C3.ai, Inc. (AI) Earnings Call Transcript & Summary

Hello, everyone. I appreciate you joining us today. Apologies, we're a little bit late. I'm Patrick Colville, a senior analyst at DB covering infrastructure software. The format of this session will be a fireside chat with a listener Q&A. There's a chat box where you can ask questions. Any questions you ask today will be anonymous, so we're not going to mention your name or company affiliation. Let's kick this off with introductions. We've got Tom Siebel, the CEO and founder of C3.ai here with us today. As I'm sure you all know, C3 AI is a pioneer in artificial intelligence in the enterprise. Tom, thank you so much for joining us.

Good morning.

Good morning. Can I just -- I guess, just jump straight in and probably the obvious place to start would be just to help us frame adoption of artificial intelligence in the enterprise and in the federal vertical and then how C3 plays into this adoption.

Well, we're seeing that this is starting to be a pretty big business. I see in your report that your research suggests that all that's going on in AI is science projects and -- from your analyst reports. And you might want to check with your CIO, Bernd Leukert, at Deutsche Bank about some of the projects that they're doing or -- and I think, by and large, actually, most of the AI projects going on in the world, in fact, are science projects. I think I'll agree with you on that. When people are trying to build these science projects, Bespoke from Databricks and Snowflake and DataRobot and Azure Primitives and what have you, those are science projects. Some of which can get in the hundreds of millions and billions of dollars. And I would say virtually, every one of our customers will have done some years, 3, 4, 5 years, building these science projects before they threw in the towel. Now when we look at -- I think that if you look at -- we do some research on the C3 customers and talk to the people, for example, at in -- Now, okay? So at Now, they're using -- doing smart grid analytics on a very large power grid infrastructure in Europe, where they've aggregated 100 trillion rows of data, okay, into a unified federated image from, I think, 47 million smart sensors, 18 instances of SAP, multiple instances of Salesforce, Siebel, Dynamo, 2 different SCADA systems. They -- some of these data are arriving from sensors, 42 million smart meters. Some of these data are arriving at, I think, 90-hertz cycles. They are aggregating data from the extranet, weather terrain and social media at the rate of 62 billion times a day. There's 100 trillion rows of data aggregated into a unified federated image that they're using for -- to apply AI to optimize the power grid infrastructure. This is predictive maintenance, this is fraud detection, distributed energy and resource management, energy efficiency, what have you. So this, I believe, is the largest AI application in nonclassified space on Earth in production, and it's been in production for some time. This is way, way, way beyond a science project. If we look at other C3 implementations like Royal Dutch Shell, they have a project called C3 AI. It's very easy to research. Just look it up on the web, okay, where they're applying C3 on top of Google -- excuse me, on top of Azure to apply AI -- upstream, downstream, midstream, across all of Shell. This is AI-based predictive maintenance for -- of offshore oil rigs in production, production optimization for LNG operations in production, okay, predictive maintenance, descriptive maintenance for valves, 150,000 valves. This involves 2 million machine learning models in production. So these are applications that are in production across at Shell. AI-based predictive maintenance, United States Air Force, F-15, F-16, F-18, F-35 Joint Strike Fighter. So with all due respect, these are very, very large projects involving hundreds of people that are in production that are well beyond science projects. And we're seeing an increasing number of these types of projects. I think the economic benefit that they're looking for at Enel, if we look at their annual report from Francesco Starace from these -- what you refer to as science projects, is EUR 5.1 billion a year in recurring economic benefit. I know there are places where that's still considered to be significantly nonzero...

Yes. I appreciate it. And look...

But I think by and large, in AI applications, writ large, probably 95 out of 100 of them are science projects where their CIO is playing with tinker toys. But I think you'll find that the C3 projects are quite different from that.

Yes. And we've actually also spoken to Shell and the U.S. Air Force and a number of your other customers and the success you're having there is, I guess, very clear. I mean what differentiates, I guess, a customer like that, where there's a clear fit and then versus others who like aren't there yet? Is it the maturation of their IT stack? Is it kind of leadership? Like how do you get, I don't know, like Aramco or Rosneft to think like Shell?

