DigitalOcean Holdings, Inc. Q4 FY2024 Earnings Call

Ladies and gentlemen, thank you for standing by. My name is Krista, and I will be your conference operator today. At this time, I would like to welcome everyone to DigitalOcean's Fourth Quarter 2024 Earnings Conference Call. All lines have been placed on mute to prevent any background noise. After the speaker's remarks, there will be a question-and-answer session. Thank you. I would now like to turn the conference over to Melanie Strate, Head of Investor Relations. Melanie, you may begin.

Thank you, and good morning. Thank you all for joining us today to review DigitalOcean's fourth quarter and full year 2024 financial results. Joining me on the call today are Paddy Srinivasan, our Chief Executive Officer; and Matt Steinfort, our Chief Financial Officer. Before we begin, let me remind you that certain statements made on the call today may be considered forward-looking statements, which reflect management's best judgment based on currently available information. Our actual results may differ materially from those projected in these forward-looking statements, including our financial outlook. I direct your attention to the risk factors contained in our filings with the SEC, including our most recent annual report on Form 10-K filed today, as well as those referenced in today's press release that is posted on our website. DigitalOcean expressly disclaims any obligation or undertaking to release publicly any updates or revisions to any forward-looking statements made today. Additionally, non-GAAP financial measures will be discussed on this conference call, and reconciliations to the most directly comparable GAAP financial measures can be found in today's earnings press release, as well as in our investor presentation that outlines the financial discussion on today's call. A webcast of today's call is also available in the IR section of our website. And with that, I will turn the call over to Paddy.

