I'm Siti Panigrahi, analyst, research analyst here at Mizuho. It's a great honor and privilege to host Tom Siebel, founder and CEO of C3 AI. Tom really needs no introduction. Those who know Tom was an early employee of Oracle. Tom, you wrote that wave of RDBMS. Next comes CRM. He pioneered, actually, CRM because that's when I was building his competitive product. I was telling him Oracle CRM. Then C3 AI. So Tom, I mean, you kind of identified the next big thing. When you think about being part of RDBMS wave, then CRM was a big thing in a big area. So on C3 AI, you started back in 2009. So what gives you that visibility and confidence that enterprise AI would be the next big thing? Well, it's true. When I studied relational database theory, did my graduate work there, there was no market for relational. There were no commercial relational database systems. We thought that had to be a big market. When we started Siebel Systems and thought about, well, we invented the CRM market, we thought that information technology and communication technology had to play a big role in sales, marketing, and customer service. I think that's a $120 billion market this year. So that turned out to be a lucky guess. Now, after we sold Siebel to Oracle in January 2006, we actually started this work in 2006. We thought we were about 50 of us. We worked in 2006, 2007, 2008. We started actually incorporated in January of 2009. We started thinking about what was next. We thought what next was about was elastic cloud computing. Understand, this is before AWS exists. This is before Azure. This is before Google. This is before the GPU. But what we thought next was about elastic cloud computing, big data, and the Internet of Things. And we thought that the convergence of those technologies would enable predictive analytics at enterprise scale, what we call enterprise AI. So we started in January 2009. So we are the first native AI company, hard stop, enterprise AI company. And we started work in January 2009. We spent $2 billion building a software stack that would enable us to design, develop, provision, operate at a massive scale predictive analytics enterprise AI applications. We've used that stack to build 90 turnkey enterprise applications today that address the value chains of health care, financial services, defense, intelligence, utilities, oil and gas, manufacturing. The problems that we solve sound kind of boring, but they're extraordinarily high value, like AI predictive maintenance, stochastic optimization of supply chain, demand forecasting, customer churn, fraud. And so fast. So we've been talking about enterprise AI in 2012, 2013, 2014, 2015, 2016, 2017, 2018, 2019, 2020, and nobody else cared. And then you get to November of 2022, and nobody could talk about anything but enterprise AI. So I think today it's not an overstatement to say there is no CEO this day who is not having a meeting, no military leader, no government leader, who is not having, or certainly no investor, who is not having a meeting today about what does AI mean, how do I use it, how big a threat is it. And I'm not certain there's many technologies that we could ever say that about. I don't think anybody thinks that is an overstatement, and it isn't. So as we power into June 2024, we have 90 enterprise applications. This promises the AI enterprise application software market promises to be north of $1 trillion market. When you incorporate generative AI, it might be a $2 trillion software market. So this is the largest market we've seen in the history of enterprise application software. The game that we're playing at C3 AI is to see if we can establish a market leadership position globally in that space. I realize that sounds a little aggressive. But we said that at Oracle in relational database in 1983. And I assure you, we did establish a global leadership position in relational database at Oracle. When we were nobody at Siebel, we set out to set up a market leadership position in CRM. When we sold Siebel Systems to Oracle in January of 2006, we had 85% market share in sales, marketing, and customer service. So the game we're playing is just to see if we establish a leadership position. This is a huge addressable market. We might fail. We might end up being number two. We might be number three. And whatever that is, I can assure you. So don't think I can't tell you what we're going to do next quarter. And it doesn't matter. As far as I'm concerned, the lights could turn off on the market for 10 years. And I assure you, or five years. And I assure you, this stock is not trading at $2 billion or $3 billion or whatever it trades at today. And so that's the big picture, Siti. Yeah. Look, as you say, AI is not new. We heard about this machine learning, predictive analytics, all that. But like you said, 2022, do you think that's the inflection point of AI? And do you think now it's ready for prime type enterprise adoption? Absolutely. There's no technology that we're using today in these AI systems that was not invented before I was born. Guys, I'm old. This was done at Princeton in 1951. This is deep learning, machine learning, AGI, Alan Turing. I mean, this stuff goes where there's nothing new here. Neural networks, Princeton, 1951. AGI, 1948. This goes back before the semiconductor. So there's no technology that wasn't invented like 75 years ago. Now that we have this huge AI winner, now with these enabling technologies, we had to wait for the cloud, elastic cloud computing, infinite computing capacity at virtually no cost, the GPU. The enabling technologies kind of are fueling this market. I can assure you, it is no longer winter. It is full-on