We are working with -- Aramco is thinking like Shell, Rosneft is thinking like Shell, and we're there today, okay? Now the -- now I think that -- what's going on with enterprise AI is similar to what we saw with the adoption of many computers, personal computers, relational database -- no, a little bit of better examples would be relational database technology or enterprise application software like CRM. When Siebel introduced CRM to the market in 1993, every IT director or CIO was going to build a CRM system themselves, right? So they'd spend some years trying to build it themselves. At IBM, they tried to build it 3 times. And Microsoft, they tried to build it 3 times. And then they'd throw in the towel and they bought it from Siebel. And so Siebel became -- but it's just what people do, they try to build themselves -- these things themselves. Today, CRM is what, a $60 billion global business growing at a pretty good rate. So all these companies have to go through their science projects before they adopted relational database, before they adopted -- everybody had their science projects to build their ERP systems. How many people succeeded at that? People have their science projects to build relational database systems. How many people have succeeded at that? So they try and fail a few times, and then they bought it from Oracle. So this is the same trend. Virtually every one of our customers, Rosneft, Shell, Coke, Enel and now United States Air Force, will have tried to build this 1 or 2 or 3 or 4 times themselves with science projects. The science projects fail and then they purchase it from a commercial off-the-shelf vendor. No news.

Sorry, forgive me. I didn't actually realize that Rosneft and Aramco were customers already. Is that...

I didn't say Aramco is a customer. You said we had -- okay, and actually, I'm not prepared to announce that Rosneft is a customer, but you said when will a Rosneft and Aramco will be doing that, and I can assure you that they are doing it now.

Yes. Appreciate the...

They may or may not be customers, but I'm not prepared to talk about that right now.

Yes. And so I guess, I mean, switching gear to the, I guess, the kind of revenue growth acceleration we've been seeing. So the kind of trend for the last couple of quarters is a really nice revenue growth acceleration coming out of COVID. How has the demand environment changed in the past 6 months or so kind of versus a year ago? And I guess, what do you expect to see for the rest of the fiscal year?

Well, I think the data suggests right now and I know it's most of the analysts, if I'm not mistaken, I have this in front of me, if I could find it, okay, let's say, right now, the revenue grew last quarter, what, 29% year-over-year. I mean, where would that be in mid-cap software companies? In the top quartile?

Absolutely.

Top decile?

Yes.

I mean, you're an expert at this. I'm not. I just run a software company. I think the -- in the COVID year, if I'm not mistaken, you can correct my remarks. You know these numbers better than I. I think last year, revenue grew in the COVID year like 17%, and most of the analysts -- I know that you don't publish, you just kind of copy -- you don't publish estimates. You just got to comment after the fact. But the show that -- which is that -- it just makes the job a little easier. But the -- most of I think the analysts' consensus is for the company is like 35% this year, revenue growth?

Exactly.

Where would that be for enterprise software companies? Would that be at the top decile?

Yes, definitely top quartile, but probably top decile.

Yes, so what you know...

And a lot faster than Deutsche Bank, put it that way.

Is it going faster than Deutsche Bank? So it's -- so right now, we -- revenue has been -- top line revenue growth has been accelerating. I think the -- and we're optimistic about the business going forward. But right now, the analyst expectations are 35%. I think that where are -- you can tell me, Patrick where -- what we guided to at -- what is our -- what did we guide to? Like 32% or something like that?

Exactly, yes.

Yes. So again, so we're in the top decile of, I believe, of enterprise software companies, and it's -- we're optimistic about the market opportunity going forward.

Okay. Yes. And I guess, I mean, a key partner both historically and going forward, has been Baker Hughes, and C3 has got an amazingly close partnership with them. And my understanding is Baker Hughes use C3 software, both internally in their environment and resells the software. Can you just help us, I guess, understand that partnership and just kind of -- just like help us understand how it's kind of going at the moment?

