Cloudflare, Inc. (NET) Earnings Call Transcript & Summary

Thanks, everyone, for joining us. My name is Sterling Auty. I'm the Head of Software here at SVB MoffettNathanson. Very happy to have with us Thomas Seifert, who's CFO of Cloudflare. Thomas, thanks for joining us. We really appreciate it.

Pleasure being here.

So there's basically a couple of big topics that want to dive into. One being AI, and two, kind of some of the changes that you've gone into market structure. AI is the one that obviously has been just front and center through the last couple of days, and we've heard a lot from companies. But maybe the investor conversations I've had, they're really working hard to try to better understand how do these AI companies use the Cloudflare developer platform, especially across the life cycle, doesn't seem to be like a one and done?

Yes. And this is Developer [ Week ]. So we put a lot of announcement out Monday and Tuesday. And then there -- in there -- I would say in typical Cloudflare fashion, AI impacts us in multiple ways. So there is AI companies becoming customers. And in 2 important ways. They consume what we call our Wave 1 product, right? So they need a large front door, they need performance, they need security, lot of fraudulent activity spots trying to misuse the platform running up costs. So they are, in the first step, just a regular Wave 1 application service customers and a lot of the revenue is generated from that. The second big step up now, and that is what we've talked about Monday and Tuesday a lot is using R2 as a means to help them solve 2 really critical infrastructure problems they face. One is finding available and cheap GPU capacity to train their models, and using R2 and Cloudflare literally to find the cheapest and available capacity that is out there. And in the second step, AI GPU hardware now gets constantly upgraded and modernized. So finding not only affordable, available, but also modern GPU capacity in a way that you can access it no matter in what cloud or hyperscale it resides in, and what region it resides without running up significant egress fees moving that data around. So those are already 2 really important use cases. We just announced on Monday a feature that we call Constellation, where we open up AI to developers to run pre-trained models natively on workers, especially latency-sensitive models pre-trained on us. So that is a rather unique use case with a lot of momentum. And then it's the other side. Customers and companies using AI, having employees play around, you have to control your exposure in terms of the data you use that gets uploaded, controlling the APIs that are connecting to large language models, so using our products literally to shield any exposure that might reside. So we put a bundle of Cloudflare One features and products together that helps companies to deal with that exposure. So 3 or 4 already large vectors that impact us. And next topic that is coming back with force is once you operate chatbots, AI-driven chatbots, latency, all of a sudden becomes a big issue. So getting this compute performance and deals as close to the [eyeballs] is possible is just what our network is built for. So now you're already at 4 vectors that are impacting us.

When you talk about one of the announcements helping AI companies find availability and efficient cost on GPUs, is that just the storage of the data? Or are you putting intelligence on top of that kind of almost like a least cost routing? Are you actually monitoring some of the costs and help point them in the right direction?

There it is a product that we launched very early in our life that's called Argo that allows you to route traffic. And in the beginning, it was a tool that allowed you to find the fastest way on the Internet. Then it became a way to find the safest route. So can I avoid certainly to your jurisdictions right around it. Now it becomes their -- the cheapest path to data or to move data. So there is -- you provide value on top of that 2, yes.

And I also want to hit on the data privacy portion, the tool that's monitoring for a company because GDPR was such a big element when it first came out. I imagine this has got to be a huge deal for a lot of your companies, not only European companies, but multinational companies?

Data privacy and data governance has become -- it, first, I think, showed up in Europe with GDPR, and then it became a topic in other jurisdictions, whether it's in South America or in Asia, Japan, South Korea or Australia. So the ability to control the flow of data on a very granular level has become really, really important. And in many ways, I remember, one of our core and star products is DDoS protection. DDoS protection on Layer 7 is really, really difficult because you have to inspect packages. You have to see if there is the person that is -- pretends to be is the right person. So you open it up, you expose log-in data for a second. Under GDPR rules, you cannot do this outside of the European Union. So it's really hard to defend against that you're violating GDPR rules. So finding a way that allow you to control traffic, bring it within the borders of the EU and then open it up, decrypt, encrypt again, has become a hard topic to solve. And together with workers and a feature that we call durable objects, it literally allows you to wrap data on the super granular level.

That's awesome. So with these AI companies that are emerging, is there kind of a consistent tip of the spear? Do they come to you for -- first of all, do they start as free customers? What's the journey that you see out of these customers?