Thank you, Melanie. Good morning, everyone, and thank you for joining us today as we review our fourth quarter and full year 2024 results. We concluded the year with strong momentum and continue to successfully execute on the initiatives we laid out at the beginning of the year. Our accomplishments included: building out our executive and senior leadership teams; significantly improving the pace of product innovation; augmenting our product-led sales motion with new strategic go-to-market enhancements; and continuing to accelerate the early success of our AI/ML platform, all of which together positioned us with momentum heading into 2025. In my comments today, I will briefly recap our fourth quarter and full year results, reiterate our strategy and priorities, and share several product innovation and customer use cases across both core cloud and AI/ML that demonstrate the progress we're making against our priorities. First, I will briefly summarize our fourth quarter and full year 2024 financial results. Revenue growth accelerated in the fourth quarter to 13% year-over-year to $205 million with one of our biggest growth levers, net dollar retention improving to 99% from 96% in Q4 of the prior year. Our efforts to improve growth and NDR in 2024 are evident in our Q4 results, as NDR with our traditional cloud services reached 100% in Q4 for the first time since June of 2023 on the back of our rapid product roadmap execution and our investments in several strategic go-to-market motions. From these efforts, we saw increased expansion from our higher spend customers as we continue to focus both our product and go-to-market efforts on these top customers. Our higher spend customers, which has traditionally included our Builder and Scaler cohorts, now represent 88% of total revenue and grew 16% year-over-year in Q4. We have now further disaggregated our scalers and are disclosing our highest spent customer cohort Scalers+, which are customers who were at $100,000 plus annual run rate during the quarter. These Scalers+ customers who are critical to our growth trajectory increased in count by 17% year-over-year and were 22% of the total company revenue in Q4. We reached over 500 of these customers for the first time in the company's history. And more importantly, we saw a 37% year-over-year increase in revenue from Scalers+ customers, which is clear evidence of both the wallet share opportunity we have with these customers and our demonstrated ability to scale with them. We also made material progress on our other major growth level, our AI/ML platform, and closed the year with continued momentum, exceeding the three points of overall growth contribution from our AI/ML platform that we had guided for 2024 with Q4 just north of 160% ARR growth, while staying true to our AI strategy and pursuing durable AI revenue. We are very encouraged with the rapid growth and customer adoption of our newly launched AI products, and I'll talk about them later in my comments. On top of the increasing growth signals, profitability remains strong as we delivered healthy 42% adjusted EBITDA margins, both Q4 and for the full year, maintaining our cost discipline while we continue to invest to fuel future growth. Looking forward, our 2025 guidance reflects this ongoing momentum, but full year revenue growth at low to mid-teens and high-teens free cash flow margins above our preliminary 15% to 17% indication. We continue to prioritize and rebalance our investments driving improved operational efficiencies while shifting resources towards our top growth initiatives. Our upcoming Atlanta data center is a good example of both these priorities, as the upfront investment in that facility, which will come online in Q1, not only provides us with incremental capacity for both AI and our core cloud offerings, but also gives us a lower cost facility and it's part of our longer-term data center optimization strategy. Matt will walk you through more details on our financial results and guidance later in the call. In my first year at DigitalOcean, we had several very clear priorities as we sought to accelerate growth. We needed to double down on product innovation to address key gaps in our core cloud platform, better address the needs of our larger customers, return net dollar retention to a tailwind rather than a headwind, and build the foundation for our longer-term AI growth strategy. We made material progress on each of these objectives and continue to deliver on our promise of making complex cloud and AI technologies simple. We also made substantial progress on making our platform even more scalable, enhancing our ability to meet the needs of larger customers. And finally, we doubled down on our heritage of being the most approachable public cloud provider by continuing to invest in support of open-source AI models and even hosting our developer conference Deploy in January. Let me now give you some updates on the core cloud computing platform. In Q4, we continued to accelerate the pace of innovation as we released 49 new products and features throughout the quarter, which is more than four times what we released in Q4 of the previous year. Most of these products and feature enhancements directly address the needs of our larger spend customers as we continue to remove blockers and implement the capabilities that they need to further scale on our platform. I will highlight several of these product releases that we have made to help our customers grow their businesses using DigitalOcean. Given that our larger customers run significant global workloads, they need the ability to securely connect different parts of their networks so that their systems and applications in separate environments, in various data centers in different countries can securely communicate without using the public internet to improve speed and efficiency while keeping their data secure. To address this need, in Q4, we announced Virtual Private Cloud peering, or VPC peering for short, which is now generally available for all our customers. VPC peering enables customers to connect their separate private clouds and establish seamless communication between resources hosted in these clouds using private IP addresses, keeping their information safe by traversing through the DigitalOcean backbone rather than through the public internet. Our larger customers also need the ability to