Yes, Baker Hughes, we entered into a strategic partnership with them some years ago. They do use the application for some internal applications like inventory optimization and what have you. And we go to market with them globally. So we have go-to-market motion with them at virtually every major -- order of 12,000 people working with us at -- in all divisions of Baker Hughes, and we have sales, marketing activities going on in virtually every oil and gas company in the world. So -- and they are in various stages -- early stages, running pilots in some cases, running production applications in places like Shell, LyondellBasell and Coke. But that's a strategic partnership, and it's extraordinarily positive. As you know, we recently announced a very significant partnership with Google Cloud with Thomas Kurian. So now beginning -- I figure, what the date it was in September, but beginning some day last month, we now have order of 4,000 Google Cloud sales people selling our products worldwide in Asia, in North America, in South America, in EMEA, in telco, in banking, in energy, in utilities, in manufacturing. And that's a very significant partnership and a -- and increases the reach of our marketing organization by 1 to 2 orders of magnitude. And so that's a big deal. The other comment I would make, Patrick, because I know that you're a stickler for accuracy is you might make a -- I saw in your analyst report, you're making a comment for lengthening sales cycles. I mean, we have published the facts of the sales cycles, okay? And the cycle of sale here have been continually decreasing, okay, over the last 5 years, right, from like 17 months down to -- I think last quarter, it was something like average sales cycle, 4.5 months, and it's decreased every month. So this is why next time we publish, we might want to take a look at that. And if you need -- if you want us to like help you with the data. But our sales cycles are -- pipeline is getting longer, sales cycles are getting shorter, ecosystem is getting larger: FIS, Google, Microsoft, Baker Hughes. And a large -- a number of large production applications is getting -- applications into production is getting -- growing quite rapidly. I think we published it, I think it's today, we have 101 applications in production. A number of customers is growing rapidly. It -- if you look at the structure of the business and you noted this, that we have a structurally profitable business, our cost of goods sold is you're going to be able to tell me exactly what it is, but I think it's -- excuse me, our margin -- our gross operating margin is like 75%, 78%, it's really quite significant. So I've made the statement that we can throw this into profitability whenever we want. I mean it's true. I mean if we cut back on marketing expenses, cut back on headcount growth, cut back on facilities growth, we would be -- it would be a non-GAAP profitable business. But I think that it's not in the best interest of the company right now. It's in the best interest of the company right now to invest in growth, to invest in market share and invest in brand. And so we're making a decision to do this. But our gross margins are quite substantial. So it is structurally profitable, unlike many companies that you guys cover where their cost of sales are greater than their revenue, right? So I mean it's hard to make that. And I'm not suggesting they aren't good companies. They might be great companies, but we have a structurally profitable business. And I think aspirationally, what are we going about here, we're going to see if we can establish and maintain a market leadership position globally in enterprise AI. Aspirationally, okay, we hope to run a healthy -- we're in a steady state, okay, I think we have the opportunity to run a healthy, profitable cash-positive business. And if we do not become the leader in enterprise AI, I think it's unquestionably that -- unquestionable that we'll be a leader, okay, in enterprise AI application software. And I think the market is developing about on schedule, and the expectations are that this is like 1/3 of $1 trillion enterprise application software market in like 5 years. So that's the nature of the opportunity. What differentiates us from all the science experiments out there that you referred to is, the science experiments are where companies need to select 20, 30, 40, 50 different projects -- products and try to stitch them together into something that solves a problem, say, like anti-money laundering for Bank of America -- or anti-money laundering for Deutsche Bank. Let's start close to home, okay? And those are -- that's what you would call a science project, and they never work. We deliver a turnkey tried and tested and proven application for anti-money laundering that includes all of the services necessary and sufficient to solve the problem: data integration, ontology, the ability to map data provenance and lineage, connect this into the enterprise, machine learning services, access control, encryption, motion encryption, rest queuing, application development tool. So we deliver the entire platform so companies don't have to do the science experiment. So that your -- so the [ patient ] doesn't need to fund one of those.

