Cloudflare, Inc. Q2 FY2023 Earnings Call

Matthew Prince, Co-Founder and CEO; Michelle Zatlyn, Co-Founder, President and COO, and Thomas Seifert, CFO. By now, everyone should have access to our earnings announcement. This announcement, as well as our supplemental financial information, may be found on our investor relations website. As a reminder, we will be making forward-looking statements during today's discussion, including, but not limited to, our customers, vendors, and partners, operations, and future financial performance, our anticipated product launches and timing and market potential of those products, our anticipated future financial and operating performance and our expectations regarding future macroeconomic conditions. These statements and other comments are not guarantees of future performance and are subject to risks and uncertainties, much of which is beyond our control. Our actual results may differ significantly from those projected or suggested in any of our forward-looking statements. These forward-looking statements apply as of today, and you should not rely on them as representing our views in the future. We undertake no obligation to update these statements after this call. For a more complete discussion of the risks and uncertainties that could impact our future operating results and financial condition, please see our filings with the SEC as well as in today's earnings press release. Unless otherwise noted, all numbers we talk about today, other than revenue, will be on an adjusted non-GAAP basis. You may find a reconciliation of GAAP to non-GAAP financial measures that are included in our earnings release on our Investor Relations website. For historical periods, a GAAP to non-GAAP reconciliation can be found in a supplemental financial information referenced a few moments ago. We would also like to inform you that we will be participating in Stifel's Tech Executive Summit on August 29th, and the Goldman Sachs Communacopia and Technology Conference on September 6th. Now, I'd like to turn the call over to Matthew.

