《No Priors Live》：SaaS“看空论”是否被夸大？MongoDB CEO 回应

Since 2022, the future of software has been in question. This is from the investor community, but also from customers. It is a very pivotal moment on the software stack. Then you look at the software stack and say, “Okay, what is the one thing that will always be there?”

How many companies today are there that have more than $10 billion in pure-play software revenue? It’s single digits. Why is that? The software industry has been around for a long time, created by many, many smart people like yourself. Why is it that only single-digit companies have more than $10 billion in revenue?

Because

Platforms are rare.

Platforms are rare. Speed matters. When technology transitions happen, are you building as fast as you can? Then are you learning on that technology shift? Whether it’s the internet age, the AI age, or mobile, are you pivoting fast?

You just have to stay ahead of that game. If you fall behind, investors or customers will always ask you that question: What is the future of your company?

Welcome to the very first live recording of the No Priors podcast with host Sarah Guo and MongoDB president and CEO CJ Desai.

Hey, everyone. I am so happy to be here with my longtime friend CJ. I know you guys have had a great day of announcements and learnings here at the conference, but I’m really excited personally to have the opportunity to zoom out with CJ to talk about the future of software, what’s happening in SaaS, and where the value is going to be.

These are important questions to me in my day job as a venture investor. CJ, you have worked at platform, enterprise software, and infrastructure companies. You became CEO of MongoDB recently. I feel like the one question that we were just talking about, that every investor asks you and then everybody in the technology ecosystem has in the back of their mind, is: What is the value of software when you can generate a bunch of software? I’d love to just get your thoughts on this.

That’s a very spicy question to start with. I like it.

I’m making sure everybody’s awake.

First, thank you for doing No Priors Live for the first time. It’s amazing, and we have a really good crowd here. It’s always exciting.

When you think about technology transitions in software, whether you look at the internet age or the mainframe all the way through AI, you have to really think through: What is the moat here? Whichever applications you create—you know, SaaS applications got created in the late ’90s. I think Salesforce had its 25th anniversary recently, so SaaS has been around for at least 25 years. From a transition perspective, and now with AI, the question is, in general, what is the future of software? What’s the stack? Do you really have a moat as a company or not?

There are folks who will say, “Hey, my moat is that I have a great customer relationship,” or, “My channel is amazing, and that’s my moat as I disrupt myself within.” But from my standpoint, speed matters. When technology transitions happen, are you building as fast as you can, and are you learning on that technology shift? Whether it’s the internet age, the AI age, or mobile back in the early 2010s, when Meta made the pivot toward mobile, are you pivoting fast?

If you pivot fast to leverage the technology, whatever the platform shifts are, I think it is fine. You just have to stay ahead of that game. If you fall behind, investors or customers will always ask you that question: What is the future of your company? That is something where you have to be on the leading edge. Not every bet will work, but from my standpoint, taking the extreme view that terminal value will be zero for some software companies is overblown. We’ll figure this out together.

Part of your career was leading product at ServiceNow. It is a franchise. It is one of the most durable enterprise software companies—or so everyone assumed until relatively recently, and now this question is up for debate. I think for a lot of people with an engineering mindset who think about buying developer tools or using developer infrastructure, the words “customer stickiness” or “distribution as the moat” feel very abstract.

Can you talk a little bit about how important ServiceNow, as an example, is to its customers and what you think about that?

One of the things is that platforms are sticky; products are not.

Okay.

No matter which software company you create today, in the age of AI or in the past, products can be replaced. My hiring manager at ServiceNow, Frank Slootman, used to say, “Tools are for fools.” If you say, “This is a software tool,” that’s never a good sign. That’s why he used to say that tools are for fools.

Products can be replaced because software is a fast, disruptive market. You want to make sure that you are positioning yourself to the customer as a platform, whether it’s a builder here in San Francisco creating a brand-new company for a use case, all the way to a very large company that is selling to, say, a large bank.

When you position yourself as a platform, maybe that increases your sales cycle versus saying, “Hey, you’re using this. Now you can use this. Go ahead and replace it.” Platforms are sticky because it’s a thoughtful decision from a customer’s perspective.