Yes. Thanks for that. And I get -- I mean you touched on a number of interesting areas. I mean you touched on, I guess, sales cycles, which is something we want to talk about. But first, the GCP partnership because I guess for me, that was probably one of the biggest takeaways of the earnings last week was the announcement of the GCP partnership. Can you just help us explain -- I guess what makes you so excited about the GCP partnership? You touched on the number of salespeople, but just like kind of to better understand it slightly. And then also, people like me care about the kind of numbers are right. And so when will this partnership contribute to C3's kind of financial results?

Okay. Thomas Kurian kind of put out a statement about 2 years ago that said that he was going to approach the hyperscale market in a completely different way. So rather than sell CPU seconds and lead with CPU seconds and storage hours at maybe a cheaper price and maybe they have an argument that their technology is superior, maybe it is superior. I'm not going to argue that, but it's entirely possible. Certainly, a lot of people think that. But he said, rather than do that, we're going to approach the market through applications, okay? And so the net effect of applications like anti-money laundering, like fraud detection, like predictive maintenance for aircraft, what have you, is that, of course, you're burning a lot of CPU seconds and consuming a lot of cloud capacity. And so when you think about it, if he -- so he wanted in his -- he's grown his sales organization from order of 400 people to order of 4,000 people in the last couple of years. I think the actual number is like 350 to 3,800 or something like that. But so just -- these are rough numbers but really substantial. Now he wanted a partner where he could arm his sales people with a complete set of turnkey AI applications that exploit all of the utilities of the Google Cloud. And he wants to do it in manufacturing, he wants to do it in energy, he wants to do it in financial services, telco, aerospace, et cetera. Now I ask you, so how many doors are there that he can knock on, on the world -- in the world where he could find 40, okay, production applications that address this opportunity? There is one door in the world he could knock on, and he knocked on that door. And so Thomas and I put together a relationship where we are jointly -- we are tightly -- we have, okay, tightly integrated the entire C3 stack with the GCP Elastic Cloud and AI services like Vertex, like Bigtable, like BigQuery, and that we are jointly -- we have go-to-market motion going on all around the world, coordinated by Rob Enslin, his head of sales; and Sam Alkharrat, our head of sales, where we're organizing around accounts, we're organizing around the markets, and we're in joint sales motion today. So I mean, you will certainly see -- I mean, I expect we'll close some business this quarter, I would think. And I would expect we'll certainly see revenue from this, this year. And I would expect next year, it could be pretty significant. But I think it's really quite a statement, Patrick, if you think how many doors could you knock on to find a full suite of enterprise AI applications, turnkey, ready to go, there's no place in the world that he could find that other than C3 AI. And so that's a -- I mean I think that's something of a statement.

Yes. I mean, it's really exciting. I mean there's no doubt about it. And I know Deutsche has publicly disclosed that GCP is Deutsche's cloud, and I'm sure that's something that the team internally is probably definitely very excited about.

And you might want to check with Bernd Leukert, he'll get -- who's your CIO, about whether he's getting ready to engage in science experiments or large-scale enterprise AI applications at Deutsche Bank or check with Bernd or Christian, and I think you'll find that they're a little bit beyond science experiments in terms of the way they're thinking about it.

Yes. Can I switch gears to Ex Machina? And I guess the reason I asked is that you touched on earlier that sales cycles have shortened. And I think that's probably been one of the biggest trends probably since the IPO is the ability for C3 to drastically shorten the sales cycle. Ex Machina clearly will kind of help in that regard. How are we in terms of adoption of Ex Machina? Is it still kind of very early innings, just getting started? Or are we slightly further along? And I guess -- is it a different cohort of customers using it? Is it not the very large enterprises? Is it more kind of smaller and midsize type firms?