Thank you, Phil. We had a strong quarter in spite of continued macroeconomic uncertainty. In Q2, we achieved revenue of $308.5 million, up 32% year-over-year. We added 196 new large customers, those that pay us more than $100,000 per year, and now have 2,352 large customers, up 34% year-over-year. Our focus on go-to-market improvements is already paying off. After we saw sales cycles increase 20% in Q1, disciplined around deals had them return to levels closer to what we saw last year. While our customers and prospects continue to be very careful around their IT spend, our improved execution led to a record quarter in new ACV bookings. My sense, talking to customers is that while the macro environment is still challenging, it is stabilizing. And for the first time in several quarters, sentiment among IT buyers does not appear to be getting worse. Our dollar-based net retention ticked down to 115%, down 2% quarter-over-quarter. Dollar-based net retention is a lagging indicator, so it will be slower to reflect the go-to-market improvements we are seeing. It's also important to note that we did not see any new competitive pressure or churn throughout the quarter. Instead, the lower dollar-based net retention is due to slower expansion from some of our existing customers. We expect that our focus on go-to-market operational excellence will improve this metric over time. Our gross margin held stable at 77.7%, still above our long-term target of 75% to 77% and in line with 77.8% last quarter. We delivered an operating profit of $20.3 million, our fourth consecutive quarter with a record operating profit. We also meaningfully outperformed on free cash flow, generating $20 million during the quarter, which represents a free cash flow margin of 6.5%. I'm proud that our team has proven we can not only execute in good times, but also be disciplined and deliver operational improvements while we're in more challenging times. We continue to see very strong pipeline growth. Q2 was another record for new pipeline generation. As we discussed last quarter, we made significant changes in our sales team to proactively address underperformance. That went very well, both qualitatively and quantitatively. Our top performers are invigorated. We saw a marked improvement in the average account executive productivity. At the same time, we've implemented robust onboarding, enablement, and training programs. Combined with the record number of applicants we're seeing for sales roles, this makes for the right formula to build a world-class sales organization. And our team is armed with great products to sell. Last quarter alone, Forrester recognized Area 1, our email security product, as a leader. IDC recognized us as the leader for two reports, in Zero Trust and network edge security-as-a-service, and we were the only new vendor recognized by Gartner for Secure Service Edge. Our developer platform, Cloudflare Workers, continues its explosive growth. We reached 10 million active Workers applications in Q2, up 250% since December and 490% year-over-year. R2 continues to grow and now stores over 13 petabytes of customer data, up 85% quarter-over-quarter. We have 44,000 distinct paying customers with R2 subscriptions, and brand name customers are beginning to adopt it as their primary object storage solution. That seems like a good segue to name some other customer wins in the quarter. One of the fastest growing generative AI companies expanded their relationship with Cloudflare, signing a one-year $1.7 million contract, less than a year after first starting to use our platform. Like many AI companies in the space, this customer relies on a multi-cloud architecture for training and processing requests. R2 lets them unlock the best prices and performance across multiple cloud providers. In their words, 'We see Cloudflare as a strategic foundational glue across all our services.' That’s great to hear from many customers, but it's especially fun coming from a company that's doing such cutting-edge work. These days, it feels like membership in the CEO club is predicated on saying AI as many times as possible in your earnings call script. I have to confess, I still find it a bit awkward. When we first pitched Cloudflare to venture capitalists back in 2010, at one point, I described us as, 'The first AI powered security company for the cloud.' The eye rolls around the table were so intense that I'm still a bit scared. But more than a decade later, here we are. And by our estimates, Cloudflare is the most commonly used cloud provider across the leading AI startups. They're using R2 to help arbitrage the lowest GPU cost to train their models. They're using our security tools themselves powered by AI or what our team would prefer to call machine learning to protect their own AI systems. And increasingly, they're using the edge of our network to perform inference. We are continuing to invest in this area and believe that we are uniquely positioned to win the inference market, which we believe will be substantially larger than the AI training market. In Q2, we hosted our Developer Week, highlighting 10 major announcements and features to extend Cloudflare Workers as the preeminent developer platform for the leading AI company. Q3 will feature our annual birthday week, and we have a lot more in store to provide picks and shovels to enable AI companies to build the future. Beyond AI, Cloudflare’s Zero Trust Solutions were another big winner in Q2. A Fortune 500 technology services company expanded their relationship with Cloudflare, signing a three-year $7.2 million contract for 25,000 Zero Trustees. That brought their annual spend with us to over $5 million. They first became a customer in Q3 last year using our application security products. Six months into the deployment, one of their senior executives said, 'Cloudflare is like magic and brought us into an ongoing competitive Zero Trust proof of concept.' Cloudflare's existing gateway products were chosen over first generation Zero Trust competitors due to our rate of innovation and ability to consolidate all their security onto a single pane of glass. One of the largest online recruiting platforms expanded their relationship with Cloudflare, signing a 25-month $2.4 million contract and bringing their annual spend over $5 million. With more than 90% of their employees remote, they were looking for a comprehensive Zero Trust solution and evaluated us against every leading vendor in the market. They decided to go all in on Cloudflare with 15,000 seats for access, gateway, CASB, data loss prevention, browser isolation, and Area 1 email security. I'm especially proud of how quickly we were able to onboard them, less than a month to fully replace their first generation Zero Trust vendor. That’s awesome. An Australian technology company expanded their relationship with Cloudflare, signing a one-year $2.2 million contract, bringing their total spend with us to over $5 million. This customer started out on our pay-as-you-go plan in 2016. This quarter, they signed a Zero Trust deal to protect their expanding workforce. They're also broadening their use of Cloudflare’s developer platform with both R2 and Durable Objects. Sticking down under, a leading Australian healthcare provider expanded their relationship with us, signing a three-year $2.8 million contract. We are replacing their hodgepodge of first generation Zero Trust vendors with 12,000 seats of access, gateway, browser isolation, CASB, data loss protection, and Area 1 email security. This is another example of a customer looking to consolidate vendors and choosing Cloudflare for their holistic network security solution. Here's another cool one. A Fortune 500 social network expanded their relationship with Cloudflare, signing a three-year $2.4 million contract. They initially became a customer a year ago, building on top of Workers and using our global network to authenticate the security of one of their messaging products. They approached us again looking to add increased privacy onto another product with our Privacy Gateway solution. They view Cloudflare as a leader in privacy based on our co-development of the Oblivious HTTP standard, and they admire us as one of the only other companies that truly understand scale. As privacy is increasingly top of mind, we believe there will be more and more of these sort of strategically beneficial relationships. I think I've only said AI 11 times so far, putting me way behind Satya. So let me end with one more AI customer win. Another generative AI company expanded their relationship with us, signing a three-year $1.3 million contract. They came to us for our developer platform, signing up as a pay-as-you-go customer. Because their developers loved us, they approached us about a security need and signed a deal to use our application security and Zero Trust products. They're only 100 seats, but they're growing rapidly and building Cloudflare deep into their entire stack. Whether you're a Fortune 500 industrial company that used us for application security and are now hiring AI developers to use our Workers' platform to drive innovation across your business, or you're a 100-person AI startup that started using our developer platform and then realized you can get the same security as the biggest Fortune 500 companies, that's what's really unique about Cloudflare. We've built the cloud that connects the world securely, reliably, and efficiently. With that, I'll turn it over to Thomas.

