Teradata Corp (TDC) September 9, 2026 Conference Transcript

Thanks for joining us for day two of the Citi Global TMT conference here. Today, my name is YC Wong. I'm part of the software analyst team at Citi. We are excited to have Teradata CEO, Steve McMillian. Steve, welcome back. I know you've been a couple of years since you've joined us.

It's great to be here, YC. Looking forward to the discussion and telling you everything that's been going on at Teradata.

Now, that's awesome. I mean, this year, definitely a lot have happened since the beginning of the year. Maybe you can just start off with your background, what have you been doing, and the company.

Yeah, I joined Teradata in June of 2020, really with a mission to look at how do we modernize the company and make it relevant in the cloud space. And so really taking Teradata's fantastic on-premise technology and making it available to customers in the cloud as they modernize their data states and started using cloud technologies to really support their data platform. And so when I joined Teradata, I said, look, at our core, we're a technology company. You know, we've been doing a lot of services up until that point, but we've got so much intellectual property in our Teradata software and the platform that we have that I think that exploiting that for the benefit of our customers is really the core for us. And over that period of time, we actually transformed the company. Now, almost half of our recurrent revenues are in the cloud. So we made tremendous progress there and developed a really open and connected data platform for our customers. But I think what's been really interesting is if you think of that as Teradata 2.0, we've actually moved into a new phase of Teradata 3.0 when we're looking at AI now. And so that's driven a number of changes for us recently.

Yeah, I mean, that's an exciting announcement back in May, I believe, when you guys have a big festival out at MISD. So what is the autonomous knowledge platform? Like, is it just a repackage of what Teradata has been doing? There's been a lot of changes. Can you kind of help us break down what it is?

Yeah, I think really addressing this new world, what we did, we brought in new talent at all layers in the organization. So a refresh management team to look at the world of AI. We have a new chief product officer that's been with us for 14 months now. And really, the whole team came together, not just to repackage what we're doing, but to really think about what is the platform of the future where AI agents and humans can work together and get the most out of their data to really cause business impact. And so that autonomous knowledge platform is essentially a complete architecture and framework where we've launched new products and capabilities at every single layer of the stack. And by the word autonomous, we're really agentifying our technology as well. So not just using agents to code, but actually agentifying the entire product stack. And I think nothing symbolizes that more than the very top of the stack, which we think of our workspaces, where humans and agents work together. We have a technology there called Terra, which is a harness that we've developed for AI agents. It means that our customers can use whatever language model that they want. They can use ChatGPT or Claude. But what's becoming more interesting is using smaller language models and working with customers in Europe, looking at some of their requirements around regulatory compliance, use models like Mistral. And what this harness enables the Teradata platform to do is use the right agent at the right time for the right outcome. And you can really optimize the cost of running your overall environment. And that's just one example of some of the innovation right at the top of our stack. On the very bottom of the stack, if you look at our infrastructure, we actually announced a new Teradata factory offering, our new on-prem technology, GPU accelerated, using the NVIDIA product stack built into the data platform so that on-prem you can run your AI solution and your data platform right next to each other without moving data around and still have all of that great data and financial governance that Teradata provides. So new innovations at every single layer of the stack, the context work that we've been doing is super exciting in terms of letting AI agents really understand what enterprise data is all about. You can do something like define what a customer is so you can ask interactive questions around the customer. So it's not just a repackaging, it's a completely new innovative technology set. Yeah, no, it sounds like the whole team's definitely been hard at work over the past few quarters here, but zooming out a little bit more, we'd love to talk about like somehow how is the enterprise AI agent adoption has been going, like why Teradata decide to go on this path into the autonomous platform yeah i think um you know we we recently did a study across um a thousand or so um data leaders um in large enterprises across the world um and i think what we're finding is that that kind of headlong rush to the cloud is kind of slowing down so we've been thinking much more about you know how do we grow our overall business both on-premise and cloud? How do we respond to our customers' requirements? Because it's not just about data modernization anymore. It's about getting value from AI. And I think what we've proven using our forward-deployed engineering capability that we've put in over the last 12 months is we can take those initial ideas and turn them into production reality at scale, utilizing our services capability utilizing the new technologies that we have inside our data platform i've got a fantastic example of um uh doing some work for the military in a european uh country we've actually been working with them we started as a pilot um to look at camouflage design and using ai models around camouflage design and um we've worked with that uh military organization to take that from an idea to production scale for their entire military operation, which is a super interesting use case.