summer in the AI world. Yeah. We agree with you that all the investments going on now with semis and autonomous AI infrastructure, but you can monetize only at the application, like the top of the stack right now. So I mean, how do you see this playing out? NVIDIA is now seeing the hockey stick valuation. Microsoft on its way. Doesn't seem to be many companies have figured it out. But how do you think that's going to play out in the applications? Siti, this works out just the way the PC market worked out, the iPhone market, the mini computer market. At the early stages, I'm not saying NVIDIA isn't a great company because NVIDIA is a great company. Jensen went public about the same time Siebel Systems did. I've known him for decades. This guy, Satya at Microsoft, is nothing short of a genius. Hard stop. If we look at the early stages of these markets, the value stack is all in silicon and infrastructure. If we think to 1990 or the '80s, an IBM PC XT cost $7,300, $22,000 in today's dollars. On top of that, so that was all silicon. Then we had Intel. Intel was king with the 8088 processor and beyond that. So Intel was king. Is Intel still in business anymore? I'm not sure. Do they matter? No. But all the value of that PC on the desk, I mean, it costs $22,000 in today's dollars. You had infrastructure from Microsoft on top of it. And then you had a few hundred dollars in applications from VisiCalc and WordStar. That was the stack. Now look at this PC that is on your lap today. I mean, this thing costs a couple thousand bucks. It costs your company $200 a year in expenses. It's got $200 in infrastructure. The infrastructure in silicon has been completely commoditized. And any one of us business people might be running $8,000 or $10,000 worth of applications per year on these machines, SAP, Bloomberg, Salesforce, et cetera. So this works on this guy. I'm sorry, I won't grab yours. I'll grab mine. This guy came out in 2007. I mean, all the value was in the silicon and the infrastructure. Where's the value today? It's all in the applications. So as we look at this value stack play out, today it's in infrastructure and silicon. Do I think NVIDIA gets hurt? No. Do I think Microsoft gets hurt? No. I mean, this is a multi-trillion-dollar market. But we're going to see the applications is going to occupy 70% of that value chain, as it always does. And so we cooperate with the infrastructure guys. We cooperate with the AWSs, the Google Clouds, the Microsofts. We cooperate with the silicon guys, at least in the case of NVIDIA. And we're playing in the application stack. That's a good point. Now we are in 2024. What do you think has gone according to what your vision was back in for C3 AI? What needs to be done for C3 AI vision starts being realized? I think it played out exactly as we predicted. I mean, in all fairness, it played out exactly as we wrote it up. Now, I don't think it could have anticipated this big boost that it got from the introduction of ChatGPT in November 2022. But that really fueled everything. By the way, for those of you who are interested in language models, I have no financial interest in this. And everybody in this room is interested in language models. And if you're not interested in language models, then you should resign your job and go do something else. Now, I've been looking for a good book on this subject for the last year. And it's all drivel. Now, this guy, Stephen Wolfram, wrote a book. What is ChatGPT doing? And why does it work? Guys, this takes 45 minutes to read. You will know more about language models than anybody in investment banking. It is a real and Stephen Wolfram, if you know this guy, he is the real deal. I mean, he's a very famous physicist, computer scientist, Wolfram Alpha, Mathematica. But he explains very clearly how they work, what the really small class of relatively simple problems they solve, the kind of problems that they will never solve. And more importantly, he makes it very clear. It'll be very clear why. Those of us who develop these models at large commercial scale, even we don't know how they work. And nobody knows how they work. And so it'll be quite a realization. It'll be very clear that nobody has ever there is Voodoo here. This is a new domain for computing because we've never done Voodoo before. Everything was deterministic. It's not deterministic anymore. Great book. Read it. If you don't like it, send me an email. I'll give you your $14 back from Amazon. I printed 5,000 copies of this just to give it away so that the people we're dealing with understand what this is about. I mean, I don't mean to minimize the importance of these things. These guys are huge. But you'll really understand it. Sorry. Now, digging into C3 AI, help us understand the value that C3 AI platform brings to the enterprise. So why is it so difficult to scale enterprise AI adoption? And how are you helping in that journey? Wow. We're dealing with data aggregates that were unthinkable in previous generations. I mean, we're dealing with hundreds of PB of data that are arriving at 30 GHz cycles. Massive data sets, massive data integration, enormous problems with I mean, a lot of the problems you deal with Enterprise AI are kind of grubby. Queuing, ETL, access control, encryption in motion, encryption at rest. But for those things, you have data exfiltration. See Samsung for details. And big cyber problems. So these are essential problems that need to be solved at large enterprise scale. The issues associated with machine learning, supervised learning, unsupervised learning, reinforcement learning, TD learning, generative AI, they're pretty interesting. But to be able