Thank you, Matthew. And thank you to everyone for joining us. During the second quarter, I'm pleased to share that we've seen improvements in terms of the impact from the external challenges that we highlighted last quarter. Specifically, also still somewhat elevated from historical levels, sales cycles shortened in part due to the implementation of more efficient processes and tactics. Our pipeline close rates have also shown improvement as we continue to refine our go-to-market strategies and operations. Furthermore, we observed a notable uptick in collections on our accounts receivable, which we believe reflects a rebound in customer confidence and financial stability. We also continue to maintain our strong commitment to being fully responsible and act as good stewards of investors' capital. We delivered our fourth consecutive quarter of record operating profit. We also prudently allocated capital with a focus on maximizing shareholder value by taking action to retire our 2025 convertible notes during the second quarter. Turning to revenue. Total revenue for the second quarter increased 32% year-over-year to $308.5 million. From a geographic perspective, the US represented 53% of revenue and increased 30% year-over-year. EMEA represented 27% of revenue and increased 38% year-over-year. APAC represented 13% of revenue and increased 23% year-over-year. We were pleased to see notable performance in the EMEA and APAC regions, with both achieving record new ACV bookings in the second quarter. The strength in APAC was primarily driven by large customer deals, and we are seeing security become an even higher priority in EMEA given the geopolitical situation in the region. Turning to our customer metrics. In the second quarter, we had 174,129 paying customers, representing an increase of 15% year-over-year. We ended the quarter with 2,352 large customers, representing an increase of 34% year-over-year and an addition of 196 large customers in the quarter. In fact, we added a record number of customers spending more than $500,000 on an annualized basis with Cloudflare. And the second quarter was also one of our highest quarterly additions of customers spending more than $1 million annually, including our largest Zero Trust contract to date. Our dollar-based net retention rate was 115% during the second quarter, representing a decrease of 200 basis points sequentially. Importantly, renewal rates in the second quarter were consistent with the quarterly average in 2022, which was an all-time high for the company. Instead, similar to the last two quarters, the decline in DNR was again primarily driven by slower expansion in our larger customer cohort. We calculate DNR by comparing the analyzed revenue from paying customers four quarters prior to the annualized revenue from the same set of customers in the most recent quarter. As a result, this will be a lagging indicator of Cloudflare’s underlying business trends. Based on our visibility, we believe the deceleration in DNR is nearing a bottom. Moving to gross margin. Second quarter gross margin was 77.7%, representing a decrease of 10 basis points sequentially and a decrease of 120 basis points year-over-year. Network CapEx represented 11% of revenue in the second quarter. For fiscal 2023, we now expect network CapEx to be 10% to 12% of revenue, underscoring the scalability and efficiency of our network even as we onboard new workloads, including AI. Turning to operating expenses. Second quarter operating expenses as a percentage of revenue remained consistent sequentially and decreased by 8% year-over-year to 71%. Our total number of employees increased 11% year-over-year, bringing our total headcount to 3,389 at the end of the quarter. During the second quarter, we addressed consistently low performing sales capacity with a focus on upgrading our customer-facing talents to improve growth, increase productivity, and drive long-term success. We will continue to pace hiring for the year based on market conditions and remain committed to raising the bar on new high additions given talent opportunities available in the market. Sales and marketing expenses were $125.4 million for the quarter. Sales and marketing as a percent of revenue decreased by 1% sequentially and decreased to 41% from 44% in the same quarter last year. Research and development expenses were $53 million in the quarter. R&D as a percentage of revenue decreased by 1% sequentially and decreased to 17% from 20% in the same quarter last year. G&A expenses were $41 million for the quarter. G&A as a percentage of revenue increased 1% sequentially and decreased to 13% from 15% in the same quarter last year. Operating income was $20.3 million compared to an operating loss of $891,000 in the same period last year. Second quarter operating margin was 6.6%, an increase of 700 basis points year-over-year. These results highlight our continued focus on becoming more efficient and more productive, not just during the currently uncertain macroeconomic backdrop, but also because operational efficiency is a long-term competitive advantage. Turning to net income and the balance sheet. Our net income in the quarter was $33.7 million or a diluted net income per share of $0.10. We ended the second quarter with $1.6 billion in cash, cash equivalents, and available-for-sale securities. Free cash flow was $20 million in the second quarter or 6% of revenue compared to negative $4.4 million or 2% of revenue in the same period last year. Remaining performance obligations or RPO came in at $1 billion, representing an increase of 8% sequentially and 36% year-over-year. Current RPO was 75% of total RPO. Before moving to guidance for the third quarter and full year, I would like to begin with our expectations and the provisions we have factored into this outlook. Despite being encouraged by the forward progress we delivered during the second quarter in terms of shortening sales cycles and improving close rates, mixed macroeconomic data points serve as a reminder that we are operating in a business environment that, while showing signs of stabilization, continues to be challenging to predict. As a result, we remain prudent and cautious in our outlook for the second half of the year and we are fully committed to continuing to adapt our tactics and strategies in response to these external variables. Now turning to guidance. For the third quarter, we expect revenue in the range of $330 million to $331 million, representing an increase of 30% year-over-year. We expect operating income in the range of $20 million to $21 million and we expect diluted net income per share of $0.10, assuming approximately 347 million shares outstanding. We expect an effective tax rate of 11%. For the full year 2023, we expect revenue in the range of $1.283 billion to $1.287 billion, representing an increase of 32% year-over-year. We expect operating income for the full year in the range of $81 million to $85 million. And we expect diluted net income per share over that period to be $0.37, assuming approximately 345 million shares outstanding. We expect an effective tax rate of 9% for 2023. After having achieved positive free cash flow in the second half of last year and again during both the first and second quarters of this year, we anticipate generating significant free cash flow for the full year 2023. For modeling purposes, we continue to expect free cash flow to trend upward on an ongoing basis, but anticipate variability in our free cash flow generation quarter to quarter. In closing, our team remains committed to driving operational excellence, ensuring long-term growth and delivering significant shareholder value. I'd like to thank our employees for their continued dedication to our mission, customers, and partners. And to our shareholders, we greatly value your continued support. And with that, I'd like to open it up for questions.