distribute traffic across resources, while still keeping it within a secure private network. To support this, we introduced a new feature called internal load balancer, which enhances security by ensuring that internal workloads remain isolated from the public internet, making it ideal for applications that require highly scalable private communications. We also have several large customers with volatile traffic patterns that need mechanisms to handle these massive spikes in volume very smoothly, while still optimizing costs and scaling them down when the demand is lower. To address this, we announced the general availability of droplet auto-scale pools to ensure that the right resources are available to handle application workloads during these surges in traffic, scaling up automatically to meet demand, while also helping minimize costs by scaling them back down when the traffic surge is over. We also introduced flexible management capabilities to our App Platform, which is our platform as a service offering for more granular lifecycle management, including archive and restore functionality and maintenance mode during the application's full lifecycle. Next, customers of Spaces, which is our fast-growing S3 compatible object storage service, have long asked for the ability to grant granular permissions to different users or teams without exposing full account-wide credentials. In response, we launched Per-Bucket Access Keys for Spaces. This highly sought-after feature provides customers with identity-based bucket-level control over access permissions, helping enhance their data security and ultimately simplifying management overhead. Complementing this accelerated pace of product delivery of sophisticated capabilities was one of our new go-to-market motions, where we bolstered our engagement with our top 1,500 customers. By helping take these new innovations to our customers and tightly orchestrating a closed loop between the various DigitalOcean teams and our top customers, this motion increased awareness and adoption of our new product capabilities, facilitating the migration of cloud workloads from other clouds to DigitalOcean and serving as a catalyst for both our improved NDR and faster growth rate of Scalers+ customers. Our higher spend customers have quickly started adopting many of these features that I just talked about that we released over the back half of 2024. Over 50% of our top 100 customers have adopted at least one of the features that we released in Q3, and we anticipate similar adoption levels for our Q4 features over time. Together, the breadth of these new features and the pace at which we are executing our product roadmap is enabling our higher spend customers to grow on DigitalOcean and is enabling us to win more of their workloads that today reside on other hyperscaler clouds. As we discussed last quarter, we've been focused on helping our customers seamlessly migrate more workloads to us and scale efficiently on DigitalOcean. One example of this is a customer called Digital Platform, a strategic software solutions development company that was experiencing high cost and latency issues with the cloud they were leveraging, which was impacting their application performance. Through our customer-facing teams, we were able to fully migrate their workloads to DigitalOcean, leveraging our optimized database infrastructure to improve their performance while providing them with substantial cost savings. Another example is Hudu, a provider of enhanced IT documentation with features and tools made for assisting managed service providers in IT departments. Hudu has been a DigitalOcean customer since 2019, and they continue to scale and grow on our platform. Hudu was an early adopter of our Kubernetes, managed databases, SnapShooter, and premium support products. As a result of the ease of use of our platform, they've been able to focus on their own scalability and have grown as a business over 870% since 2021. Another example of our customer's ability to scale with DigitalOcean is Moments, a fitness and wellness platform that manages bookings, communications, and memberships. Moments needed a larger instance to house their database to meet the requirements of their rapidly growing customer base and continue leveraging our platform. The DigitalOcean team was able to provide architectural guidance by crafting a solution with existing DigitalOcean products while delivering a new product they requested, a 48 vCPU storage optimized droplet with scalable storage. Let me now switch gears and give you a quick update on our AI initiatives. We remain committed to and are executing well against our AI strategy that we articulated last year. In that context, I'm very encouraged by the emerging innovations in this space, like DeepSeek, that drive down the cost of AI adoption and improve the quality of open source models, which will ultimately enable more customers to use AI. We see innovations such as DeepSeek and even reports of some hyperscalers potentially moderating their data center commitments as reinforcing our conviction that while a lot of action to date has been at the infrastructure layer, that innovation and value creation will occur at the platform and application layers, where we are highly differentiated and well positioned to grow as we democratize AI for our customers. Our prudent approach to AI investment also allows us to ramp up investment where we see customer demand. And as a case in point, we are increasing our allocation of our GPU capacity for our GPU droplets, where we quickly ran out of capacity after launching at the beginning of Q4. Each of these three layers, infrastructure, platform, and AI applications have their purpose and very distinct customer targets. And although most of the action is still in the infrastructure layer, we are now starting to see more narratives in the market around the higher layers of the stack in platforms and agentic applications, which is a good validation of our AI strategy that we laid out last year. We've been making excellent progress, enhancing our AI infrastructure offerings, as well as innovating at the GenAI platform and agentic application layers, as we build towards the goal of democratizing AI, by enabling our customers to quickly experiment and build AI into their real-world applications. On the infrastructure side, we are seeing strong adoption