Excellent. Let's get started. My name is Josh Baer, Software Analyst here at Morgan Stanley. We have the DigitalOcean team. Before we do introductions and get started, some disclosures. For important disclosures, please see the Morgan Stanley Research Disclosure website at www.morganstanley.com/researchdisclosures. If you have any questions, please reach out to your Morgan Stanley sales representatives. So we have CEO of DigitalOcean, Paddy Srinivasan, and Matt Steinfort, CFO. Thank you so much for joining us. Really appreciate it. Thanks, Josh. Wanted to open up, Paddy, with a question about your strategic priorities. You came in about a year ago, roughly, and you've made a lot of change. Could you kind of review what some of those changes and where you've been focused on this past year and how you're thinking about your strategy into 2025? Sure. We'll be happy to do that. Thank you, Josh. First of all, thank you for hosting us here at the fireside chat. Great to be back here at this conference, so yes, I've been here for about 12 months now, and when I stepped in, my biggest priority was to look at the core business first, so when I looked at the core business, one of the biggest things we noticed was that we, as a high-volume, high-transaction-oriented business, our NDR or NRR was stuck at 96%. And we came to the realization that for us to reaccelerate growth, that's one headwind we need to turn around and make it into a tailwind for us, and for that, there were two things we had to do. One is look at the product that we are, the cloud platform that we are building, and ensure that that really addresses the needs of our customers, especially the larger customers, which is where we saw a lot of our pain points experienced by our customers, and the second one was to have a complementary go-to-market motion to ensure that all the good stuff that we are shipping is getting into the hands of the customer, so we did that, and you can see that in our results over the last couple of quarters. The product velocity has picked up significantly, and not only that, our product strategy has evolved into meeting our customers where they are, especially the larger ones. For those of you that are new to our story, we define our larger customers as scalers, which are about 18,000 of them, spending on an average $25,000 per year. And that's where a lot of our NRR issues were, which meant that we really had to go deep into understanding exactly what the gaps were in the product. And over the last two or three quarters, we have started addressing those gaps in our product portfolio, whether it is manageability, security, high-bandwidth networking, and those kinds of things have been a big focus of ours. And the second one was building a complementary go-to-market motion. That was a big focus of mine. And the second, so that was in the core cloud business, and the second one was defining our go-forward AI strategy, which we've been talking about for the last couple of quarters, and we are showing robust momentum in our AI business as well. So that's what has kept us very busy over the last 12 months. Perfect. Great overview, and we'll dig into all of that. Matt, do you want to pull you into the conversation and just touch on your reported earnings last week? I want to ask about the demand environment and some of your results that showed improving trends following a lot of these strategic priority changes that have come. Where can we see some of the results of the improvement show up in your numbers? Yeah, I think the first and most prominent area was in NDR, as Paddy said. We saw NDR improve from 97 in the prior three quarters to 99. And we immediately looked at that and said, "Okay, is that something we did, or is that something that just we're lucky the market kind of improved?" And we unpacked that. And I'd say about a third of that improvement was, "Hey, we just had better comps," and so that was good, but it's just math. But two-thirds of it was because the customers were growing faster. And if you take that two-thirds and we could further unpack it, we said that we were able to identify that half of that, so one-third of the total improvement, was from customers that we'd either touched with one of our new go-to-market motions that Paddy had implemented, or it was a customer that had adopted one of the new features that we had pushed out. So a third of the NDR improvement was from actions that we had taken, we believe, in the last kind of nine months or so, which is a really positive signal because it tells you that we are working on the right products, the customers appreciate those, and they're able to grow. And it was pronounced in the larger customers. So those larger customers, the Scalers Plus that Paddy described, saw the most prominent increase in NDR. They've grown 37% year over year, which is very, very strong. And then if you take the last third of that improvement, I think that's where we're assuming it's just more macro improvement that we're seeing because we couldn't identify anything that we had done in particular to touch those customers. So that gives us a lot of encouragement going into this year where last year NDR was at 98 for the whole year, and this year we think we'll get to at least neutral and hopefully slightly positive on NDR for the full year. And we're also seeing increases in the top of the funnel. So our customers, our new customer revenue, that's any revenue from a customer in their first 12 months, we're starting to see signs of that picking up as well, which is a good long-term leading indicator. Perfect. So I want to stick to you both mentioned Scalers and Scalers Plus, your larger customers. How should investors think about that as far as the type of companies that these customers are? Are they enterprises, or is it your traditional digital native SMB customers that just happen to be spending $25,000 or $100,000 on the platform? Yeah, so the Scalers or Scalers Plus that we have subdivided our big customers into, they are typically digital native customers, and a vast majority of them, if not all of them, are independent software vendors in the sense that they make money selling or licensing software. And when you look at the footprint that they have on us, especially the top 500 customers as an example, what defines an enterprise is what is the percentage of their technology spend that they're deploying on the cloud. And these companies, even though they may be smaller in footprint, a substantial amount of their technology spend is spent on building the infrastructure on the cloud. So in that sense, you can think of them as an enterprise, but these are companies that are digitally native software application vendors that are doing mission-critical things on the cloud. At least 500 of them on our platform are spending more than $100,000 with us. It's new for us to break it out, obviously, but what is really important is the focus we're putting on these customers in ensuring that we can get our fair share of their cloud footprint. For that, we have shaped our product roadmap in such a way that it really resonates with them, and