Thank you. We will take our first question from Shaul Eyal with TD Cowen. Your line is open.

Thank you. Good afternoon, guys. Congrats on the results and the outlook. Matthew, can you talk about some of your displacement activity this quarter? And maybe my second part of the question will be, do you see AI accelerating your displacements and win rates? Thank you.

Yeah, Shaul. I really appreciate the question. We've had a lot of noise on the line from the operator. So hopefully, we can get things muted. Apologies for that. So I think that when we look at displacement, it really is across three different parts of our business. The first area is our conditional application security business. In that space, we continue to displace a number of traditional hardware vendors, the people who are providing web application firewalls, load balancers, funding various services that people had in those areas. We also see point cloud solutions that are doing just one of those things getting displaced by us where we can pick up a significant amount of their business, and that's been the case for quite some time. In our second part of our business, which is our Zero Trust business, the first is sort of the front door of your business, the second is kind of the back door of your business. The Zero Trust business is about protecting employees and data. In that case, we're more and more going head-to-head with the other first generation Zero Trust providers. So the Zscalers, Palo Alto Networks, Cisco Umbrellas of the world. And, again, we really like our win rates in this space. In a number of the examples that I cited, we were specifically either in competitive solutions there, or in some cases where we were actually doing takeouts, people who have made bets on the first generation of Zero Trust providers and wanted to upgrade to us for better ROI, a much better user experience, and a much faster and more performant network. And I think that that's one of the areas that I'm the most excited about. The third area is really our Workers' business, which is our developer platform. In that space, it really depends on what's going on increasingly, we're doing takeouts from object stores where people are moving data off of more traditional object stores onto R2, which is our object store. But a lot of times, we're also just moving individual applications or individual functions to us. And so it's not a complete displacement. Oftentimes people will use us alongside a more traditional hyperscale public cloud. But we can see that working together. In the AI space, in particular, I think that, again, it's such a new space that I don't know that we're displacing people as much as we're just helping AI companies get what they need. And the two big areas around that are first around training, where GPU scarcity is significant, and the cost with the traditional hyperscale public clouds and moving data to wherever there's cheap GPU capacity or even available GPU capacity makes it cost prohibitive. R2, because we don't charge for egress, has been a real boon for a lot of AI companies to be able to adopt wherever they can find the cheapest GPU at any moment in time. And that again, that’s an area where a lot of that growth has come from. And then increasingly, we think that the inference market is really going to be fought between two areas. One is going to be on your device itself. If you have a driverless car, you don't want when a ball is bouncing down the street and the kid is chasing after it, for that decision on whether or not to put on the brakes to have to go out to the network now. You want that to live in the car itself. And so a lot of inference and models can be run on devices. But we think if they're not run on devices, if they’re too large, if they need too much capacity, from either a GPU or memory or network access space, in those cases, it's going to make sense to run it in the network itself. And in that sense, Cloudflare is uniquely positioned to win in that inference market for those models that make sense not to run on the device themselves, the more complicated models that make sense to run out at the edge of the network. And that's exactly what we're starting to see from more and more of these really innovative AI startups.

Got it. Thank you so much. Speak to you all.