of GPU droplets, which we made generally available to all our customers in October. As a reminder, GPU droplets allow DigitalOcean customers to leverage on-demand and fractional access to GPUs in a self-service way in just a few minutes, vastly simplifying a very complex series of steps. Let me now give you a couple of real world examples. Prodia, a company specializing in integrating generative AI into their own applications, leverages DigitalOcean's GPU infrastructure globally to efficiently manage their own products. Prodia accelerates generation speeds, offering an easy-to-use API for AI-powered image generation. Another example of an AI/ML infrastructure company is Commodity Weather Group, a company that provides advanced weather model forecasts to their clients and runs AI-based weather models to enhance decision-making with additional insights. They also leverage DigitalOcean's AI infrastructure for its scalability and ease of use. These are just a few examples of how our customers are leveraging our infrastructure to develop and sustain data-intensive software to be able to meet the needs of their own customers, all while leveraging the simplicity of DigitalOcean's AI/ML infrastructure. Moving up the stack, I'm very excited about our GenAI platform, which is now in public beta. The DigitalOcean GenAI platform is one of the simplest platforms to create, deploy, and integrate AI agents into real-world applications. Stepping back for a second, AI agents are software applications designed to autonomously perform multi-step tasks that involve reasoning and decision-making leveraging AI and ML. Our new GenAI platform gives customers everything they need to build AI into their own applications without the need for advanced expertise in AI or machine learning. Customers can easily and quickly build AI agents leveraging DigitalOcean's infrastructure by adding their data to their pre-trained third-party GenAI models and can seamlessly integrate those agents into their own application via secure endpoints or chatbot plugins. In the four weeks since we made the GenAI platform beta public, we have seen well over 1,000 agents created on the platform, with the most encouraging fact being that roughly 90% of these agents were created by existing DigitalOcean customers, which is further validation of our belief that our typical DigitalOcean customer wants to leverage AI into their software stack and are willing to do it if you make it very simple and integrate it with the rest of our cloud platform. In our latest version of the DigitalOcean trends report that we published earlier this month, we found that almost 80% of our target customers are interested in leveraging AI, but over 70% of them said that cost and lack of expertise are the two major impediments to AI adoption. Our GenAI platform makes it very simple by abstracting out most of this complexity associated with creating AI agents by having templatized agents, click-through wizards, and so on, and by providing easy access to a variety of open-source models, including Lama, DeepSeek, and Mistral. At the application layer, we introduced cloud-based co-pilot in public beta, which is a suite of AI solutions designed to bring intelligent managed hosting to small and medium businesses. Starting with AI-powered diagnostics to give customers recommendations and alerts to fix issues before they become problems. This helps our customers automate tasks, monitor performance, and provide them with insights to keep their websites up and running smoothly. One such example is ClickIT, a web design and development agency which is already leveraging the newly announced cloud-based copilot in AI. ClickIT is finding a four times reduction in the time spent manually handling issues and taking care of web servers. We also started using GenAI agents for our own internal DigitalOcean cloud operations. For a variety of operational incidents, GenAI agents are invoked, which analyze our service logs to determine the root causes. This has resulted in a 39% improvement in our time to resolution, and is one of the most sophisticated uses of GenAI agents in the industry today. We're building these agents and using them, not just to improve our own operations, but also to deeply understand the pain points and complexities of building AI agents so that we can incorporate these learnings into our GenAI platform and make it even simpler for our customers to use. Beyond our product and customer progress, another recent highlight was the Deploy Conference we hosted in January in Austin, Texas. This event brought together customers, partners, and some DigitalOcean employees, amplifying our presence in the developer and AI/ML space and building out a strong position as the most approachable public cloud. At Deploy, we introduced a slew of new product capabilities, launched an AI variant of our popular startup incubator program called Hatch, and hosted many of our technology and channel partners. At Deploy, we also launched a new migration program designed to seamlessly transition cloud workloads from the hyperscalers to DigitalOcean. This program will eliminate migration-related complexity, deliver lower operational costs, and provide seamless technology assistance through our partner ecosystem and a newly formed DigitalOcean team of solution architects skilled at migrating cloud workloads. To recap, as I close my prepared remarks, we entered 2025 with increasing momentum. In Q4 alone, we released more than four times as many products and features as we did in the previous year, increased net dollar retention to 99%, grew revenue 13% year-over-year, and delivered 18% adjusted free cash flow margin. Our focused efforts on our higher spend customers and our continued traction in AI drove quarterly revenue for our top 500 plus customers, representing 22% of our total revenue to grow at 37% year-over-year. This shows clear progress on our strategy and builds on our leading position as the simple, scalable, and approachable cloud. Before I turn the call over to Matt, I'm very excited for our upcoming Investor Day, which we will be hosting at the New York Stock Exchange on April 4th, starting at 9 A.M. Eastern time. During this Investor Day, we will share more on our longer-term strategy, including more details on our progress and key metrics, and we'll share a view of our long-term financial outlook.

