your organization's use case and, and what you guys do as a as an organization or it may not. But really the first thing you should look at is where is my data and in what state is it, you know, is it, am I utilizing a database data warehouse? Or is it in a, you know, a large unstructured data lake that we may need to spend some time to clean? It's an old adage. It's used pretty often. But bad data into an AI platform leads to bad results. So looking at your data from a real honest and holistic perspective and understanding what does it look like today, where does it sit today and what would I potentially need to do in order to have clean usable data go into an AI platform.

You also need to look at what your needs are. You know, when you're looking at AI needs, do you need a custom AI model or, or can you use an existing pre-built model that, you know, all three of the cloud platforms that we're going to talk about in a minute have access to, you know why why would I need a custom AI model versus a prebuilt model? You know, the answer to that question is there are 1,000,000 answers to that question, but companies do have unique requirements that general purpose models can't effectively meet. Some factors that may play into that is I don't know if I'm a company in healthcare or maybe I'm in manufacturing and I have a jargon or document types or workflows that a pre-trained model just won't understand very well. A custom model can be trained on specialized terminology and, and different data sets to improve the accuracy and relevance of that data and of those AI models. Or I don't know, let's say you are a company that has a lot of custom workflows for approvals or, or maybe niche customer service rules or maybe a proprietary logistics operation that a standardized template model won't really fit well.
So the off-the-shelf models aren't going to align with those specialized processes that could potentially produce unreliable results. Or, you know, the AI model to give you data that is not and insights that are not exactly accurate.

On top of that, you know, do you have the internal skills and resources to manage an AI or cloud platform model for that matter? You know, probably unrealistic for a company to say, Oh yeah, I just found a guy who was an expert in building AI models or, or adapting existing AI templates to our use case. But you may have someone who is either willing to learn those skill sets and manage it with the help of let's say a third party, or you can have a a specific individual that is a is was a cloud engineer in a past life or has some cloud engineering background.

On top of that, and, and in my opinion, not should never be the most important factor in these decision-making processes, but should certainly be top of mind at all times is cost predictability. Both the cloud platforms that I'm going to talk about and the subsequent AI models that they offer can have cost that can spiral out of control if you're not careful. So you know, how important is those fixed versus variable cost? How important is CapEx versus OpEx to your organization? How much are you willing to spend and what sort of platforms are you willing to entertain or look into to fit within that budget cycle?

And what I like to use is a, is a performance per dollar sort of metric. And there are certain cloud platforms and AI platforms that shine when you say, OK, per dollar, this platform or this model is going to deliver these results. From there also there's of course integration requirements and that is both from the cloud and the AI perspective. How easily can I integrate my existing ERP and ancillary systems into this cloud or AI platform? In this case probably specifically JD Edwards, but you know, I don't know, maybe Avalara, you're using Avalara for tax. There are a number of different third-party ancillary applications that you would need to consider in this.

And then of course, not to be left out, compliance and security. I mentioned a hybrid cloud approach previously. There are always industries that are subject to compliance and regulation that should be considered making a decision like this, whether that be GDPR or Fedramp or whatever it may be. You know, maybe you keep some of those core financial systems on premise or, and then the other half is in the cloud or you know, any number of different variations that should be considered when you're looking at, OK, is the cloud a viable option for a potential future in AI?


AI Tools and Services from Cloud Providers

Jumping right ahead, I wanted to talk through some of the AI tools and services that are offered from the different cloud platforms. We can first talk about AWS. We'll talk about Bedrock to start.
That is a service for essentially accessing those foundation models. So if, if, if you're an organization that needs to have access to specific foundation models, you can look at things like Bedrock, you know, it's, it's, it's a really strong and mature platform. There's really no training needed. It's serverless. There's fast scaling associated with Bedrock. That said, it is an extremely high-cost solution. It's a high cost per token versus Azure's open AI or Oracle's comparative offerings as well.

You can also look at SageMaker, which is a full end-to-end machine learning platform. So you can build, train, deploy those ML models. And again, similar to Bedrock, it is a very mature, best-in-class solution. It's got very deep controls. It's got a lot of customizations. It shines from that, from that standpoint, but it's extremely expensive for large training jobs. So, you know, smaller jobs of course could be optimized with some savings plans. But it is and this is going to be a recurring theme with, with any AI model via AWS, it is almost always going to be vastly more expensive than the other two options.  And that goes for the cloud platform as well, but we can talk in more detail about AWS versus OCI versus Azure from a cost perspective. But from a performance per dollar, if you're looking at it from that standpoint, you know certain cloud platforms offer very strong financial incentives for a migration from either on Prem or maybe a hybrid other setting or even another cloud platform if you if you're thinking about replatforming.