So often, with Cloudflare there is not really a typical answer. So opening up what we Cloudflare Constellation is a developer tool that could be a free tool. It could be a pay go tool where you give developers the opportunity to run pre-trained models on us and connect with workers natively. So could be free to very low prices, I would guess. And then you have large AI companies that come to us in a first step to protect themselves and then the opportunity of R2. R2 today, and I think we said that in the last 2 weeks a couple of times, is small in terms of numbers, dollar numbers, but large in terms of growth rates. So that is opportunity moving forward, but it's covering the broad spectrum of go-to-market and dollar opportunities.

That's kind of a good segue. The other question I wanted to lead into is, okay, given the ratable model that you have in terms of revenue, and when you look at the consumption patterns for these companies, what should invest -- put it bluntly. Investors have come to me and said, geez, AI is exploding. Why isn't it an even bigger part of revenue already? So what should be kind of the anticipation of how will these kind of grow and eventually move the needle?

I think they will grow because the opportunity is just so significant like -- but like so often, in Cloudflare's case, we are not pushing dollars first. We push adoption first, and then you move to revenue and a second step that is the more durable path, that's our experiences, the stickier path that really worked well for workers pushing for adoption before we push revenue. And I think we'll see the same on AI. We'll see broad adoption. We'll see sticky models. We get exposure to many different use cases, and then we'll take it from there. I got the question this morning whether it will be hundreds of millions of dollars by year-end, and for sure not. Yes.

That would be pretty significant.

Yes.

Before I actually ask the next question, let me just remind people that there is a QR code up on the screen. If you want to submit a question, just scan the QR code and it will come up to me on the iPad. One of the other elements is, all right, so ChatGPT OpenAI built on the Microsoft platform. You've got Google with Bard. These are 2 massive companies with huge networks of their own. I get the question put simply to me, why the hell do these companies need a Cloudflare at all? Why can't they do it all inside a Microsoft and Google?

Well, I think the answers are pretty much the same across why do other large customers come to us in the first place because hyperscalers usually have the same challenges that drives customers to select our network for their Internet-facing properties. And in the first place, there is still this fact of -- there's speed of light, so closeness to the eyeballs is really, really important. This is where we excel, and that's why the largest web properties on the planet choose us because of our reverse proxy abilities, that is one. Then we talked about our application service products. They are just really, really good, way better than most of the products hyperscalers have. And then there's the third topic that is what we just talked about on where the benefits of R2 are is avoiding login, our ability to be literally this neutral layer between those hyperscalers and private and public clouds. So it allows you to move data without egress fees at this point and move freely and avoid this function. I think those are the big 3 arguments why this makes sense. And they need really, really big front doors, and that's why some of the largest companies work with us.

And you had mentioned the latency. I mean for any of you that have used ChatGPT 4.0. And you put in your query and you wait and you wait, and you wait. It's clear that reducing that latency is going to be something that's especially important when I also think about the extension, which is, all right, a lot of investors think about the LLMs, the OpenAIs, the ChatGPTs, but there's that whole ecosystem of companies that are building on top of them. How are those companies potentially utilizing Cloudflare?

Well, how do you connect to third parties? You have to open up your APIs, and use APIs to connect. And how do you protect those APIs is a product called API Shield from Cloudflare that is especially that control these connections, who connects, what connects, what data has moved and have visibility around those use cases. That's going to be really important. I mean, you load up your model and it's public. All of a sudden, your IP is gone. So keeping your arms around that is going to be really crucial.

Makes sense. All right. I want to switch topics and hit the other, which is kind of more of the changes in the go-to-market and the outlook. Maybe just to kind of kick it off, when you just reported you lowered the guidance in terms of growth down to the low 30% range. May be articulate for investors. What were the driving forces? What caused that to happen?