Thanks, Paddy. Good morning, everyone. Thanks for joining us today. As Paddy discussed, we delivered a solid Q4 and full year 2024 on all financial metrics and made meaningful progress on the key initiatives and goals we set in place at the beginning of the year. In my comments, I will review our Q4 results in detail and cover the full year 2024 financial highlights before sharing updated first quarter and full year 2025 financial outlook. Revenue in the fourth quarter was $205 million, up 13% year-over-year. Annual run rate revenue, or ARR, in the fourth quarter was $820 million, as we added $26 million of incremental ARR in the quarter, up from $24 million in new ARR in Q3. Of note, we made a methodology change to how we report ARR, where we now calculate ARR by multiplying the quarterly revenue times 4, rather than the final month of the quarter times 12. We decided to use aggregate quarterly revenue to calculate ARR to reduce potential volatility in this metric given the project-based nature of some of the training workloads customers are running on our AI/ML platform. More details on this as well as a reconciliation between the old and new methodologies can be found in our Form 10-K. Revenue from our builders and scalers, which are our highest spending customer cohorts and represent 88% of total revenue, grew 16% year-over-year, and customer count increased 6% year-over-year. This quarter, we also began to disclose further disaggregation within our Scalers cohort, and are now separately disclosing our largest spending customer cohort, or Scalers+. Or customers that monthly spend is more than $8,333 per month during the quarter, which is more than $100,000 on an annualized run rate or ARR basis. In Q4, revenue from Scalers+, who represent 22% of overall revenue, grew 37% year-over-year. Driven by a 17% year-over-year increase in Scalers+ customer account, coupled with their expanded usage of our core cloud services and continued growth of our AI-related solutions. The net expansion increase we saw from our top customers drove our Q4 net dollar retention rate up to 99%, up from 97% for the first three quarters of 2024. The increases we are seeing in our net expansion levels, coupled with churn that has remained stable for the last two years, is moving us closer to reaching and exceeding 100% NDR and shifting NDR from a growth headwind to a tailwind. Within this overall improvement NDR, we saw the NDR rate of our traditional cloud services reach 100% in the first quarter for the first time since June of 2023. All of this progress is despite the fact that we are still lapping a few headwinds from our managed hosting products that we've spoken about previously. And this progress gives us confidence in our baseline growth rate heading into 2025. Turning to the P&L, gross margin for the quarter was 62%, which was 300 basis points higher than the prior quarter and 500 basis points higher than the prior year. The increase in gross margin quarter-over-quarter and year-over-year is primarily driven by the increase in revenue as well as the change in useful life for our servers from five to six years as we've been able to extend the utilization of our equipment. More details on this change in useful life can be found in our Form 10-K. Adjusted EBITDA was $86 million, an increase of 17% year-over-year. Adjusted EBITDA margin was 42% in the fourth quarter, approximately 200 basis points less than the prior quarter, but 100 basis points higher than the prior year. We feel confident in our ongoing ability to appropriately balance our growth investments with our efforts to improve operating efficiency, as is evidenced by our continued delivery of healthy adjusted EBITDA margins. Q4 adjusted free cash flow was $37 million, or 18% of revenue. This is higher than Q3 by approximately 500 basis points due to timing of capital investment payments, which will continue to create quarter-to-quarter variations in adjusted free cash flow margins. Finally, non-GAAP diluted net income per share was $0.49, an 11% increase year-over-year. This increase is a direct result of our ability to increase our per-share profitability levels by continuing to drive operating leverage while mitigating dilution through share buybacks. For 2024, total revenue increased 13% year-over-year to $781 million. This growth came primarily from continued strength in our durable customer acquisition engine, which in 2024 was augmented by new customer revenue on our AI platform and from growth in our highest spend customer cohorts. Given our 98% NDR in 2024, the vast majority of our revenue growth in 2024 came from customer acquisition and new customers, including those that consumed our newly launched AI services, which delivered north of 160% ARR growth in Q4. Despite the NDR headwind in 2024, our Scalar and Scalar+ customer cohorts collectively contributed approximately $445 million and 57% of total revenue in 2024 and grew revenue 18% year-over-year. Turning to the P&L, our gross margin for the year was 60%, which was 300 basis points higher than the prior year, which I noted earlier in my comments is primarily driven by revenue growth that is faster than our growth in COGS as we delivered further operating leverage. On the profitability front, we delivered healthy adjusted EBITDA margins in 2024 at 42%, up 200 basis points from 2023. We generated a 17% adjusted free cash flow margin in 2024, down as we had planned going into the year, 500 basis points from 2023 margins as we ramped our investment in our compelling AI growth opportunity. On our customer metrics, as I mentioned previously, this quarter we are now disclosing further disaggregation of our Scalar customer cohort into Scalars and Scalars+ as these higher spend customers are the focus of our investments and their performance is key to our growth strategy. In Q4, the number of Scalars+ grew 17% year-over-year and the revenue growth from Scalars+ grew 37% year-over-year. We are also seeing an increase in average spend within Scalars+, where average revenue per user, or ARPU, grew 18% year-over-year. The traction we are seeing with our highest spending customers is the result of our continued efforts across product development, targeted customer success and account management motions, and new go-to-market investments that are all focused on these customers, and this is an encouraging sign of our ability to drive future growth. Our balance sheet remains very strong as we ended the quarter with $428 million of cash and cash equivalents. We continued to execute against our share repurchase program in the quarter with $28 million of repurchases in Q4, bringing total share repurchases to $57 million in fiscal year 2024 and bringing our cumulative share repurchases since IPO to $1.5 billion and 32.6 million shares through December 31st, 2024. With our healthy cash position and ongoing free cash flow generation, we are well positioned to continue investment in both organic growth and share repurchases while maintaining the appropriate flexibility to address our 2026 convertible note at the appropriate time, which is likely before it goes current at the end of this year. Moving on to guidance, I will now share our financial outlook for the first quarter of 2025 and for the full year. For the first quarter of 2025, we expect revenue to be in the range of $207 million to $209 million, representing approximately 13% year-over-year growth at the midpoint of our guidance range. For the full year 2025, we expect revenue to be in the range of $870 million to $890 million, also representing approximately 13% at the midpoint of our range. As was the case with our 2024 guidance, our 2025 guidance is underpinned by our baseline growth foundation as we entered the new year. Our primary 2025 growth levers are bolstering customer acquisition and continuing to drive customer expansion. Customer acquisition includes revenue from any new customer, including new AI/ML customers who are within their first 12 months on our platform. We anticipate customer acquisition and revenue from new customers to again contribute the majority of our growth in 2025. But by continuing to improve NDR in 2025 and by expanding usage by our existing AI customers that will soon have been on our platform for more than a year, we expect customer expansion to improve and to have a neutral to slightly positive impact in 2025 where it was a headwind in 2024. For the first quarter of 2025, we expect our adjusted EBITDA margins to be in the range of 38% to 40%. For the full year, we expect adjusted EBITDA margins to be in the range of 38% to 40%. For the full year, we expect non-GAAP diluted earnings per share to be $0.41 to $0.46, based on approximately 103 million to 104 million in weighted average fully diluted shares outstanding. For the full year 2025, we expect non-GAAP diluted earnings per share to be $1.85 to $1.95, based on approximately 104 million to 105 million in weighted average fully diluted shares outstanding. On adjusted free cash flow, we expect adjusted free cash flow margins for the full year to be in the range of 16% to 18%, slightly ahead of the preliminary indication for 2025 that we provided last quarter. Consistent with our historical guidance practice, we are not providing adjusted free cash flow guidance on a quarter-by-quarter basis, given it is heavily influenced by working capital time. However, we would like to highlight that our 2025 expenditures will be front-end loaded, which is driven by additional AI-related capital expense as we scale those services, one-time startup costs to get our Atlanta data center online and the usual Q1 cash expense events including bonus payments and higher payroll taxes. Given this front-end loaded investment and expense, Q1 adjusted free cash flow margins will decline from Q4 levels, but we will quickly return to higher adjusted cash flow margins in Q2 and across the balance of the year. And we remain very confident in our ability to deliver the 16% to 18% full year adjusted free cash flow margin in our guide. That concludes our prepared remarks and we will now open the call to Q&A.