to apply those at the scale of a Shell or the United States Air Force, you need to solve some enormously difficult problems. Now, there are many companies out there that are doing highly customized implementations. This is not what we do. Kind of there's the Accenture of the world that do this. There's a software company in this space. They position themselves as a software company. They're just doing custom implementations. What we're building is production turnkey software products we sell again and again and can that optimize supply chain, identify customer churn. The value that we bring at Shell is $2 billion a year in economic benefit. $2 billion a year at Shell. At the United States Air Force, where we're doing predictive maintenance for aircraft, they have 5,000 aircraft. As a result of our system, where we've fused the data from all the weapons systems in the United States Air Force, F-15, F-16, F-18, F-35, KC-135, F-22, et cetera. A B-1 bomber alone has 42,000 sensors emitting data at 800 cycles. That's just one airframe, each airframe. So this, guys, is a lot of data. And we analyze these data. We identify system and subsystem failure before it happens. You can fix it before it breaks, net net. We're increasing aircraft availability on any given day at the scale of the United States Air Force by 25%. Holy moly, guys, 25% at the scale of the United States Air Force. This is a big deal. So that's quite a value proposition. Yeah. And with this GenAI, how do you think that use cases evolve? Do you see now it's more pilots that you're seeing? What are you seeing with the? I announced this in February of 28 days of February. 10,500 executives, senior executives, CEOs, CFOs, SVPs supply chain, 10,500 executives in 28 days in February contacted us wanting to know about generative AI solutions and how they could be applied to their business. 10,500. I think last quarter, I announced it approached 50,000. This quarter, it'll be 90,000. All those executives were for companies greater than $500 million in sales. So nobody's ever seen anything. I believe that I have 90,000 executives approach us from around the world. What are the use cases? First of all, I think the most important use case associated with generative AI is largely not acknowledged. It is that it fundamentally changes the nature of the human-computer interface model in these enterprise applications. Guys, we've done a pretty dreadful job of the human-computer interface model in enterprise applications. See Bloomberg Terminals for detail. So you all know what that looks like. I'm not saying Bloomberg isn't a great company. It's a great company. So is Oracle. So is Siebel. So is Salesforce. But all their products are basically unusable by virtually anybody. Even you guys can't use the Bloomberg Terminal. And you use it all day long. And you might be able to take advantage of 4% of the power that's there. Well, imagine you have a generative AI interface on it. What does that look like? It looks like a Mosaic browser. You all know what that is. It came out of the University of Illinois in 1993. You know it is the Google browser. They copied it. And so you have just a command line. And you ask any question. What's the correlation point in the stock price of IBM and solar flare-ups? You ask the question. It'll give it to you. What are the risk factors associated with the name of the company? It'll give you the answer. And there's no more you ask it in 131 languages. It gives you the answer in 131 languages. And there's no more control, shift, F3, and then stamp your foot on the table to figure out how to get the answer out of Bloomberg. And I'm not taking shots at Bloomberg. I think it's a wonderful product. It's just I'm incapable of figuring out how to use it. So I think that's the biggest application. Now, as it relates to AI, this is really important because it makes AI dramatically more accessible. Think about this predictive maintenance application that I'm doing for the United States Air Force. This is highly, highly technical stuff used by maintenance technicians. You put a generative AI place on it, as we have done. Anybody in the Department of Defense can, by the way, we enforce all the access, data access controls. But it can ask any question in plain English and get the answer right now about any weapon system or any weapon systems in the Department of Defense. I'm sorry, United States Air Force. What about readiness levels for F-35 Squadrons in Central Europe? Think processes rather than a minute. Shows the map. Tells you where the squadrons are and what the readiness levels are. I mean, right now. And so that's use cases. We have a large law firm that you'll be able to relate to this use case where they're based. Now this becomes a small language model. Think an enterprise language model where we train the language model on the corpus of EDGAR, SEC. So every S-1, every 10-K, every 10-Q, that's what we use to train the language model. Now they want to bring some company public. Whoever's coming public next, I don't know. You do. Maybe Databricks. I don't know who it is. They put in the financials, type in what the company does, hit the carriage return. 