We can also talk about Recognition. Recognition is probably one of those solutions that's not going to be very enterprise-oriented from a JD Edwards customer base perspective. But it, it's used for video analysis, pre-built images, so we can analyze things like faces, objects, text. And it's, it is very accurate. It's it's quick to deploy. It competes pretty well with the Azure vision, but it's less integrated into ancillary applications. It's mid-priced, it's not outrageously expensive. And, and again, I want to say this all with the caveat of different use cases may have different cost structures associated with it, but in comparative, in comparison with the other two similar offerings, it is at least roughly competitive.

And then of course you have Comprehend. It's a natural language processing service.
It's, you know, entity extraction, language detection. I think at one point I had someone much smarter than myself explain like I was a 5-year-old what a natural language processing service actually does. And I think his analogy was, OK, so I'm going to tell you like you are a 5-year-old.
Let's say you have a a robot toy and you tell that robot toy, you want it to tell you a story.
It understands your language, it ingests that language and then it spits back out a story that you want to hear. That's far oversimplified, but it is a solid NLP service.  It's fairly accurate. It's it's fairly accurate. Azure's cognitive services are slightly easier to integrate in business applications.
But all that to say, it is also reasonable in terms of cost.

The bottom 2, recognition and comprehend are the 2 that are fairly easy to digest from a cost perspective, but both not exactly ripe for an enterprise utilizing JD Edwards.
I'm not going to go through every one of these of course. I will provide this slide deck to anyone who would like to read through it, including the cheat sheet that's attached as well.
But just to quickly cover the OCI offering as well.

The OCI AI services is a pre-trained model for easy application and integration. It's, it's tailored to businesses and ERP data types, especially Oracle workloads, including JD Edwards. So it's got a really tight integration to JD Edwards Fusion. It is a tighter, more robust integration than AWS or Azure. It is made for business applications and it is specifically made for Oracle business applications.  Now not to say that other ancillary applications can't, it's not friendly with those, but it is a very simple integration process for existing Oracle workloads. And on top of that, it is significantly lower in cost comparatively than the AWS or Azure comparative offerings. I have seen them as different as 30 to 40 percent, 50% even in some cases for comparable solutions on the Azure and AWS platforms.

And then you have the Oracle Data Science on OCI, which is that is building and training, deploying those models with open frameworks. So when I was talking more about the custom versus the prebuilt templated models, this is more of the custom model for, for building, training, deploying models with, with kind of open frameworks. It's got really great flexibility and, you know, it's less expensive from a direct GPU comparison standpoint. You know, I've seen it as low as 20 to 30% lower than the other two comparatively from the GPU standpoint.


Comparing AWS, OCI, and Azure

Jumping over into my little cheat sheet that I've put together for you guys, it is using this this legend, whereas these little money bags goes from one to three. Same with the lightnings that that pertains more to performance And then of course integrations.  So when you're looking at utilizing cloud platforms for a potential use in AI, you want to look at it from, OK, what is it going to do? How much is it going to cost? Will it perform up to my standards as an organization and how tightly and how well can it can it integrate to my existing ERP system and ancillary applications that are tied to it as well.  So if we're talking about OCI, we're talking about Oracle AI Services and Oracle Data Science. You know as I mentioned earlier, it is the lowest cost solution on the market. There are exceptions to every rule. Please do not come knocking on my door and telling me that it was ended up being 2% or 1% less expensive or more expensive than I said it was going to be. There are exceptions to each rule, but from my years of doing research, market research on these various platforms both from the cloud platform perspective and the AI offering perspective, including the Oracle Analytics Cloud or Power BI, these are and have held true. Now Oracle is going to be the lowest cost for both of these. The performance. Oracle has a very robust and well-built AI model. It is going to do the job. It is geared towards enterprises. So whereas some of the other applications may have a broader scope, AWS certainly has a much broader scope of AI offerings. If we're talking about what you would be utilizing as an organization as a JD Edwards user, you're really using it for what Oracle has geared it up to be, which is as an enterprise application and AI model for performance and an increase in efficiency within the various departments that you're using it for, whether that be finance, warehousing, or shop floor operations, any number of things you can apply these models to.

And I actually have a session tomorrow, not to plug my tomorrow session in the middle of my existing session, but I have a session tomorrow at 1:30, I believe Mountain Time, 3:30 Eastern Standard Time, Talking specifically about these use cases, what we have seen, what I have seen specifically and how these can be applied to you as a JD Edwards user and your organization.  It also, of course, has a very tight Oracle application integration. It is, you know, purpose-built for Oracle data. Of course it is friendly with other ancillary applications. And as you may or may not know, Oracle has entered into agreements with Microsoft Azure and AWS Google Cloud to be friendly with those in terms of integrations and utilizing their solutions as well. But that is to say that it is going to run more easily and more easily integrate to the Oracle applications, JD Edwards included.  Of course, when we're talking about AWS, as I mentioned before, they have a very broad deep offering. That said, it's going to be on average significantly more expensive to run. It's very good on performance. It's very customizable. But oftentimes the integrations can be, you know, DIY integrations, maybe it's a stand-alone platform. It could be fairly flexible, but again, you're dealing with a solution that will certainly have what you want. But if we're talking about it from a performance per dollar perspective, it’s typically going to get beat out by OCI and Azure.