Thank you. We will now begin the question-and-answer session. And your first question comes from Josh Baer with Morgan Stanley. Please go ahead.

Great. Thanks for the question. At your recent Deploy Conference, you talked about several customers that were disappointed with their experience with the hyperscalers and ended up migrating to DigitalOcean's platform. I was hoping you could expand on that and touch on what types of customers you're targeting, what types of workloads? And then I have a follow-up.

Yes, thank you, Josh, for the question. So, yes, what we highlighted during the conference was a handful of customers that are looking at alternatives primarily for two reasons. One is the simplicity of running a fairly sophisticated workload on a hyperscaler. It's not an easy undertaking. You need specialists for many of the nuances like storage and networking and compute and so forth and also the total cost of ownership, especially if you have spiky workloads and you're unwilling or unable to commit to really long-term contracts at a substantial amount of minimum commitment, then it becomes a really onerous thing to keep running your workloads on some of these hyperscaler clouds. So what we are offering as part of the migrations program that I was just talking about is a couple of things. One is the ability to get our partner ecosystem involved to facilitate a smooth transition of the workloads, but more importantly, provide a super compelling, scalable platform which is far easier for most of the mainstream workloads to run and operate on the solution platform. So we're seeing all kinds of customers, Josh, and we are staying true to our target customer segments, which is tech native, digital native cloud application software companies that are running globally distributed workloads that are typically network and bandwidth intensive. Some of them require very bursty spikes in their traffic patterns, so that need to be supported in a very elastic manner. And what attracts them to us is our core value proposition of being simple yet scalable, but most importantly, a very approachable cloud that really cares about them.

Thanks, Paddy. And just wondering on the EBITDA guidance, I mean, initial EBITDA guidance for this year was 36% to 38%, you ended at 42%. I was hoping you could kind of provide a just high level of the main drivers of that degree of outperformance and then how we should think about that level of conservatism in the initial 2025 guide for EBITDA? Thanks.

Great question, Josh. We mentioned last quarter that with the arrival of new executives and our need to speed up the product roadmap, we created some flexibility in Q4 for the R&D team to bring in contractors as needed. We thoroughly reviewed our spending and successfully redirected resources, with Bratin and his team doing an excellent job of prioritizing key initiatives and delivering important products. As a result, we didn't require that extra surge. You'll notice that our full-year guidance, as well as for the first quarter, has a broad range for EBITDA. What we're indicating is that you shouldn't get overly concerned with the fluctuations in EBITDA from one quarter to the next, as we might bring expenses forward or identify areas where we can save costs. However, I want to emphasize our commitment; we've increased our free cash flow guidance from an initial 15% to 17% to 16% to 18%. This is what we are focusing on more intently. I expect EBITDA will vary a bit more, but we are dedicated to improving our gross margin, increasing operating efficiencies, and enhancing the leverage we have in the business. At the same time, if we identify an opportunity to enhance product capabilities that can boost revenue, we will pursue it, which may temporarily affect EBITDA margins. This is why we've provided a broad guidance range for the fourth quarter and for 2025.

Your next question comes from the line of Gabriela Borges with Goldman Sachs. Please go ahead.

Hey, good morning. Great to see the NDR numbers. Thanks for taking my question. Paddy and Matt, I wanted to follow-up on the math you've given us in the prior quarter on how much ARR you're able to capture per dollar of GPU-related CapEx. Maybe just refresh us as you move up the stack from IaaS to PaaS, are you able to generate more revenues per dollar of CapEx? And Matt, maybe you can comment on the gross margin profile of the AI services business as you move into more differentiated services? Thank you.

Yes, both are excellent questions, Gabriela. What we've discovered, particularly with our GenAI product, has been an insightful experience and strongly supports our strategy. When someone comes in wanting to create a chatbot, they bring their knowledge base and aim to connect with another model. They can definitely utilize our GenAI capabilities, which have significantly higher margins compared to traditional bare metal or infrastructure layers. However, what stands out the most to us is the additional revenue this will generate from cloud services. Each chatbot requires storage for the knowledge base, bandwidth for communication with models, and a variety of our other database infrastructure. Therefore, we believe the incremental revenue from the GenAI services will far exceed the actual revenue from GenAI alone. That’s a very compelling aspect. As we've indicated and is apparent in the market, margins on the infrastructure layer, particularly for core GPU services, are not exceptional. Pricing is quite transparent in the industry, leading to considerable competition for initial workloads. Despite new entrants like AMD and others enhancing their capabilities alongside NVIDIA, the industry is still largely dominated by one vendor, and costs remain very high. We anticipate these costs will decrease over time, allowing us to leverage more infrastructure for inferencing, which will result in more consistent and higher-margin revenue. We believe we are moving toward a revenue model that increasingly relies on higher-level services, which not only possess better margins on their own but also lead to greater high-margin cloud revenue.

Thank you.

Your next question comes from the line of Mike Cikos with Needham & Company. Please go ahead.