And we will take our next question from Matt Hedberg with RBC Capital Markets. Your line is open.

Thank you for taking my question. Thomas, I wanted to ask you about the positive results this quarter and the optimism expressed in your prepared remarks. You mentioned improved win rates and shorter deal cycles. Observing the sequential growth for Q3 and Q4, it appears to be stronger than what we experienced in the first half. What are some key factors contributing to this increased optimism for the second half? Also, does your guidance assume that the macro environment remains stable or potentially improves?

Well, thank you, Matt. We saw the first data points that made me a bit more optimistic. We still think we need to apply a good portion of caution to our outlook. One data point doesn't really make a trend. At this point, we do not assume that the macroeconomic environment is improving significantly. We still see significant mixed macroeconomic data points that we factored into our guidance. And therefore, it sets data to be prudent and cautious, not to over-interpret just on one data point.

Got it. That's helpful. Maybe just a quick follow-up. Sort of double-clicking on the strength that you saw this quarter, there was a lot of conversation about shorter sales cycles, better win rates. Have you noticed any quantifiable benefit from your new CRO, Marc, as he's come on board and maybe improved sort of the sales focus?

We're making good progress in restructuring our team and adapting our tactics and strategies to move upmarket. However, it's important to remember that onboarding new team members takes time, which is slightly shorter for the mid-market and longer for the enterprise segment. We have not yet seen most of those improvements and do not expect to see them for the rest of this year. While we are making significant progress, we remain cautious about what we have included in our guidance for the second half.

Thanks, guys.

And we will take our next question from Trevor Walsh with JMP Securities. Your line is open.

Great. Thanks for taking my question. Maybe, Matthew, just for you, first. A lot of the comments around AI seem to focus more around the developer piece and Workers specifically, which you kind of have as your part of your Act 3. Does that become something accelerated into more of Act 2 for that part of the platform? Can you just maybe talk a little bit about the changes of the game from that perspective?

Sure. Thank you, Trevor. I believe that the sequence of focus remains as follows: Act 1 is application security, Act 2 is Zero Trust, and Act 3 is Workers. However, I must say that we've been pleasantly surprised by the rapid growth of the Workers' platform. At this point, we are not prioritizing maximizing revenue from it. Historical analysis of developer platforms shows that adoption is key. What encourages me is that every day, as new companies emerge and trends like AI gain momentum, individuals are gravitating towards this platform. I don't view this as a temporary phase; it feels substantial and enduring. More companies and experienced developers are increasingly turning to the Workers' platform for solutions unavailable through traditional hyperscale public clouds. While I recognize the importance of our progress in Act 2 products this year, which has been impressive, I see exciting prospects in both the second and third acts.

Thank you, Matthew. I appreciate the insights. I have a quick follow-up for Thomas. Of the 196 large customers you added this quarter, could you share whether these are existing customers reaching that size, or if there is a significant number of new customers in that range? Can you provide a general sense of where these customers are coming from?

In the past, it was pretty even, fifty-fifty between new customer sign on and defender expansions. I would say, in the last quarter specifically, we had probably a larger share of new customers signing up right beyond the $1 million. So it shifted slightly away from expansion into a new logo sign on.

Awesome. Great. Thanks again for taking the questions.

And we will take our next question from Jonathan Ho with William Blair. Your line is open.

Good afternoon. This is John Weidemoyer for Jonathan. Thank you for taking my question. I would like some clarification on your net retention rate. If I understood correctly, you mentioned in your prepared remarks that the retention rate is a lagging indicator that should improve due to your go-to-market and sales replacement initiatives. However, I believe you also stated that this is influenced by your largest customer cohorts. I would not expect large customers to be significantly affected by your sales or go-to-market strategies, as it seems that mostly existing customers would be. Could you help me understand how this impacts the retention rate and what some of the factors at play are?

All right. Well, just to make sure, the expansion rates stayed high. Expansion in the large customer cohort was a little bit lagging. That is what we already saw in the prior quarter. It's also one of the reasons for my previous answer that the large customer growth was pretty much coming more biased towards new logos than it was coming from expansion. Expansion, as I said in my prepared remarks, is a lagging indicator because it's pretty much a look back to the four prior quarters. That is what you compare to a sign on. So any movement we see in an existing quarter will take time to show up in DNR. We think we are seeing bottoming of the DNR development. So we are quite confident we'll move that upwards to where and beyond where we came from. But because we are so conservative in how we measure DNR, it's very much a lagging indicator. So it will take a while before all the improvement that we are initiating in getting expansion going again will show up in DNR.