Then jumping over to Azure, you know, Azure has a very compelling offering as well. It usually falls somewhere in between Oracle and AWS. They have really tight integrations. If you are a Microsoft-heavy shop, Microsoft 365, Teams, Dynamics, you know, it has very tight integrations, very native application integrations to those Microsoft workloads. But that's probably not news to you if you are a Microsoft shop. I'm sure you know that already. Again, all that to say, those existing agreements that are in place between Oracle and Azure make it pretty easy to integrate into those as well.   Just to go through a kind of a final consideration and boil it down, AWS is going to be a really good option for custom flexibility, but it does in turn have the highest cost risk. I have seen companies utilizing AWS AI offerings and their costs have spiraled out of control very quickly.
And I'm not saying that to scare anyone on this call. I'm really only saying it to say that it's something that you need to consider and review internally and keep a tight leash on when deploying. If that is your inevitable choice.  OCI will still contend and is going to be great for Oracle ERP-driven workloads and is going to be the lowest cost while still offering a very strong AI solution both in the custom and templated model versions of it. And then Azure, it's going to be really good for enterprise-wide AI adoption as well and of course has a very strong integration into existing Microsoft products. So if you're a Microsoft shop, of course, that is something that you're going to want to review.


Closing Thoughts 

My closing thoughts though, when it comes to this, and I'll try to leave as much time at the end for questions as I can, but I want everyone on this call to consider matching the platform to business realities. Don’t look at AI as a sort of hype like you would have maybe looked at cloud 15 years ago. Your data, your ERP ecosystems, and your AI maturity should shape that cloud decision, not just what platform has the most features and what is the most bells and whistles. AI is not a one-size-fits-all sort of approach. Some companies need very deep pipelines, others want intelligent automation or better forecasting and can apply templated models to that and look at it as a crawl, walk, run approach. I think you'll hear that quite a bit over the next few days. But if you can find quick wins to apply an AI model to, then you can expand from there.  You don’t need to bite off a piece of the apple that is too large right off the bat. You can, you can really prove these models out over time and I think that will lead to a much better result as an organization and a lower risk. And then of course be strategic about cost when it comes to approaching both the cloud platforms and the AI offerings associated with those cloud platforms.

So as I've mentioned, OCI is going to offer the lowest total cost of ownership for those Oracle-heavy workloads and really does have a very attractive GPU/AI price. Azure really does provide some cost efficiency for organizations. We don’t want to rule them out. They have a, they have a strong offering as well, but you know, you’re not limited to Microsoft AI. You can look at ongoing future partnerships to understand how you can maybe look at a hybrid approach.  And AWS offers a wide breadth of innovation when it comes to cloud. They were the first on the scene. They have the largest and broadest offering, but cost controls can require some serious discipline when reviewing these, especially if you have really inference-heavy or custom training use cases. You really need to consider that before diving too deeply into the AWS AI offering.

The bottom line here is that the right cloud is not the one with the most tools, most bells and whistles. It is the one that lets your business teams turn data into intelligent decisions.
You know the processes that you have into intelligence and cost into value.  So with that, I appreciate your guys’ time. If there are any questions, I'll be happy to answer them and I'll take a look. Also, of course, while we look into it, I am Stuart Peterman. I’m the Enterprise cloud executive at ERP Suites. I appreciate everyone’s time and if you need to get in touch with me for any reason, here is all of my contact information. My email is speterman@erpsuites.com. And of course, we will be at Blueprint. We’ll be at the various user groups. Atlanta, Houston—will be in Plug June 9th, Houston July 21st. And of course, I've actually got a few podcasts out there that I think should be released over the next couple of weeks on the "Not Your Grandpa’s JD Edwards" podcast, presented by ERP Suites, available both on YouTube and Spotify.

All right, that’s 30 minutes. That is all the time we have guys. If there are any questions that have gone unanswered, we will follow up with them. And again, if you guys need to get in touch with me for any additional insights. I was forced to, I was forced to go through this in more of a quick bite-sized portion of it. But I have spent a lot of time on these in the past, both in front of customers and, you know, from a research perspective. So I'd be happy to offer my two cents if you and your organization need it. Thanks again and I hope you guys have a good rest of your day.

 

 ChatGPT

Nate Bushfield

Video Strategist at ERP Suites