Great. Thanks for taking the questions guys. Great to see the improvement in the NDR that you're talking to as well. I guess first question I had for you was related to the AI/ML. Great to see the growth there still remaining triple digits north of 160%. Is there any way to give us a little bit more as far as the size of that AI/ML ARR base today, what it represents as a percentage of the total ARR? And then the second piece on that point would be, how do we think about the ARR composition today? Is the vast majority of that coming from the Scalar+ cohort? I know we have these new disclosures and you guys are seeing broad-based adoption, but just interested where the revenue dollars specifically are coming from for that piece. Thank you.

Thank you for your questions, Mike. We're not disclosing the specific annual recurring revenue for AI for specific reasons. The revenue is intertwined with other parts of the business. When we receive some revenue from GenAI, it often leads to additional revenue in different areas. We don't separate our business into distinct AI product groups and other product groups; instead, we view AI as one of our products among many services like infrastructure as a service and platform as a service. When we acquired Paperspace, we gained about 15,000 customers, primarily small ones that resemble our DigitalOcean customer base. Many of these customers use the AI platform and the legacy Paperspace notebook capabilities. Most of the new customers adopting the GenAI product are existing customers, and their profile is similar to our current customer mix. While there is a significant portion of AI revenue in Scalars+, it's not the entire amount, and the majority is outside of that for now, but there is a notable presence.

Thank you for that. Can I just tack on maybe one more on the gross margin? I know that we have the Atlanta data center coming online in Q1, which we're all looking forward to, but can you help us think about how gross margins are likely to move through the course of the year with that Atlanta data center? And then we obviously have the news that the useful server lives are being extended a year. How do we put those two pieces together when we think about how calendar 2025 looks? Thank you.

It's not expected to change significantly. There may be a slight decrease in gross margin at the start of the year, but it should rebound, similar to what we observed previously with Sydney. Initially, there are upfront costs that aren’t yet matched by revenue. Thus, there will be a minor dip in gross margin, but we don’t anticipate a major shift in our current gross margin range. Moreover, Bratin and the teams are actively working to improve gross margins over the long term as we continue to optimize our data centers and seek better ways to utilize our existing infrastructure. Therefore, while there might be a small drop in gross margin at the beginning of the year, it will recover. I want to emphasize that we are confident in achieving a free cash flow margin of 16% to 18% for the year. Ultimately, everything should balance out to yield improved margins as the year progresses.

Your next question comes from the line of Patrick Walravens with JMP Securities. Please go ahead.

Oh, great. Thank you. So Paddy, you've been here a year, right? What has gone better than you thought it would and what has proved a little more difficult?

Hello, Patrick. Good morning and thank you for your question. It's quite early for you, and I appreciate you joining us. Regarding what has gone better than I expected, I would say our product innovation and our understanding of customer needs have exceeded my expectations. The adoption and positive reception from our customers indicate they truly want to remain with us. We are seeing early signs that if we take care of them, they will expand their relationship with us, which is fantastic. The AI landscape is evolving in intriguing ways, and I'm very encouraged by our commitment to our AI strategy. We're not getting sidetracked by the rush to deploy excess capacity or chase the GPU-based business model. This has been a significant learning experience, and gradually, some of our assumptions in the AI space are proving accurate. The two main barriers to adoption, as I mentioned earlier, are complexity and cost. We can simplify the complexity for our customers, especially by using open-source models, which provide us more freedom to ease the process. Additionally, we aim to make our offerings more affordable. We've begun to implement on-demand fractional access to GPUs and token-based serverless endpoints within our GenAI platform, particularly utilizing the open-source model, which is also gaining traction. Those have been the two key insights for me over the past year.

Your next question comes from the line of Wamsi Mohan with Bank of America. Please go ahead.

Hi, thanks for taking my questions. It's filling in for Wamsi today. I have two. First one for Paddy. You launched a lot of products in 2024. Paddy, can you talk about the areas of investment in 2025? And thanks for disaggregating the Scalers+ at 22% of revenue. How do you see that percent, that cohort growing in 2025? And to do that, are you happy with the go-to-market investments you've made, or do you need to hire more salespeople? And I have a follow-up for Matt.

Thanks for the question. I’ll respond quickly. Regarding our product focuses this year, I’ll divide it into core cloud and AI. In core cloud, we aim to address the needs of our larger customers, focusing on enhanced management, security, networking, and a wider range of specialized droplets. We still have work to do to support these customers better and facilitate the migration of larger workloads to DigitalOcean from other cloud services. On the AI front, we will strengthen our infrastructure, which is currently robust, scalable, and well-regarded by our users. The GenAI platform will drive significant innovations, making it simpler for software companies to integrate agents into their applications. Recently, we launched the site reliability engineering copilot, which started as a cloud-based offering and will be available more broadly. Our goal is to develop more AI agents that address concrete business challenges rather than focusing solely on research. As for our go-to-market strategy, we have enhanced our customer engagement efforts, and I will provide updates on our go-to-market initiatives as the year progresses. We have a substantial customer base, allowing us to promote our expanding product platform and capture a larger share of their spending. While acquiring new customers through our efficient self-service model remains important, much of our efforts will concentrate on increasing spending from existing customers, which is a more effective strategy than solely pursuing new logos. We will pursue both strategies and I will keep you updated on our progress throughout the year.