Thank you. I would like to ask about the initiatives you mentioned, specifically regarding the new sales personnel who may not be fully productive until the end of the year. That makes sense. Can you provide an update on your progress with the sales processes? At what point do you expect to be halfway done or fully completed, and have you identified any additional process improvements since our last discussion a quarter ago? Thank you.

Yeah, I'll respond to that. I believe Marc is doing an excellent job overseeing the organization. Last quarter, we pointed out some areas of underperformance, and we've taken steps to address that. Over the last three months, I've been very attentive to the organization, meeting with many of our team members. What I'm hearing is that they feel highly motivated and appreciate the additional training and support we've introduced. Our sales team is now equipped with the necessary tools to help them successfully close deals. We still have a robust sales pipeline, and I believe you'll see positive results as our new representatives join and we continue to provide support to our existing team. This is very encouraging for our future prospects.

Thank you.

We will take our next question from Brent Thill with Jefferies. Your line is open.

Good afternoon, Matthew. If you take your crystal ball out in the second half of the year, I'm just curious if you feel things are starting to slowly improve? It seems like a lot of your security peers are starting to see some decay and perhaps you're gaining some share here relative to your architecture and the platform. I'm curious if you can kind of maybe stitch the back half together and how you see it at a 40,000-foot view.

I’m not sure how accurate my predictions are, but it doesn't seem to me that things are getting better; rather, they appear to be stabilizing. The first quarter was quite challenging. A 20% increase in sales cycles within a single quarter was significant and concerning. While it was early in the earnings season to highlight worries among IT buyers, other companies echoed similar sentiments after us. In the second quarter, we noticed that sales cycles aligned more closely with what we experienced in 2022, which was still higher than in the previous years. My impression is that we are in for a tough road ahead, not just for Cloudflare, but across the entire economy. However, this challenge is pushing us towards operational excellence. Our team is committed, working diligently to optimize every process. You’re right—one aspect that sets us apart is our ability to achieve high gross margins with our products, which reflects our differentiated platform. As customers look to consolidate vendors and maximize their return on investment, they are increasingly turning to us. I believe that while the challenges ahead will be tough, they will ultimately strengthen us as a company.

Thank you.

And we will take our next question from Keith Weiss with Morgan Stanley. Your line is open.

Thank you for taking the question, and it's encouraging to hear that things are stabilizing. Matt, I want to explore your comments about Workers and Cloudflare being a strong platform for inference. Can you explain the technical reasons behind this belief? Since we are still in the early stages of these technologies, understanding how inference will develop over time would be valuable for both the industry and the Cloudflare discussion about your positioning. Thomas, when we think about inference, we often associate it with being GPU intensive and having lower gross margins. You seem confident that this won't affect gross margins in the near term. Is this mainly because it's still early and volumes are relatively small, or is it possible that as this grows, it could have a more significant impact on gross margins over time? Thank you.

So, Keith, I appreciate both questions. They're really, really important to understand the advantages that we have in this space. So, first of all, what are the challenges with inference? And I think there are really two. One seems like a bigger deal and is probably actually not as big a deal. The other doesn't seem like it's a bigger deal, but it is a really big deal. And in both, I think that they shape how we think the inference market is going to work out. So the one that kind of feels like it's a bigger deal is around performance, which is that if you're playing with the various generative AI companies, if you're trying to do something, that wait time between when you submit a query and when you get back a response that's going to become a bigger and bigger differentiator between different AI platforms. And so anything that you can do in order to make that performance as fast as possible is advantageous. And one of the ways to do that is to move the actual inference as close as possible to the person who's requesting it. And so, again, we think that inference will primarily be done on device or very close to where the end user is, inside the network. We won't get, again, if the ball is bouncing across a street, you want that inference to be done on the device, on the driverless car itself. So we won't win every inference task. But there will be a lot that makes sense to run in the network where we have, again, almost infinite network capacity, almost infinite storage and memory, and very, very significant CPU and GPU resources to be able to run those inference tasks. That I actually think will be the lesser of the two advantages for us. The larger one, which again doesn't feel like it's a bigger deal, but we're already seeing it play out some of the regulatory efforts that are happening around the world is that a lot of times for these inference tasks, the data that there is very private. People and governments want that to stay as close to the actual end user as possible. So we've already seen action in Italy that has restricted the use of certain AI tools because it sends data out of the country. What Cloudflare can uniquely do because we're positioned across more than 250 cities worldwide, we are in the vast majority of countries worldwide is that we can actually process that information locally. So again, we think that on device we are very close to where the user is on Cloudflare’s network, is going to be the place where inference is going to take place. I'll take a quick stab at your second question as well and then hand it off to Thomas for anything that he would add.