Yeah, I know there's certainly a lot of opportunity on the public sector as well, but you mentioned FDE. Can you give us a sense of, everyone's been talking about FDE at this point, like Palantir kind of started it a few years ago, Salesforce, ServiceNow. Can you give us some flavor of how is your FDE working and what is the opportunity that you see with FDE?

Yeah, I think the great thing about our forward deployed engineers, we had a super consulting and sales and SE team, and it really just formalized the go to market structure around working with customers on what the real business problems are. And I think that manifests nowhere as much as in the context layer. So working with customers to actually look at their data models, their business knowledge that has been imbued into their systems over time, and then working on a specific business problem. It may be a customer care problem. It could be a supply chain problem. And developing the data products to actually help solve that problem in a very pilot phase. So that's what our forward deployed engineers do. And they're organized by industry. They're organized by solution set. And they bring that knowledge and capability to our customers very quickly. And then we can use our AI services team to actually scale that out and use our technology to deliver those ideas into production in real time. And that is really the challenge that a lot of the data leaders that I talk to every day are the core challenges. How do I move from pilot to production at enterprise scale? And that's the problem that Teradata can help them solve.

Okay. No, there's definitely a lot of different levels how FDE is able to help an organization get more ROI. Is there any internal metrics that you trace, oh, this is actually worth my time to invest in? Because what we heard, like FDE costs a lot of money, it's an expensive services part of it. Like, what are you seeing that makes you want to continue to invest in FDE?

Yeah, I think so our FDEs usually are developed in some form of proof of concept with the customer. So we track how those proof of concepts are moving into real opportunities, how those opportunities are then translating into incremental ARR. They start off by potentially generating a services engagement to do that implementation and then generating technology or product-based ARR as a result of that. So we have a whole pipeline measurement and management system that gives us those leading indicators of moving from that proof of concept right the way through into technology implementation. Okay.

Is there any other like hard numbers that we can say, hey, this is improving my sales cycle, improving my delivery time?

Yeah, I think if you looked at our earnings comments over the past 12 to 18 months, we've seen a continuing increase in number of proof of concepts that we've been doing. I gave some of those numbers in our earnings calls. But the really interesting thing is now seeing those turn into fruition with, you know, major automakers, governments around the world, financial services organizations, telcos really taking advantage of the technology now and implementing.

Okay. With FDE, what is the conversion cycle for you to make? We're going to services upfront. What's the return that you're seeing? Is it six months out, a year out?

I think what we see is an enterprise software sales cycle emerge. It starts off with that thought and then it runs through a sales cycle. Six to eight months is a pretty good indicator of that kind of sales cycle. And it's why, you know, we knew that we, as we came into the year, we would have a tremendous amount of innovation. And you can just see that in terms of the press releases and the capabilities that are going into a general availability for us over the first part of the year. And, you know, our Teradata factory going live, our new dynamic compute engine going live, some of the new context offers becoming available in the marketplace. But it takes time for those to monetize. So as we came into the year, we knew that we returned the company to ARR growth last year. We knew that we would accelerate that this year, but we didn't bake into our number of any large upside from the new products for this year. So as we look out into the future, we see a real opportunity to continue that growth acceleration into 2027.

That definitely takes some time. We look forward to seeing that trajectory improving. Maybe going towards the product side, instead of too much finance, maybe sovereign AI, that's one thing that has been a bigger topic with Teradata Factory that you mentioned earlier. Can you give us a sense what is the opportunity with sovereign AI and how is your customer conversation with?

Yeah, we see it very clearly, especially in regulated or highly regulated industries or in governments around the world. And I think the important thing is to think about sovereign AI. You can break it down. So there's sovereign infrastructure. So making sure that you have control over your infrastructure. And we've seen a number of customers choose to deploy workloads on-prem rather than deploy in the cloud. That's one of the reasons why we oriented our investors to look at what's our total ARR growth, not just thinking about how well we're doing in the cloud as an indication of how well we're going to do in the future. Cloud's always going to be important and be a key part of our driver, but looking at the total ARR growth for the company, that's really what's going to drive our company forward. So there's an infrastructure choice there, and Teradata Factory gives our customers the opportunity to run AI workloads in their own data center right next to their data platform. And then there's AI sovereignty from a data perspective, enabling customers to choose where they put their data. They can put their data in a cloud, in a private cloud. They can put their data on-prem. And then there's AI sovereignty. So where does the AI model run? Does it run, you know, do you run it in a general purpose model like Claude on the public web, or do you run that in a cloud environment, or do you run those models on-prem? And we've developed our platform to be able to have sovereignty at all layers in the stack and give that capability to our customers.