45 minutes later, it generates the first draft of the S-1. Guys, it used to take 10 associates two weeks to draft that first draft of the S-1. So those use cases are Baker Hughes. We've put it on top of Workday and come on, Bill McDermott, ServiceNow. And they have 68,000 employees. I believe that number could be wrong. But I think it's 68,000. They're roughly a $25 billion company last time I checked. They operate all over the world. And these are their HR systems. Anybody in any language, Farsi, Italian, Arabic, Hebrew, English, can ask any question about any HR issue. They get the answer immediately. What are my vacation days? What are my benefits? How do I enroll for this? But the applications that we're seeing are really neat, highly diverse. It's huge. Yeah. So it seems like now with the proliferation of use cases you're seeing with this. So let's dig into a little bit how you make money. So you license your platform or as you expand. How do you expand your opportunity within a customer? Well, usually our entry level is quite, let's say it's a Dow or a Shell or wherever it might be. I'll bring the application live for $500,000 at the scale of Dow in six months. $500,000 in 6 months, optimize the supply chain. Get your mind around that. That would be three years and $200 million for Accenture. We do it in six months for, say, predictive maintenance for polyethylene cracking units. six months, $500,000. And then they have new use cases, new production. And it grows and grows and grows and grows and grows. And I think a large customer for us would be doing probably order of $100 million. But it starts with a $500,000. We're bringing it live in six months. If you like it, keep it. Can I talk about cash flow? Yes. I was going to go to that. With this opportunity, you see definitely how you're investing and how should you think about your cash flow? OK. It's Siebel Systems. Some of you will recall I ran a cash positive profitable business just the day we shipped a production product. I forget what year that was, maybe 1995. And we got compensated by that for the market loved it. I think we had about a $52 billion market cap in 2000. That's when $52 billion was still a lot of money. So it's hard for you guys to relate to. Now, and then when we get into remember 2019, 2020, 2018, this is the days when Masa was king. And all the banking analysts are in your office saying, you can't spend money faster. It's about cash flow. It's about cash flow. It's about growth, growth, growth. And now the pendulum with positive interest rates, the pendulum has swung hard stop to the other area, which says, well, you need to be cash positive and profitable now. Well, you know I've been thinking about that. When I ran a cash positive profitable business at Siebel, maybe I was a chump. Maybe that was stupid. Let's think about my friend Marc Benioff. I know Mark pretty well at Salesforce. I know his business pretty well. I might have invented it. Anybody remember how long it took Salesforce to be profitable? That would be 25 years. That would be a quarter of a century. Anybody remember how long it took Jeff Bezos to make Amazon profitable? That would be 29 years. How'd that work out for his shareholders? Anybody check the market cap of Amazon today? How many of you recall how long it took Apple to be profitable? 21 years. How'd that work out for the shareholders of market cap? What's their market cap? $4.5 billion or something. Today went up again, 10% today. Oh, my God. So I do think that the pressure of the markets to run a cash positive profitable business in the immediate run is not really. I'm not sure how healthy it is. And when we raised $1 billion in four minutes because the market said you could do it then in December of 2020, and I said, we're going to invest in technology. We're going to invest in market share. Three Zoom calls in four minutes. I mean, it was like that. And I said, I'm going to raise $1 billion to invest in technology, invest in market share, invest in human capital, and invest in brand. And that's what we've been doing. Now, we haven't been profligate. I mean, I think last quarter, I think I disclosed I had something like $750 million cash. Last quarter, I think we were $20 million cash positive. But for market constraints of the investment community, and we have to respect investors because we serve investors. They're part of my but for the current where the pendulum is today, I'd take the $750 million I have. I'd invest it in market. I'd invest it in technology. I'd invest it in brand. And I'd go out and raise another $2 billion. But I'm not going to do that because I don't think the market will tolerate it right now. This is a multi-billion dollar market opportunity. I want to be the market leader in this space. Hard stop. I mean, I am a cash positive profitable guy. I have demonstrated that at Siebel Systems over quarter-after-quarter. We are building a business. We are a structurally profitable business. By that, I mean our cost of sales is substantially less than our gross margin. So we're structurally a profitable business. So I can make this profitable tomorrow. Is it tomorrow I can make it profitable? All I got to do is slice some headcount. Is it in the best interest of my shareholders to make this company profitable tomorrow? No freaking way. No way. So you're going to expect us to be investing in the market. I expect to be at this time to be cash positive next year. By the way, there's some analyst out there who has a math error of the various analysts that covers this. There's some analysts out there who has a math error in his or her spreadsheet that has this massively cash positive in the first quarter of this quarter. Let me help you out. Ain't going to happen. So get over that. We need to talk to that person. It's just inadvertent error, I'm certain. I mean, there's no way, no how it's just cash positive. It was cash positive last quarter. It's not cash positive this quarter. Right now at this time, I expect it to be cash positive next year. And we are focused on growth. What has happened to this company as I changed from subscription-based