This is actually not the general advice of many people in the startup and VC community, so I want to talk about this a little bit, and then I want to get to your second point. A lot of people talk about having a wedge, and for ServiceNow, a service desk would have been considered the wedge. Is that not the right—is that a wrong retelling of history here?

You need an initial use case, and that initial use case has to be a killer use case. If you go in front of a large bank, a healthcare company, or a manufacturing company and say you are building something here, you can say, “Okay, this is a disruptive way to think of a legal use case, a finance use case, or a help desk use case.” That’s great. That’s your entry point.

But if the entry that was easy for you to get in was disruptive, your exit would be similarly easy, because they have not built things around you, right? That’s the main point. Today, it might work for your $0 to $100 million, $0 to $10 million, or $10 million to $100 million—whatever steps you want to go in—but it gets harder and harder from $100 million to $1 billion, $1 billion to $5 billion, and then $10 billion-plus.

$10 billion in just pure-play software revenue.

How many companies are there? It’s single digits.

Okay.

Why is that? The software industry has been around for a long time, created by many, many smart people like yourselves. Why is it that only single-digit companies have more than $10 billion in revenue?

Because—

Platforms are rare.

Platforms are rare. So one is, your dream or aspiration as a software company should be that you become a platform. Once you are a platform, that means N equals at least 2: You have 2 or more products being used by your customers. Whatever you are offering, they all work in unison with each other, so it’s sticky truly from a technology perspective.

Then I would argue further that all the integrations your customer has to do with their existing systems also matter. Remember, if you go to a large bank, some banks have been around for 100-plus years. You go to a large insurance company, and they’ve been around for 100-plus years. You go to a healthcare company, and it’s the same thing. You look at the Fortune 10, Fortune 100, and Fortune 500—that’s where the TAM is.

Mhm.

If that’s where the TAM is, and you go there and it is just a product, eventually you’re going to max out, and then you’ll have to add multiple things. If you’re a platform, your products are sticky, the products work with each other, and then those products work with all the systems you have.

I’ll make it very specific. Speaking to a bank on behalf of MongoDB, they run their commercial banking applications on top of MongoDB. They have built a lot of other integrations. They have done all the security checks, governance, and all of that stuff. I said, “Wow, gee, how many applications have you built on MongoDB?” They said, “Very critical, and they matter.” I said, “How many?” Finally, the CTO tells me in London, “30.”

I said, “Wow, okay, that’s great. 300 applications built on MongoDB.” He said, “CJ, don’t worry. Thank you for coming to London. We are not going anywhere.” I said, “Can I just ask you what the denominator is? I understand the numerator is 300.” He said, “9,000.” I said, “9,000? That’s a great opportunity for MongoDB.”

He said, “I’m not going anywhere.”

I’m not going anywhere. That’s when the more they use us, the stickier we get, and then we are in the fabric of their infrastructure.

The other premise that some builders, buyers, and investors now have is that those 9,000 applications, with vibe coding being possible, or with engineering and some code generation, are just going to be made on demand or in niche ways by every company. What’s your take on this? Is that the way that this bank, or whoever it is as a customer, is going to get exactly what they want, and they’re not going to use horizontal, more standardized, or even vertical applications anymore?

If you’re trying to sell to banks, they have huge budgets for technology, right? So, okay, you used a vibe-coding platform A, B, and C, and you created an app. Great. Your app velocity has increased. If you used MongoDB, it accelerated—just kidding—but your app velocity is high. Got it.

But you still need that go-to-market channel. How are you going to approach the bank? What is your truly disruptive way? Then the bank will ask you, “Hey, we talk to regulators a lot more than we speak to our customers and vendors. Will this work? Will it pass our regulatory test? We need resiliency.”

“What do you mean? You’ve just built it in AWS. It doesn’t work in GCP. I need multicloud resiliency.”

“I really need this banking application on-premises, truly sandboxed—or, a better way to say it, in an air-gapped network.”

These are enterprise-class things that you need where the TAM is.

Mhm.