One clarification. So in that world view, it sounds like you think we're going to see more kind of smaller open source and distribution of a lot of very small models versus like a world view that everything is going to come up into a big GPT or LaMDA model over time. Is that correct?

Not necessarily, but we run enough capacity to see out at the edge of our network that we can run fairly large, I mean, very large models out at our network. And what I think is a little bit confusing is most of, if you're trying to do the training of the models, then having the absolute latest, greatest GPU, the H100 from Nvidia right now, there's a lot of constraint in getting those chips. But there's actually a sweet spot for inference tasks, which isn't necessarily at the absolute cutting edge of the models. And so Cloudflare is not the right place to actually process the training of models. That makes much more sense to do in a more traditional, centralized data center model, much like much of the traditional hyperscale public clouds. In those cases, you have to have the latest, greatest GPUs. But when you're doing inference, again, a lot of that's going to run on your device. And a lot of that is also going to run inside the network. And we're going to be able to, with a much lower capex spend, leverage the edge of our network, in order to be able to do that processing extremely efficiently. And maybe we don't need the H100. Maybe we can live within A100 or whatever is again a generation or two behind. But that's also the difference between training and inference. Inference doesn't need necessarily the latest greatest GPU. Does that make sense?

Yeah, super helpful. Thank you.

What I would add is that to truly understand the competitive advantage of Cloudflare and the efficiency of the business model, it’s essential to start with the network and how it is effectively optimized using standard hardware and a fully integrated, uniform software stack that allows every product to run on any server at any location. This extensive, globally distributed network is designed not only to manage large volumes of data but also to handle numerous simultaneous requests efficiently. This makes it well-suited for inference tasks, which inherently involve processing multiple requests at the same time. Additionally, it requires less computational power than training models. Therefore, the architecture of the network itself places us in a favorable position, which is why we are confident that the business model will remain robust. It also explains why we are able to reduce AI's capital expenditure ratio for the year, even as AI workloads increasingly utilize our network. It's crucial to revisit the efficient architecture of the network itself to uncover the reasons behind our confidence.

Excellent. Super interesting guys. Thank you.

And we will take our next question from Andrew Nowinski with Wells Fargo. Your line is open.

Okay. Thank you. And congrats on another great quarter. I wanted to shift gears and ask about Zero Trust. Is there any more details you could provide on record Zero Trust contract you talked about and whether that was a displacement of another vendor and maybe why they selected Cloudflare. And then I have a quick follow-up. Thank you.

Sure. In almost all of the Zero Trust deals we encounter, we are competing with some of the first generation Zero Trust vendors. In many cases, there’s an incumbent vendor that we are displacing. This usually occurs because the usability of the existing Zero Trust vendor has been poor. For instance, it’s unacceptable that some leading Zero Trust vendors have issues like the captive portal on United Airlines Wi-Fi not working when you try to use your laptop. While that may have been tolerable during the pandemic when travel was minimal, it’s no longer acceptable now that people are traveling again. One of Cloudflare’s advantages is our perspective as a consumer-focused company. We offer a Zero Trust product that can be downloaded to your phone and has been successful on millions of devices. As we collaborate with device manufacturers to integrate our network into their applications, we gain visibility to prioritize performance and end-user experience. This enables us to replace many first generation Zero Trust vendors, who often lack the same user experience and performance. Therefore, in these situations, we’re competing with and often displacing traditional Zero Trust vendors.

That's great. Thank you, Matthew. And then as a follow-up, I think most people assume that Microsoft's new Entrust solutions will be targeted at that SMB sector at the lower end of the market, but that is a market that I think Cloudflare can also serve. So I'm just wondering what you're seeing in terms of competition with the new Entrust solutions.

Yeah. That's my sense of how Microsoft is thinking about this. And in fact, Microsoft's long been a really great partner of ours. And specifically in even their announcements of this, we looked for ways to actually highlight the part of the market that's the SMB market. But we're not satisfied just winning the SMB market. We're winning some of the enterprises. And I think the biggest enterprises that have the most interest here. And so I think Microsoft has been a partner to Cloudflare. We are directly integrated into their Edge browser. The network that we have delivers them very unique benefits, but we also respect them as a competitor. And so I think what we have as an advantage is that network, a network that they use themselves, but it's something where over time, I think their entry into the market has just validated the market, it's defined it, and we look forward to competing in the places that we do and cooperating in any other places where we don't. And what we see is that what customers really want is a network provider that can protect your front door and your back door. What customers really understand is that there's a reason that we have accountants and auditors, and those are separate things. And that what customers really want is they want a solution that works across not just one vendor's products, but the entire IT stack. And I think that's what we see time and time again as the reason why customers are selecting Cloudflare Zero Trust solutions.