pricing to consumption-based pricing? Subscription-based pricing. I used to do deals $10 million, $20 million, $30 million, $40 million, $50 million at a time up front. Then we switched to when we wanted to scale the business globally, we switched to kind of the current pricing standard of the realm in cloud computing, which is consumption-based pricing. So as a result of that, consumption-based pricing, rather than pay $50 million up front, you pay $500,000 up front for your reading it live. And then you pay $0.50 per VCPU hour or something. Well, that's cash neutral to us over 10 quarters. Cash neutral. But it enables us to do a much larger number of transactions. The effect is your growth rate comes down, down, down, down, down while you make the transition. It hits zero. And then it goes up, up, up, up, up as you get more customers. As your number of customers increase, your revenue becomes more predictable. Your quarter to quarter operations become less slumpy. And we've made it through that transition. I think we had in the last decade growth rates of like 50%, 40%. Then it went down, down, down to zero as we made the transition. And then up to seven, 11, 18, 20. Last quarter, it was 20% top line growth, 41% of your growth in subscription revenue. And we've kind of returned to accelerating growth. That's great. Let me pause here. If there is any question, raise your hand. Hi. Just a quick question. I'm sorry. Who's talking? I'm over here. Hi. What's your name? Jonathan. How are you, Tom? Jonathan, what do you do? I like to look at guys like you guys. I visit you guys and share something. OK. Welcome. One of the things that I remember you speaking quite a bit about was kind of software in a box. I think you kind of mentioned it today in some ways. You're trying to simplify the process for everyone. Try to simplify what? Simplify kind of a whole AI solution, neural net solution for kind of customers. Yes. So since most customers now seem to be pretty comfortable moving down the path of kind of AI, obviously is much more generative versus kind of predictive, I'm kind of curious. Are you seeing kind of customers more comfortable doing an in-house solution? Or are they more comfortable using your Jupyter, your Ex Machina type platform? Everybody tries, Jonathan, once, twice, or three times to build this stack internally. By the way, we've seen this movie before. When we introduced relational database to market in the early 1980s, everybody was going to build their own. Who succeeded at building their own relational database? Tell me the company. That would be nobody. When we introduced later on ERPs, CRM systems, everybody was going to build their own. Who succeeded at that? Virtually nobody. Now, this is a stack. It took me $2 billion in 10 years to build this stack. And these were some pretty experienced guys, not like IT people. Come on. IT people manage Deloitte to spend $2 billion doing the upgrade of SAP. That's all they do. Or try to figure out how to get SQL sign-on to work. And the idea that they're going to build a stack to build enterprise AI, there's just no freaking way. At the same time, they delude themselves into thinking they could do it. And they try and fail, try and fail, and try and fail. After they tried and failed three or four times, CEO is not happy. Executive team is not happy. CIO is put over to the sidelines. And they call us. And we get the job done. But I think everybody tries to build it once, twice, three times themselves. Every one of my customers did. And they almost have to go through that journey to be ready for us. Great. I have a question. I know you can hear me, but. I can hear you. I'll repeat the question. What's your name? Jenine Presidio. So I work for Presidio. Jenine. Thank you. Jenine. Hi. Hi. Nice to meet you. Very good presentation. Thank you. Cybersecurity. Yes. Can you speak about that? It's huge. It's really important. Existential. And how much of a threat is introduced by AI? A lot. It increases the surface area available for attack. Generative AI opens enormous portals for cyber attack. Being well documented now by Zico Kolter and others at Carnegie Mellon. I mean, really well documented. Generative AI has enormous cyber problems, data exfiltration problems. So we cannot pay enough attention to this. I would say the level of care that many organizations have, I mean, simple prophylactics like changing your password, keeping your operating system current. People don't do this. And they subject themselves to simple kind of best practices will probably solve more than 95% of the problems. But there are bad actors out there, aside from just the hackers. I mean, we have Russians. We have Chinese. We have North Koreans. We have Iranians. I mean, these people are competent. Pure evil does exist, people. And I mean, they could turn off the power grid from a cell phone in St. Petersburg any minute. Any minute. They could shut down the global banking network from Beijing any minute. And everybody knows it. And everybody knows it. So this is like really scary stuff. There's a book out there called The Perfect Weapon by David Sanger. It is a great book. It's a history of every state-sanctioned cyber attack. SolarWinds, the Chinese OPM, Office of Personnel Management Attack, where they got, I think, 21 million records of everyone who's ever been considered or granted a security clearance in the United States. The Chinese just walked off with this. The North Korean hack of Sony. They're all well described. David Sanger, The Perfect Weapon. Read it. Thank you. Great. With that, Tom, this is a great discussion. Thank you so much. Thank you, Siti. Thank you.