Right. And so that's what could hold you up: Is this truly an enterprise-class application that you can take to a healthcare company or a public-sector federal customer and do that? So yes, vibe coding will allow you to create an app fast. You have a great use case, and you have some disruption in mind. That's excellent. But then there are a lot of things that you need from a go-to-market perspective to be able to break in, pass all their checks, governance requirements, security audits, and things like that.

I want to talk about the decision to join and lead it in a minute, but having worked at incumbent platform companies like ServiceNow and Cloudflare, what would you do if you were them or any other large enterprise software vendor today? What do you think is the path to success 5 or 10 years from now that the investor community does not understand?

Wow. What would I do if I were there? So, recently at Cloudflare, I would say the TAM for these platforms still exists, and the TAM is still large. So that's a good thing, because if you feel like your TAM is decreasing or all of a sudden no longer relevant, that's an issue. But whether it's MongoDB or anybody here who is working on their company, I would say you really, really need to understand what that moat is, and you need to protect that moat or maybe strengthen that moat even more using AI—whatever that moat is.

If the moat is truly that you are the platform already integrated with 50 different systems in that large healthcare company, great. Why can't you now integrate with 100 more companies in there? Why can't you create additional products for additional use cases really fast using AI and continue to show, I want to say, reacceleration of growth—that AI is really helping us innovate more and sell more? Because if you say you're innovating more but you're not selling more, then you potentially have issues, no matter who you are. That's a generic comment, but can you innovate more, can you disrupt within, and can you sell more?

That's what, if I'm an investor—and we speak to investors all the time—I'm looking for. Would AI reaccelerate this company's growth? And unless you show reacceleration, they're going to say, "Okay, maybe I'm neutral," or, in some extreme examples, "I'm a bear."

When you were at ServiceNow, you got a lot of calls for different jobs. And at Cloudflare, you got a lot of calls for different jobs. I know because I called you about several of them.

There's actually been this big shift of dollars in the investor community. I'm really thinking about public markets, but also private markets, from business software to essentially AI infrastructure, the model layer, and maybe hyperscalers.

And hyperscalers. Yes.

The data and developer infrastructure layer has not been the focus of that dollar movement, and you could have done any of these things. How did you think about what sectors would be long-term relevant? It sounds like you're pretty committed to being a platform, so be a platform.

Yeah, it's absolutely true that I had choices, and choices are always hard. Cloudflare is a great company, and I could have stayed at Cloudflare.

From my perspective, the first thing I look for is: Is there a durable TAM? I started my career out of college with Oracle Corporation and understood how they scaled the database platform, truly created apps on top of it, and did a bunch of other things during Oracle's organic growth days before they started buying various assets.

I learned a lot about how they really thought about the database platform, then the middleware, then the application layer on top of it, and so on. So, one, with Cloudflare, having an understanding of the database market and its large TAM, that's great, and then MongoDB is also a large TAM. So that's the sequencing between Cloudflare and MongoDB.

Second, with MongoDB, when I truly talked to—I want to say—many, many customers, when I did my own diligence on MongoDB, one of the surprises I had was mission-critical apps. Whether it's an e-commerce app for a retailer, a commercial banking app, a healthcare app, or an insurance claims-processing app, these are very critical apps. In the database industry, Oracle will celebrate its 50th anniversary in a year and a half, so it has been around for a long time.

MongoDB has been a truly disruptive force since being created in 2007. So it's been only 18-ish years. So there's a large TAM that could be disrupted, whether you believe the database market has been around 50 years or 60 years. MongoDB is truly the only disruptive force.

And then the second thing I learned by speaking to some of the companies here in the San Francisco area was that some of the digital natives around the 2010–2015 timeframe, and some of the AI natives that I spoke to, were building on top of MongoDB. I'm like, wow, okay, so there is something.

When the founders created the company, they didn't know that it would be full of unstructured data, that you would want very high velocity, and that you would need to be able to search on this. That's what AI applications are like: The data is so messy. MongoDB is perfect for that.