Yeah. Is sovereign AI just kind of mainly a compliance discussion that you're having, or does it also involve customer wants to make based on better performances, the cost involved, or the data gravity of a certain...

Yeah, I think the initial discussions are certainly being driven from a compliance perspective. We see a lot of our customers in Europe and in Asia really thinking about that sovereign AI infrastructure and not running on public cloud. However, what we do find is that all of the benefits of the Teradata architecture come to life when we see very high volumes of queries, very high volumes of users, query complexity being very high. That's where the Teradata engine really starts to shine in terms of executing this workload. So it's not just a regulated industry. It starts to get into the cost of owning and operating these platforms. How much does it cost to run? Because Teradata, we solve complexity with great software rather than scaling out our compute as some of our competitors do.

Yeah. Trying to tie into what you're seeing on sovereign AI, especially outside of America, How do you see the opportunity going to be potentially impact your ARR number longer term?

Yeah, I think as we've looked at it in the past, really our on-prem business, I think, was kind of flat to decline. But I think what we're now seeing is actually we can see growth coming from our on-premise instantiations. In fact, just an interesting statistic, half of our business is on-prem, half is in the cloud, as I said before. But for the 50% of our ARR that's in the cloud, half of our customers that are in the cloud with us have also retained on-prem environments. And they're creating a fabric across their entire data platform, across the cloud and on-prem to have an integrated data environment, no matter where their data is stored. And that's a real advantage that our customers are taking and utilizing to have the best possible data platform for their particular use cases.

Okay. Does that kind of impact how you're thinking? Because historically we think about cloud transformation project, there's a certain uplift to it. How do you think about balancing customer wants to remain hybrid at this point versus moving to the cloud?

Yeah, I think that's the great thing with the Teradata offer. we offer our customers choice. And so when they want to keep that data sovereignty, when they want to keep that data inside a highly controlled environment, inside their own four walls of a data center, we offer them the capability to do that. If they want to run a VPC in cloud, their virtual private cloud environment, we can run inside that VPC with them. If they want a fully managed Teradata SaaS solution, we can run it as SaaS for them.

So we offer all of those different types of deployment models, and a lot of our customers are responding really well to that because they see it as a real advantage from a flexibility perspective in terms of if rules and regulations change, how can they dynamically respond to that? okay yeah that sounds there's a lot more opportunity like people are definitely talking about hybrid um maybe just moving pivoting a little bit with some of the transitions we are seeing with ai coding tool right um we have astra launching last week definitely causing a lot a little bit of a stir within the software industry like curious to see how are you seeing because you have terra coding you have like a difference right terra claw right a lot of new product coming out. What are the opportunities on AI coding, either just helping with modernization use cases or day-to-day work within your customers?

Look, the challenge that I've given to Sumit, our chief product officer, is to use AI and the agentification of the entire Teradata platform as an opportunity to leapfrog the competition. And so Terra transforms the way that agents and humans can interact with the Teradata platform. You can use natural language interface. You don't need to learn how to code in SQL anymore. You can use natural language interface to do administration tasks of the platform and manage the control plane. But one of the really interesting things about our Teradata harness is it actually optimizes and governs the use of agents. So a lot of our competition is essentially passing through token cost in terms of the revenue model to their customer. We don't do that. We allow our customers to use the right model at the right time, which has been particularly beneficial in countries like France, where the French government are promoting the use of Mistral, as an example, as a language model. You can plug that right into the Terra harness and utilize that as your core language model rather than using a cloud or a chat GPT. And we're seeing customers create really interesting use cases using that technology and deploying to massive numbers of users inside their environment because it takes away that skill requirement to understand how the data is constructed, what the table schema looks like. It completely leapfrogs the way that you access your data platform. So we think about Terra really as the Claude for data, if you can think about it like that.

Yeah, no, absolutely.