Thanks for all the details there, Paddy. Matt, just a quick follow-up. On payables, it looks like sequentially it was up meaningfully. Is that because of the Atlanta data center investments, or do you get a new contract that is driving that up? And what would ARR be under the old method in fiscal 4Q? Thanks for taking my questions.

From a payable standpoint, we are preparing for Q1 for the Atlanta data center. Bratin and the team did an excellent job accelerating the readiness of some services, which allowed us to purchase more equipment earlier than anticipated. We managed to stay within the limits of our free cash flow. The goal was to ensure Atlanta was well-prepared for a strong start. Regarding ARR, I don't have the specific numbers handy, but they can be found in the reconciliations in the 10-K. It would have been higher in the fourth quarter if we had reported the December figures over 12 months without changing our approach. It's important to remember that we operate on a consumption-based model, including our AI services, which means we don't have committed contracts. Therefore, consumption varies over time, leading to oscillations in our metrics. Averaging over the quarter provides a more consistent metric, and while it made our ARR appear lower, we believe it is the right approach for the market by providing a more stable and less volatile metric.

Your next question comes from the line of Mark Zhang with Citigroup. Please go ahead.

Good morning, team. Thank you for addressing our questions. I appreciate the extra discussion on the Scalars. How should we understand the differences in upsell and cross-sell rates between Scalars+ and non-plus going forward? Why are we observing a faster growth rate in the non-plus Scalars cohort, considering the product and go-to-market strategies you've implemented? What factors could enable this group to outpace the growth of builders? Thank you.

The reason we separated the Scalers+ and the customers contributing over $100,000 is that the most frequent question we received from analysts and investors was whether we have a graduation issue. Specifically, whether our largest customers are leaving us—these are our most valuable customers, and we don’t want to have a situation where we’re losing the most significant part of our customer base. There was some truth to this concern, as we indeed faced challenges with these customers in recent years, not meeting their needs as effectively as we could have. This situation contributed to a decline in our Net Dollar Retention, taking us below 100. Since the arrival of Paddy and Bratin, combined with our strong focus on product innovation aimed at these large customers, we have made significant improvements. While a large customer for us may not be a key account for hyperscalers, they are still essential to our business. We've committed to focusing on this group, ensuring we listen to them, and rapidly providing the capabilities they need to grow on our platform. The feedback has been very positive, and we are now discussing migrations for these customers to bring their workloads back to us or to us for the first time from the hyperscalers. This has resulted in a significant turnaround. We will discuss this in more detail at our Investor Day in April, but if you look at the improvements in Net Dollar Retention and year-over-year revenue growth, there's been a remarkable change among our top customers. As Paddy and I have mentioned, we believe there's a substantial opportunity to increase our wallet share with our customers. While you've noted that the Scalers below the $100,000 threshold are not growing as quickly, we are still addressing this. We are attempting to identify the customers in that group who are most likely to spend and where we currently have the lowest wallet share, so we can pursue them effectively. What you're seeing now is challenging work that requires considerable effort as these are important customers, but we are prioritizing our largest customers first. We are working through the other layers as we move through the year.

And the only thing I'll add is, the Scalers and Scalers+, they all look the same. These are very similar looking customers. So that's the best news because now we can go and get and start working on the Scalers, as Matt just talked about, and get as many of them graduated into Scalers+ and keep expanding their share of wallet. And as we've been maintaining, Mark, there is a tremendous wallet share opportunity because they're all running substantial workloads, just not on us. So that's a great opportunity for us to go earn the trust of these customers and get more of their workloads.

Your next question comes from the line of Raimo Lenschow with Barclays. Please go ahead.

Hey, thanks for squeezing me in. One simple one at the end. What do you see out in the market in terms of the end amount, et cetera, and how is that helping you for this year? Thank you.

I'm sorry, could you repeat the question, I didn't catch?

Oh, I was just asking for what are you seeing in terms of end demand? If I look at the small business index, et cetera, that all starts looking better. Does that help you? Do you see help coming there from that one or is that all neutral for you? Thank you.

Yeah, so Raimo, thanks and nice to hear from you. From an end-user demand perspective, we have not modeled any major variation in our guidance or our plans. So we are expecting it to be stable and just like how it has been over the last few quarters. And from our NDR improvement perspective, I mean, obviously there's a lot that we have done to control our own destiny and earn the right to keep our customers and keep them expanding on our platform. But hopefully the macro and the end-user demand stays neutral to even positive, but we have not baked any of that into our projections.

And ladies and gentlemen, that does conclude our question-and-answer session. And with that, that does conclude today's conference call. Thank you for your participation and you may now disconnect.