So I said, okay, the cloud transition is going on; it's still going on. When I speak to Fortune 500 companies, I can tell you without hesitation that they are all still talking about, "I need to move X% of apps to GCP, AWS, Azure, some combination, Alibaba in Europe," whatever the case might be.

That's one. But, second, the AI transition has just started. If cloud started with AWS, we are approaching the 20th year, and it's still going on. AI will still go on, and this is a layer you must have. That's it. TAM, must-have layer, no risk of disruption.

When you think about the investor focus on the model layer versus the application layer, they're anxious about the SaaS applications. They're anxious about data infrastructure because it feels like the way you build applications is still evolving very quickly. Okay, right?

And tell me if you disagree with any of these assumptions. What do you feel more confident about in terms of ways in which applications will be valuable in the future?

Just even speaking to people since 2022, I think that was probably a very pivotal moment, with ChatGPT in the fall. Since 2022, now we are 3 years plus into that journey. I have never seen this because it's been pretty static for a while. The future of software is definitely in question. This is from the investor community, but also customers that are asking, "Hey, should I use X or should I use Y?" So definitely, it is a very pivotal moment on the software stack.

Then you look at the software stack and say, okay, what is the one thing that will always be there? I mean, LLMs will be there for the foreseeable future when you are truly building AI applications that rely on that stack. You have also seen a lot of innovation. You look at xAI, which came from nowhere, kind of, and how well they are doing overall. But that stack will be there in the agentic software framework.

The data layer has to be there because you need to store data somewhere, so the data layer has to be there. That's the second one. Everything that is around that is going to evolve, and you better show true value on whether you use the platform analogy or whatever.

Whether it's the top layer of the stack, where you really understand the use case for the insurance industry, and you are building an AI-native company for the insurance industry—the insurance industry has a multitude of use cases—you say, "Okay, please move from old SaaS X to new Y, and with this new Y, the speed to value is going to be this fast and we will always be ahead." Things that you thought were not possible with the old SaaS are now possible because of AI.

That use-case focus on the top layer, besides the LLM and the data layer, will still always be critical.

MongoDB has startup and individual developer customers all the way to the Fortune 10.

Yeah.

Almost all of them. What do you hear from the buyers and builders at the very largest companies in terms of their real perspective on AI value and what they're excited about or skeptical about now?

Yeah. I would say I feel like it's a total failure of a week if I don't speak to at least 10 customers a week. That requires a lot of prep and a lot of follow-up, but it's usually at least 10.

Okay.

I’m constantly getting these data points and trying to do pattern matching on what’s going on. The first thing I would say is that when you think about the Fortune 500, Global 2000, and the community there—some of them are here—it is still not moving very fast. A lot of them tried the office-productivity-type copilots, and it’s unclear how much value they got out of them.

They’re like, “Okay, does this really work with my Excel? Can I really do this thing or create a PowerPoint slide using natural language?” The feedback is not great on the value they got. The feedback on coding assistance, which took off in 2024 in a meaningful way, is very, very positive. 2024 and 20 started with GitHub Copilot, then a few others, all the way to Anthropic and so on.

So 2025 was a breakthrough year, from my perspective, for coding assistance, and that’s still going on. I get very positive feedback from customers: “Hey, I’m using this particular coding assistant, or this one, and it has improved the innovation velocity, security, or whatever I’m looking at.” People are still tinkering around with customer support. On customer support, they’re like, “Okay, truly, if I’m a large, say, telco, or I’m a large healthcare company, can that fully be done by this AI-native company that does customer support?” Not there yet, right? They are going after initial use cases, and that’s great in certain industries.

You mean the end-to-end customer experience?

End-to-end customer experience. The question that I get from these customers—they ask me my perspective on SaaS—is, “Should I think about this AI-native company in customer support as an ‘and’ or an ‘or’?”

Mhm.

So, I have a system of record with one of the companies for customer support—

We can say Salesforce.