We are seeing a lot of efficiency gain with coding tools, Claude, Codex. just from like investments like how where do you see Terra could help drive a change within the ARR or even margins right help drive better efficiency within the organization yeah so I think a couple of things one we use a lot of AI coding tools to really increase the speed of our innovation and delivery I think if we look at the last 14 months in terms of the product development and product engineering that we've been able to execute. A lot of that has been as a result of using these coding tools. But for our customers, we actually expect Terra to drive significantly more usage of the Teradata platform. And we see it as a mechanism to allow other agents to drive usage of the Teradata platform. Now, one of the great things about the way our platform works and some of the patents that we have is we are designed for AI workload in our active compute. Our massively parallel architecture that we have says we can work with incredible high volumes of users. And so if you think about enterprises of the future, they're going to have tens of thousands of agents. They're all going to be heading the data platform at the same time with queries, so they're going to have lots and lots of concurrency of usage. And then the complexity of those queries are going to increase over time as the agents develop more and more complex queries that they're going to ask. And if you look at those different parameters around the use case, that's exactly what the Teradata Active Compute Engine was designed to address. But we've also just announced in June our dynamic compute engine, which is targeted specifically to essentially deliver the same kind of workloads as Snowflake and Databricks from an agent. So these are ephemeral compute engines that an agent or a human can spin up and spin down and save the environment. It's an offer that we can run for our customer, but not only that, they can run it inside their own environment too. And so we can just essentially sell it to our customer as a software-only solution where essentially they run it inside their environment. And that's going to be a very flexible pricing model that we're going to be able to take to our customers into the future. And it's another level of innovation that we're doing in what we call our compute substrate. That's essentially where all of the engines are that interact with the data platform. And so I think as these agents start opening up new queries, new business use cases, they'll drive consumption both to our active compute engine, which we'll do in a very nicely financially governed way, but also start to spin up dynamic compute capability inside the environment. And the agents and the governance that we put around those agents will use the right engine at the right time to deliver the optimal solution for our customer. Because nobody else in the industry has that active compute engine that's allowed Teradata to work at enterprise scale globally for the past 20 years.

Yeah, certain advantage that Teradata has been around for a long time, unlike some of these newer data platform companies that you brought up, Databricks and Snowflake. When we were out at Snowflake Summit or Data AI Summit, we recently heard a lot about how coding tools help them do faster modernization migration project from legacy platforms. What are you seeing in the last few quarters or months from comparative nature between this cloud data native platform?

Yeah, well, I think from our perspective, our retention rates have improved. They improved last year in 25, and they continue to improve into 2026. And so I think what we are seeing is our customers are using these tools to think about how to get business value out of AI. And modernization of the environment is something that Teradata offers now with our new architecture. And so it's not become as much about the modernization of the data platform. It's turning into a discussion around how best can I solve this business problem? And we've developed a framework and an architecture from Terra all the way through to our infrastructure layer that enables our customers to solve those business problems in exactly the way that they choose. We don't lock you into a particular language model. We don't lock you into our particular context capability. we are developing technologies in each layer of the stack that gives our customers choice in terms of how they execute.

Okay. It sounds like you are more willing instead of just a replacement, they're replacing certain product or you are replacing them. You're more cooperating together at certain use cases where Teradata is better at or Teradata Snowflake is better at, right? How does that solution work from a customer perspective? Do they have to get more integration needs between in order to integrate Teradata or autonomous platform?

Well, I'll give you a good example. In our context layer inside our architecture, our very first announcement from a context perspective didn't just develop context for organizations for data that's stored inside the Teradata platform. It also developed context for Google BigQuery. And what we see inside our customer environments is, especially these very large organizations, they'll have multiple engines and multiple capabilities. What the platforms that will win into the future have to be open and connected. And that's something that we built in as we designed our cloud first strategy. We knew that as Teradata, we had to be very focused at what we're good at. You know, we have $110 million or so of R&D. We have to be very focused in terms of where we invest that and the differentiation that we have as a platform. But what we believe is the platforms of the future and the agentic future will enable these agents to choose the right capabilities at every layer in the platform to solve the problem that they're trying to solve. And we believe that in every single layer of the stack, we've got the best technology that can differentiate from our competition, either from a cost per query or a total cost of ownership, but not just that, also the capabilities of the platform. And I think we'll start to see those agents driving that kind of workload into the future. You know, not very many people would, or people would be surprised to learn that Teradata appeared for the first time in the AI ML Gartner Magic Quadrant. And the very first time we appeared, we ended up as a visionary in terms of the capability that we were looking at developing for our customer set. And those products are starting to come online now, and we believe will help to drive significant growth as we move forward.

Yeah, it sounds like the big focus, definitely focus on getting enterprise ready to get agentic workload into production, right? Is there a specific vertical that Teradata focus is seeing are leading the pack?