Great question. First of all, if we look at reduction in sales cycle, it goes on well before the IPO. So I think in 2000 -- in '19 -- this is from memory, so this is off a little bit, okay? But I think our average sales cycle in '19 was like 17 months, and then it went down to 15, and then it went 12, then it went down to 7, then it went to 5. And I believe last quarter, it was 4.5 months. So it's been consistently declining as we've restructured our sales organization to move from -- as you correctly noted, from kind of large, big gun, elephant hunters. And now we also have people who are compensated to be deer hunters and rabbit hunters, okay? And so we're going through enterprise sales and middle market sales, and we have products today in our marketplace and telesales where we're selling now to even our enterprise stack to relatively small businesses, whereas we used to only sell to Shell or Bank of America. And now we sell to companies like Morsco that are relatively small businesses. Let's talk about Ex Machina that you referred to as a nascent market, okay, for -- I don't think that's true. In other words, how long has Alteryx been in business?

Sorry. I didn't say nascent. Or if I said it, I missed -- I meant a nascent product. I think it only went in GA in January.

You're talking about a nascent market. I don't think it is a nascent market for basically citizen data scientists and the people for analysts who want to do data science. Alteryx has done a pretty good job. You also have in their activity -- there's a couple of other players in there, maybe DataRobot and one other. Now I think that, though, the way that we're going to sell -- the way that our go-to-market motion with -- and Ex Machina is a mature product. I encourage you to download it from the web today. You can get it for free, okay? And you can use it for 30 days for free, do some analysis on how about the sales cycle for C3 over the last 5 years. That would be a good analysis to do, okay? And they are -- but really, download it and try it and send me a critique. And by the way, anybody on this call, download and try it. Don't forget to leave your e-mail. I think you have to leave your e-mail because we'll be calling you in 30 days to try to get you to pay some money for it. The -- I'm sure my guys will do that. And the -- but I think the real opportunity for it is not to sell one at a time or 10 at a time, which is the way you would have done it in the old days like Alteryx. There's a real desire virtually -- I've been in discussion with one company that wants to bring 1,000 data scientists live in their F&A -- citizen data scientists live in their F&A organization, okay, in 90 days. How many CIOs don't have a desire to bring a 100 citizen data scientists live? Or 50 data scientists live? Or in the case of this organization, like 1,000 data scientists live? So I think you're going to find our primary -- while you can download it from the web, you can use it for 30 days. If you want to keep it, I think you pay $4.95 a month forward or something, okay? And I think you'll find it's a pretty good -- better than a pretty good product, okay? And it's quite mature for being able to drag in CSV files and open source data and do point-click, drag-drop, serious data science for non-data scientists, it is great. It's a fine product. But I think the real opportunity for us, because what -- if there's anything we're good at, it's large enterprise sales, is serving the CIO at Deutsche Bank, at Bank of America, at Coke, at Shell, at General Motors, who needs 100 to 1,000 citizen data scientists. So we'll be -- if somebody wants to bring 1,000 citizen data scientists live, we'll do it for free, okay? If you use it for free for 60 or 90 days, we'll train your data scientists in data science, we'll train your data scientists on the use of the application. We'll help them build their -- we'll provide white glove customer service to these organizations. And I think a non-zero number of these trials will convert to paid licenses. So that's how we're doing it. I think it's going to be a -- it hasn't -- does it contribute to revenue? It does. Is it a major contributor to revenue? It is not, okay? Will it be, okay, in the out years? I'm certain it will be a nice business. The other thing that I'd point your attention to and take a look at what we're doing to AI-enabled CRM. I mean what we've done there is really quite remarkable with C3 AI CRM. I mean, CRM is a $120 billion business this year, right, $120 billion, $60 billion in software, $60 billion in services. Accenture alone does $3 billion in services. $120 billion this year alone. So we think about it, what do we need -- this is Salesforce, this is Siebel, this is Dynamics, this is Veeva. So if it's $120 billion this year alone, you tell me, what's the size of the installed base? This has to be maybe $0.5 trillion or $1 trillion installed base out there. Well, the way that C3 AI CRM works is it sits on top of that installed base, on top of Salesforce, on top of Dynamics, on top of a Siebel, on top of Veeva. And you install it and your investment you have, which is in some cases, many hundreds of millions of dollars, some people pay Deloitte -- I think Salesforce alone pays to Deloitte $100 million a year just to maintain their implementation, right? And so now you can -- your Salesforce system, your Siebel System, your Dynamics system, it's completely AI-enabled. AI-enabled revenue forecasting, product forecasting, next best product, next best offer, customer churn. Very, very precise revenue forecasting and just amazing data visualization. We fundamentally changed the user experience, a paradigm for CRM. I know something about CRM, Patrick. I invented it, okay? And I did. I did, okay? And the -- and now we're reinventing it. But we're not competing against all the installed products in the world. We're just -- we're cooperating with them and enhancing them because everybody wants their CRM applications to be predictive. And as we know, these revenue forecasts that come from our sales organizations are replete with human error, okay, and highly inaccurate, and we -- the first installation, when we install it, we reduce the average forecast error from 18% to 3%. I mean this is -- I mean this is really, really exciting. So I think there's a big market opportunity there. And it's a very -- this is -- CRM is not a nascent market, okay? And I -- so I finally found one that we're in that you won't -- I'm confident we both agree is not nascent. Again, and the -- I think this is going to go off like a bomb.