Yeah. Yes, Salesforce. And if I have this disruptive company that comes and says they can solve these problems and handle these support cases, should I think about that as an “and” or an “or”? I ask them the same questions. We use these systems-of-record SaaS systems of record, and when somebody comes and says, “Are you a layer on top of the system of record, or can you replace the whole thing?” you will have my attention as a leader if you say, “I’m going to replace it. It’s going to be cheaper, faster, and better. And, oh, by the way, I have disruptive pricing, and for the value you get, that’s when you pay me the money.” That’s a conversation I’ll have every single day.

That’s a very surprising attitude, actually. I think it is risk-taking, and the value that you’re asking for is much higher, but you’re showing openness to a lot of pain and being willing to ignore a bunch of the sunk cost of the implementation of the systems you already have. Having built systems of record yourself, do you think that’s feasible—that other people will do that too?

I mean, absolutely. Our CIO here is in the audience, Deepa, and she gets approached by AI-native companies daily, multiple times a day. When she comes and asks me how to think about that—whether it’s for go-to-market, sales, marketing, or whatever the case might be—I say this is how I think about it: If this allows us to hire fewer people and makes our people more efficient, then we will use that budget.

I want to be an AI-first organization on behalf of MongoDB, to say that we are transforming our business—not just making people productive. Being productive is okay, but we are transforming our business using AI.

I think that’s really interesting, because a lot of the conventional wisdom would be that the system of record is so embedded that you’re going to come in with a wedge or a layer on top. That is different from what you’re describing, which is, “I’m going to sell a platform, and I’m willing to buy a platform if the use case is good enough.”

That’s correct. But I actually think, going full circle, that’s a very interesting opportunity for MongoDB in particular. One of the reasons you might actually replace these systems of record in this age is that you want to keep much richer information about interactions or whatever else it is in your system of record. It’s messy. It’s got to go somewhere, right? It might not be Oracle anymore.

I was talking to a European retailer the day before yesterday at NRF in New York, and they said they tried a bunch of systems of record for ERP—expensive, failed implementations, lots and lots of issues on supply chain all the way to financials—and decided they were going to invest in just building it themselves. They are building that on MongoDB, and that’s a great use case. I’m like, “Okay, you had me at hello,” to do that.

But if these kinds of organizations are going to transform or disrupt within, I asked Deepa, our CIO, the same question: Are there things that we can build ourselves to disrupt within on MongoDB? That’s the story and the compelling value we can articulate to our customers as well.

I want to take the last couple of minutes that we have together to talk a little bit about leadership, especially as a product person. I think of you, first and of course as a CEO, but first and foremost as an extraordinary product person.

You do talk a lot more about business strategy, and have for the decade I’ve known you, than many product people. When did you start thinking about that as a product and engineering person? When you’re talking about defensibility moats, how people buy, and so on, that’s a lot of business orientation for somebody who’s also thinking about the product itself.

It was our CEO at the time, John Thompson at Symantec. This was in the early-2005-ish time frame, and I learned a lot from him. Specifically, his bar for how you interact with customers, how you sell to them, how you serve them, and how you show up in front of them was very, very high.

Through my early career as a product person—I was director of product management—I still remember that it made a huge impression on me.

How was it higher than other people’s? I think everybody would be like, “I’m going to be a high-quality product director.”

What he basically taught me, Sarah, was: When you speak to customers, don’t just ask them how we can serve you better. Ask them what other problems they’re having and what pain points they’re having. You fly there, you meet them, you have coffee with them—whatever it is—and truly understand, because it allows you to see around the corner.

One of the best pieces of advice he gave me was that you cannot be a great product and engineering person unless you speak to customers all the time. All the time. That will allow you to not only do a pattern match but also see around the corner.

So even when we have a platform story, I’ll say, “Hey, this was another retailer at NRF in New York whom I met on Monday or Sunday.” He’s a CTO who reports to the CEO, and I said, “Oh, so you use MongoDB for your e-commerce application?” He said, “Yep, we’re very happy. E-commerce is 20% of our revenue.” I said, “Great. Do you know that you’re not using our search? Do you know we also have vector search?” He said, “Oh, I didn’t know that. Does my team know that?” We are following up with him.

This kind of customer intimacy makes you much better as a product person. I would say any product person who thinks, “As you build, they will come”—that does not happen. It gives you the orientation to understand how they think about deploying your product, how long it will take to deploy that product, and how much value they expect.