So it always -- I think it's -- stepping back, what did we see last year, how did the year develop up to the guidance we gave at the beginning of the year? And what people tend to forget this, we are a subscription model. So even today, with a higher share of variable revenue, we are, to a very significant extent, a subscription model. And subscription model is like you build pipeline, pipeline turns into ACV and that ACV gets recognized over time. So last year, what we saw was, at the beginning of the year, that pipeline decelerated significantly in the first and second quarter, we talked about it. When we started to talk about the slowdown in the first quarter of last year, people said it must be you. It cannot be the macro and then a couple of months later, quarters later, it became a broader acceptance that there was a slowdown. So that pipeline that you lose there is not -- is ACV that shifts out, its revenue that shifts out. And when we gave guidance for the year, originally, we said, we -- the first -- and the risk originally is literally in the first and second quarter because we knew that the pipeline slowdown from last year turning into slower ACV would impact revenue in the beginning. And then in the second half of last year, we saw pipeline accelerate again. We talked about this, third quarter was picking up. Fourth quarter was even stronger, even stabilizing into the first quarter of this year. Sales cycles were ticking up slowly, but nothing too alarming. So when we entered the year, we thought strong pipeline, slightly extended sales cycles, more pressure on guidance in the first and second quarter because of what we've seen last year, but more optimism in the second half. That is how we framed guidance entering the year. And then we saw things changed significantly in the first quarter with the event of one bank, which really changed the buying behavior significantly. And our sales cycles in the first quarter of this year literally took a step function. They went up by 27% on average and some parts of our business especially in the expansion business with very large customers, it went up by 45% to 47%. You have to understand that our sales cycles are really short. So normally, we close business within a quarter. So if a sales cycle goes up by 27%, we lose a significant amount of ACV in the quarter. And that means that ACV we lose in the quarter automatically is revenue or loss in the quarter. So it's like a boat -- that is building up a wave that is building up in the front of the boat, you just pushed this out. And that is what caused us to adjust guidance. So that revenue -- that ACV is revenue loss in the first quarter, and we just assume that with what we see in the market, the sales behavior is going to stay with us and caused us to take guidance down for the year. So not so much a matter of, if we see revenue or if we see that business, but when will we see this business coming in. And the subscription businesses are slowing the way down, but they're also slowing the way up. That is just the nature of the business. So that is what caused us to adjust guidance in the first quarter.

So what gives you the confidence that this is the right level, that this is at a level that you can underpromise and overdeliver?

Guidance inherently has risk, right? So you cannot just eliminate all risk. Otherwise, why guide in the first place. So you have to make certain assumptions. We always thought we are prudent and thoughtful in how we think about business. We model the heck out of the -- of various outcomes before we decide on the guidance range. This explosion in sales cycles was way beyond any confidence in the world that would have made sense as the basis for guidance. So we think we are thoughtful in terms of what we assumed for the remainder of the year. We said for the second quarter, we assume that only the extended sales cycles. We assume that linearity would not change on top of extended sales cycles. We saw dramatic nonlinearity in the quarter. So that would mean also very little to almost no revenue recognition from new ACV in the current quarter. And we assume that sales cycles will not improve. So we think we are thoughtful about what we assume for the rest of the year, and we'll play to it.

Yes. I received a lot of questions saying, well, geez, you made a change at the head of sales level very late last year. Why isn't this just a reflection of disruption that came about. Or to put it another way, is this timing of sales execution improvements playing a factor? Or is there a disruption that is still to come?

The -- it's not like we woke up one morning and said, there's room to improve on the sales execution. That was something that has been building up over time. We've been working on almost everything that is in flight today has been in the works for an extended period of time up to third quarter, starting already in the third quarter of last year. The -- you're also not hiring a new Head of Sales overnight. That is a process we started at the very beginning of last year already. So we knew there was room for improvement and there needs to be room for improvement, right? Our journey from where we came, freemium, pay-as-you-go, mid-market, to now our largest and fastest-growing cohorts from a revenue perspective are the largest accounts. So there is room to adjust. There's a necessity to involve the business model. So that was in flight. And the previous guidance already assumed that this would happen, and we didn't model any improvements as we said from that, from those opportunities into the last guidance. So now you can talk about whether the timing was fortunate or not to put this all in one quarter. But there's also a time where we think it made sense for us to execute on this. When your revenue growth is still extraordinarily high like last year, your opportunity cost for disruption are higher. The labor market for goods -- for very good enterprise sales [ people ] was really tight and expensive last year. So -- and we thought this is not the right time to execute on all the measures to the fullest extent. So they fall together, but the guidance impact is -- adjustment is a macro topic and was not driven by the things that were in flight already from a sales execution perspective.

So you recently held an Investor Day and your new President of Revenue, Head of Sales, did a really good job. And in talking with him, it feels like a lot of the changes have been instituted already are really good blocking and tackling items that we've seen in other companies like Zscaler when Dali came in and elsewhere have had significant positive impacts. When you look at what's been instituted thus far, what are some of the changes that jump out to you?