Yeah, I think we've certainly, again, I'll talk to our context announcements. Financial services, telco and healthcare have been some of the initial industries where we've been developing that context for our customers. Now, you have to put into consideration the fact that we have developed comprehensive industry data models across multiple industries. For those of you that don't know Terri, we serve all industries from manufacturing to governments to large banks, large insurance companies. And we also do that on a global basis. So we've got great dispersion in terms of where we get our revenues from. But having a lot of experience and working with the largest enterprises in the world has enabled us to build up these context maps that we can make available to AI agents to quickly get them off the ground to deliver real business value. And we do that through our technology. We do it through our AI services capability and bringing all that together for our customers. But the initial set of industries are financial services, telco. They are really important to us as well as healthcare.

I guess now maybe we have like five minutes left here. I'd love to see if there's any questions from the audience. For a few minutes. Yeah, one from Joe.

Talk about your relationship with NVIDIA and what the hugging face acquisition, what opportunity might mean.

Teradata. Thanks, Joe. Yeah, so our Teradata factory is a ground-up re-architecture of our on-prem technology set. It's something that we co-developed with Dell. And Dell have actually committed to work with us from a go-to-market perspective to take this capability to market. But it's a GPU accelerated on-prem technology, which also runs on the NVMe, the NVIDIA software stack for running language models. And we have already deployed that technology with customers. There's a large bank in the U.S. where we've deployed that with Hugging Face sitting on top. We also did it in a bank in Australia. So Hugging Face on the NVIDIA software stack on top of Teradata to solve customer complaints analytics and also customer service interactions. What the announcement between NVIDIA buying Hugging Face is we'll see tighter integration of that software stack. And as opposed to us integrating it for a customer, it will be a self-serve software stack that you can run on Teradata Factory. So we see this as a really exciting opportunity to work with an established partner like NVIDIA to really deploy that on-prem.

Yeah, no, the open-weight model is certainly a big discussion right now. How do you see this cold-source versus open-source model impacting how Teradata communicated our life with customers?

Yeah, I think we are already seeing a lot of our customers wanting to use those open weight models. They're cheaper, they're smaller, they're more efficient, more effective. Also, those smaller language models can be trained very effectively in a particular domain. And that's why we designed our Terra harness, so that you could take advantage of the right kind of language model deployed in the right technology, so deployed on-prem or deployed in the cloud. And so that, for us, is a key part of our value proposition in terms of enabling our customers with that choice from a language model perspective to really optimize their environment, make sure that they've got tokenomics. and working out for them, right? So that's a key part of our value proposition as we move forward.

Yeah, I think Terra Harness is that still preview or it's not officially GA, right?

It's in preview just now. And we are actually thinking about giving the source code to the harness to our customers because they want to develop their own harness with certain features in it. So we will maintain a code line that is for Teradata and we will manage and run that for our customers or they can utilize our code base to really look at how they optimize the harness inside their environment and so by having that kind of open source kind of layer inside our product stack, we think that will open up the opportunity for lots and lots of customers to ultimately use the Teradata platform as their data platform as that integration that harness is able to integrate through all layers of the stack yeah i guess we're still very early on it every um all the frontier labs data talk about harnessly as well like what have you seen from your customer that uses the teradata harness i think they like the fact that it's uh so open and also the fact that it's focused on data um you know we're not trying to be um uh to use language models or a harness to develop applications where we are looking at Terra as the cloud for data.

Okay. Is that going to help you drive like faster consumption?

More focused consumption straight through to the data platform. And also as organizations implement agents, they know that they can use the Terra agents as their data experts inside their entire infrastructure. Right.

So we have a last minute here. Maybe just talk about The autonomous platform helped you go into a different frontier. If you're thinking out a year, two years out from here, what do you believe is the next major phase of innovation that Teradata would be part of?

Yeah, I think in today's world, it's difficult to forecast more than three months out. But I think the key point there is, what we've seen in Teradata over the last 14 months is a tremendous amount of innovation that's allowed us to reposition the technology proposition that we've got and reposition the company to accelerate growth as we move forward. And that's certainly our ambition as we look to 2027 and beyond.

Yeah, great. I think there's going to be a big autonomous conference out in Dallas as well coming up.

Yeah, to work for a member. It's nearly sold out. We just had to move to a different venue. So I'd encourage anybody that's interested to come to come along. great thanks steve thanks everybody thank you thanks