Yes. It's -- I think we're running out of time, but I really want to squeeze one more in because we discussed a lot of the kind of macro trends, a lot of these major changes happening in software digitalization, AI. Can I just talk about -- one question was more about the minutiae. On the earnings call last week, you mentioned some deals slipping out of 1Q. Is there any color you can give us on, I guess, why those deals might have slipped? And then have those deals that you're referring to closed today?

I don't think I referred to anything that were closed today, okay? A matter of fact, I'm quite confident that I did not, okay? The -- have some closed after the -- since the first quarter? Yes. Have I referred to any specific ones? I'm going to assure you that I have not. Okay. The -- one needs to be careful about these things in 2021. The -- where am I going? Oh, I mean, our deal size -- well, our deal size has gone -- average transaction value has gone from $16 million down to $4.5 million. $4.5 million is still large, Patrick. I mean this is probably an order of magnitude larger than any other software company you and I know, okay, with the single exception of Palantir. But that's hard to tell because they're clearly conflating products and services into the same line. So it's hard to tell what's really going on there. Here, we're talking about software sales. But $4.5 million is still quite large. And I have put into place the -- and as you and I communicated and I communicated with many of the investors here, before we went public, that we had a plan to -- a concerted plan to restructure our go-to-market motion and bring down our average contract value through lower-priced products, through different distribution model and it's working. We've gone from $16 million average to like $4.5 million now, and it will continue to get smaller going forward. That being said, $4.5 million average contract value is still pretty big, probably an order of magnitude larger than any other software company you and I know. And the fact is, some deals moved out that we were expecting to close, okay, in the last month of the quarter, moved into the next quarter. And it happens. It's life in the software business. The good news is when you have a recurring revenue model, the way that we do, it really doesn't have much impact on revenue and the -- because the -- because what we're building is a revenue flywheel. But yes, some deals moved out. And when we get to this quarter, it was -- I have not yet taken all the lumpiness out of this business. I think it's probably going to take -- you can expect this business to continue to be -- the billings line to be lumpy for probably the next 7 to 8 quarters until we get it all smoothed out. But I think we're right on plan just as we communicated, and I'm very pleased with the progress that we've made to date.

Yes, I appreciate that. Well, Tom Siebel, thank you so much. I mean, we've discussed how customers are not using C3 for science projects. They're using it for transformational projects. We discussed how the pipeline is getting longer but sales cycles are getting shorter. We discussed the GCP partnerships. It's been really fantastic. So thank you for your time. I really appreciate it, Tom.

Thanks for your time. I really enjoyed it. Nice to talk with you.

Cheers. Bye-bye.