I mean, there's a lot that was in flight already that we talked about before and folks tend to forget that. One of the -- we -- our journey over the last year is quite remarkable already from a product perspective, right? We have about 50 products that are revenue contributing today. So how do you bring those products to market? How do you deal with the complexity of pricing? So moving to bundles, what are the bundles, how do you price the bundles, what billing engines do you need? But then also, what is there -- how do you change your go-to-market behavior? What are the sales specialist overlays you need for what product? How is the enablement going to change the personas that are needed are going to be adjusted that we target so the campaigns have to be redone or adjusted to that. And so a big part of our pipeline improvement last year was already a reflection of the campaign changes that were made. So a lot of it was in flight already and this has shown already with lead time then that we see a remarkable improvement. But it's -- I think for me, the most impressive changes are the one where product meets customers, so to speak, in terms of bundling, pricing and enablement.

And then what's left to kind of come online? I mean you talked about the headcount changes, but I think the improvements are well beyond just changing out 100 heads.

Yes. I mean you could argue productivity is a race that's -- without a finishing line. So it will continue because this journey in terms of more products consumed by more customers, customers getting larger and then we grow up the stack is going to continue. So our go-to-market has to evolve with it. But in the near term, it's onboarding new employees, enabling new employees, making sure that we are staying within the productivity targets we have and our productivity ramps are pretty interesting. So we -- mid-market ramps to full productivity within about 4 quarters -- 4 months, 6 months about for an enterprise AE. So you're making sure that we enable that and we see productivity improvements within the current year.

Say that one more time because I think that's an element that a number of investors miss. I think about a traditional enterprise sales rep takes 4 to 6 quarters to ramp effectively. Sounds like what you're saying is both middle market and enterprise ramp much faster for you.

Yes, yes, 4 to 6 months.

4 to 6 months is a huge difference.

Yes, it is.

Because we tend to think about sales transitions or changes taking upwards of 3 quarters to really start to show improvement. If it takes 4 to 6 months to ramp a new rep, it suggests that your time line should be much more abbreviated.

Yes. You have to keep in mind that the products have become more difficult, but the implementation of products for a customer to get on our network are really, really fast. Once you're on a network, from an expansion perspective, literally, every product is just a mouse click away, right? There -- even for complex products, they're hardly any professional services impacted. We onboard the largest financial institutions under the most sophisticated DDoS attacks within hours, right, not within days or weeks or months. So the efficiency of the products also caters to that.

It's a good position to be in. Because of the bundling, I think it's always been difficult to parse out. How is your security business versus your performance business, et cetera, but you did provide some commentary. How do you look at that? Is there -- is it -- are you able to see which is being impacted more from the macro side, for example, at the moment?

Not really. And I think that some products are more complicated than others. Magic Transit, where you onboard customers with complete corporate networks are more complex sales. They might be impacted a little bit more. But I don't think you can differentiate down to a product level. It's -- we see this across the board.

Yes. Now you talked about that pipeline generation went through a cycle through last year. You start to see improvement in the back half of the year, stabilization in the first quarter. Is there a sense either geographically size or company, industry, what were some of the trends that you're seeing in the pipeline?

Not a lot to be very honest. And we thought we would see more impact in Europe because of the war in the Ukraine, but that didn't turn out to be the case. If at all, Europe again in the first quarter was one of the strongest regions for us. There's not a big differentiation in terms of products or in terms of regions or in terms of customer sizes, with the exception that I think expansion in large enterprises had the longest -- so had the most elongation in terms of sales cycles. And. My interpretation was that large organizations are very disciplined, very rigorous processes, probably reacted the fastest to a change to a higher degree of uncertainty after some of the banks got wobbly and just step on the brakes significantly faster than others in terms of reacting and holding back and measuring twice, putting additional signature loops into their procurement processes.

Did you see -- so that happened in March to the stuff that you can comment, kind of which is when you gave guidance on the quarter. Was there any noticeable change? So was it kind of a pause and a little bit of a bounce back and going into the guide or...

We saw 2 -- we talked about this. I can comment here. We saw 2 things that really jumped at us there, that sales cycles really exploded, went up 27%. Collection behavior changed. The payments came in really, really slow. Payments recovered as soon as the first one or two banks got rescued. So payment and collection behavior recovered. Purchasing behavior has not really changed, didn't deteriorate, but it did not bounce back either.