That’s great. Thank you very much.

And we will take our final question from Mark Murphy with JPMorgan. Your line is open.

Thank you, and I’ll add my congrats. Matthew, can you elaborate on your vision for how Cloudflare can protect companies from leaking sensitive data out and maybe having some of that land inside a generative AI model. What kind of opportunity do you see? And since that traffic to Microsoft and OpenAI is so tremendous in that area, and you have the partnerships there that you've mentioned, is there a role you can play directly to try to help control some of the data flows in that Azure OpenAI service?

Yeah, I think that this is an area that we really listened to what customers' concerns were and built a product that specifically addresses the concern that you're highlighting. Because it's one that is on the mind of just about every GC and CISO that's out there. And that is that whether they're telling their employers or not, the best data is that almost half of workers that in knowledge industry companies are using AI in one way or another in their jobs. And the risk with AI is that if you send a piece of information up to one of these models, it gets incorporated in the same way that, if your two-year-old hears a bad word, it's really hard to get them to unlearn that thing. And so the key is really making sure that the data doesn't actually leave. And so what we've created is leveraging our existing data loss protection products, the DLP products, and making them specific to tag data in such a way that you can say, here's a piece of information; maybe it's from our marketing website. I am totally fine with that going out to the public, Zero Trust, actually, not Zero Trust, out to this public AI vendors and being able to train on that. In fact, it's great if they're trained on my marketing messages. Here's another piece of information which is much more sensitive, where maybe I'm going to restrict that specifically to my own internal or sandboxed solutions that are the AI training models for that. And then maybe there's something else, that sort of cliché example is the secret formula of Coca-Cola or maybe what all of your internal pay schemes are. Maybe that you don't ever want to get out to anything. And so we've used our DLP solutions in order to specifically address the concern that CISOs and general councils have about data leaking out. And what we think that can do is over time add not only those controls but then start to add things like this particular task has a certain value, so I'm going to send that to maybe GPT 3.5 rather than GPT 4 to save money on that. And so we think that because of the position we're in, because so much of the AI universe relies on us, that it puts us in a great position to not only provide security to the AI companies, but also provide security to anyone who is using AI in their business. And so we can benefit from both sides of the equation.

Thank you. It's extremely helpful.

And we will take our final question from Alex Henderson with Needham. Your line is open.

Thank you. I would like to discuss the significant shift in tone between the first quarter call and the current one, which shows an improvement in nearly every metric you monitor. Could you please highlight the differences across various regions? Did the improvements occur in large enterprises, mid-market, and the SMB sector as well? It would be helpful to have more details on where this change in tone is occurring, especially since you've mentioned that you've seen advancements over the past year. If there is indeed a shift in inflection and tone, I would love to understand its origins.

I'll start and then Thomas will add some insights. Q1 was concerning because we experienced a 20% increase in our sales cycle within a single quarter. Previously, we had seen a similar increase over the entire year of 2022. When we saw the spike in Q1, we were unsure if this trend would continue with another 20% increase in sales cycles for Q2. Fortunately, that did not occur, and the sales cycle normalized. However, I wouldn't say it has returned to a level that makes us feel very optimistic about the overall economic environment. The macroeconomic conditions remain challenging, and the upcoming period will involve significant effort. We did not experience the same dramatic rise in sales cycles that many in the industry, including ourselves, observed in Q1. This is a key factor, from my viewpoint, supporting what you described as a more optimistic tone.

Let me provide some additional insights. We noticed significant forward improvement, and we've discussed this. However, there were numerous inconsistencies in the environment that made it more challenging to forecast what the second half will bring. The Asia-Pacific region was very promising, as previously mentioned, with many large deals coming from there. The European business was largely driven by security and showed encouraging results, likely due to the specific geopolitical situation in Europe. In contrast, the Americas were generally more subdued. We did see strong performance in the Middle East and South America, while North America was somewhat less dynamic overall. There wasn't much to highlight regarding verticals, but the performance was consistent across the board, with no particular vertical standing out for underperformance.

Just want to take a second to thank everyone at Cloudflare. Again, as the macro environment continues to be challenging, I'm proud of how our team has stepped up and executed towards operational excellence. Everyone is working incredibly hard to help live up to our mission of helping build a better internet. So thank you to all the Cloudflare employees, to all of our customers, and to all of the investors and we really appreciate everything that you're doing for us.

And ladies and gentlemen, this concludes today's call, and we thank you for your participation. You may now disconnect.