I never say that a large bank has bought, say, ServiceNow or Cloudflare. I will say it is Jeremy at this large bank who made a bet on Cloudflare, for example. That, for the last many, many years, has given me a lot of insight into how sales teams show up, how we price for customers, and, if something bad happens—an outage or something—how we show up during the crisis. Those are the things that have made me a very grounded product person and helped me understand the go-to-market channel and the business strategy.

Some of the step-function jumps that a lot of SaaS and infrastructure companies don’t make are from a single product to multiple products, right? Or maybe it’s because they were never platforms to begin with, but let’s say, to multiproduct. And then there are technology transitions, right? You got Atlas to be great and dominant, but it was a question at some point—I don’t know if we’re allowed to say that, right? Many companies struggle through the cloud transition and will now have to address the AI transition. What do you think differentiates a product and engineering organization that does it versus one that fails?

Yeah, you get comfortable. I still remember we were early on ServiceNow on AI, and when I would speak to our engineering team, they were like, “Oh, this is just something that’s out there. Not sure.” I said, “No, not leaning in is not an option.”

Mhm.

Not leaning in—this is a platform. Whether it matures 2 years from now or 4 years from now, we have to do that. I think it is more of a change-management thing, because if you are doing something really, really well—

I’m going to date myself: Think about Nokia handsets. They were doing really, really well. Even if you think about BlackBerry, do you know that when the iPhone launched, I want to say 3 or 5 quarters afterward, BlackBerry was still selling a lot and was not being disrupted until it got really disrupted.

So this transition is more of a change-management thing, and that's when you achieve the step function to say, like, MongoDB did the Atlas transition or multicloud transition nicely, and they have to do the AI transition nicely. Fortunately, a lot of architectural advantages are there, but they still have to nail it and get the trust and information from the customer, and that's when it happens. Otherwise, you are on the bear thesis, and I'm not sure. The only way you prove investors wrong on the bear thesis—because sometimes they don't get it right—is by reaccelerating and showing that, hey, now we're back.

We're out of time here, but I think this is such an interesting question: Can the incumbents get AI right or not? One of the dynamics that I think is worrying people is that one of the ways you sell as a very large incumbent is you bundle a bunch of products together.

Yes.

And then you call a piece of the product that may or may not be working for customers cloud, or you call it AI, and you do pricing hijinks to make the number. I look at this, and I think it's so interesting because so many talented people work at any of these incumbents. My observation would be: You absolutely need people who are committed to those transitions—a leader with innovation as their North Star—and you also need some guardrails for intellectual honesty in the organization between what is working for the 10 customers you talk to every week and the Street.

Yeah. Right. I agree, and the only thing I would say is, that's why, when MongoDB published our Q3 results, I got asked this question over and over again: Is this because of AI? And I said, absolutely not.

You're not allowed [laughter] to say that.

Yeah. No, I said this is our core, and on CNBC I said it's our core. Yes, we have AI-native companies building on MongoDB, and we have hundreds of them. But that's not because, once I give them X% of this, this is how I think about AI-native companies, then they're like, "You can optimize for that."

Yeah. Yeah. And then you go into a different cycle, and I said, yes, whenever they do something at scale—because let's look at it—if you think about however you define success for AI-native companies, whether it's $100 million ARR or $1 billion ARR, pick one, there are like 10 companies like that today.

Today, right? There are not that many companies, and if there are not that many successful companies, they won't have a lot of data. Some of them who have data do use MongoDB, right? So, from my standpoint, this will be an and, not an or: When the AI wave really takes off, that will of course add, but our core data platform and core business is still growing. That's the answer I gave, because that was the truth.

Thank you so much for doing this.

Yes, and thank you very much. I really appreciate it. Thank you. [applause] Find us on Twitter at no prior pod. [music] Subscribe to our YouTube channel if you want to see our faces. Follow the show on Apple Podcasts, Spotify, or wherever you listen. That way you get a new episode every week. And sign up for emails or find transcripts [music] for every episode at no-priers.com.