we are starting to have increased expansion of their footprint on us. A lot of that, some of it is new workloads, but a lot of it is our workloads moving from other clouds to DO. Perfect. Could you double-click on some of the product innovation that has put you in a position to win those workloads over either some customer examples or specific products that your customers were really looking for that maybe prevented them from putting those workloads on your cloud? Yeah. So as I was explaining previously, the category of product innovation over the last couple of quarters has been focused on helping our customers run mission-critical workloads. So one example is the slew of product innovations that we released over the last few weeks around advanced networking capabilities. So as the customers become larger in footprint, typically they are running multinational workloads or workloads that span multiple data centers. And for them, it is really, really important for them to be able to build network backbones between our different data centers so that they can avoid going through the public internet, but use our DigitalOcean backbone to communicate between different parts of their application. Not only that, we also introduced a new advanced feature, which is VPC Peering, which opens up a direct pipe between a DigitalOcean data center and an AWS data center. So that is also a huge part of winning over the hearts and minds of these large enterprise customers to distribute their workload, even though they're running some parts of it in AWS and other hyperscaler clouds. Now they have a very secure and very high-performant, low-latency way of peering that with parts of the workload that are running on the DigitalOcean platform. So that's just one example. There are several other examples around high-performance Droplets. Droplets is our basic unit of infrastructure. We have released half a dozen special variants of Premium Droplets. It could be Droplets that have high compute. For example, we have a V96 CPU variant of a Droplet. We have high-memory Droplets now available. So these are all examples of innovations that are helping customers pick the right infrastructure footprint for their specific workload. If your workload is running an in-memory database, as an example, you need a large memory footprint on your compute environment. So now you can use that memory-optimized Droplet to run that workload. So these are fairly advanced, sophisticated capabilities that we are releasing now for our customers to be able to run mission-critical workloads on us. Great. And then to complement or follow some of the product innovation around go-to-market, can you talk through some of the initiatives that you've been working on just around customer success, account management? Sure. Yeah. So as we were just describing, these are fairly sophisticated workloads. Now, it's one thing to have these capabilities in our product platform, but it's another thing to ensure that our largest customers are taking advantage of this or, A, first know that we have these capabilities and start taking advantage of these capabilities. So what we did last year in the second half of the year was put a team in a dedicated fashion focused on the top 500 customers. And you can see the results in our latest earnings print where the technical account managers started putting their arms around these customers, engaging with them at the right time with the right sales plays to ensure that they're taking advantage of our latest and greatest. Now we are expanding that concentric ring into the next 1,500 accounts to make sure that we are taking all the lessons learned from the top 500 and applying them at a slightly larger scale to the next 1,000 customers, and we'll keep doing that throughout the remainder of the year, so this is number one, figuring out a propensity model to analyze the product usage triggers to engage with the right customer at the right time. We don't want to keep calling them every day, but we want to engage when the customer needs our engagement, and number two is a series of sales plays that really help our customers, so we are iterating and learning as we go along to make sure that we are in the service of our customers when they need us. As Matt explained, when we decomposed our NDR improvement, it is a combination of this one-two punch of product innovation along with dedicated customer engagement, is what is helping us recover the NDR and drive up expansion. Thanks, Paddy, and to sort of wrap up the topic of focusing on your scalers, your larger customers, I think one thing that has been really consistent since IPO for investors that are new to the story or maybe familiar with AWS or Azure. The initial response is, "How do you fit into this hyperscaler market competitively?" And so these efforts to focus on your larger customers, what's the right takeaway as far as your strategy around competition and just with regard to the hyperscalers? Where do you fit in? Where do you see yourself? Where we see ourselves is in the service of our customers. I'm not trying to give a trite answer, but we define our strategy in the service of our customers. We've been competing with hyperscalers since our early days as a startup. AWS Lightsail was released, what, in 2017 or something. Hyperscalers are fierce competitors up and down the stack. They fight for every little customer. I don't think that has changed. We don't wake up saying that, "Oh, we need to go and take on hyperscalers today." We wake up every day saying, "Okay, how do we get a bigger share of wallet? How do we earn the respect from our customers and earn their trust to get a bigger chunk of their cloud spend?" I think that is absolutely a part of our strategy, and some of it comes from hyperscalers. Increasingly, a lot of it is coming from moving workloads from hyperscalers, and last month, we announced a more proactive, programmatic way of facilitating these migrations from hyperscaler clouds to our cloud, but it all starts and ends with, "What do our customers need and what are they telling us? Awesome. I want to shift to AI and machine learning, and that's an area where we've also seen a lot of exciting announcements, an acquisition and ultimately a lot of innovation up and down the stack. So could you help to articulate the strategy around AI and machine learning? Yeah. So about nine months ago last summer, we started articulating our AI strategy with two core tenets. One, it will be a strategy that is meant for us and for our customers and not copying someone else's strategy. And number two is the strategy is going to be built around software differentiation, not hardware differentiation. So what does that strategy look like? The strategy looks like three layers of AI offerings. On the bottommost layer, you have infrastructure. The middle layer is our platform offerings, and the top layer is our Agentic Layer. In the basic infrastructure layer, we have two offerings. One is bare metal GPUs, and the second offering that we have there is what we call GPU Droplets. Now, each layer and each offering has its own distinct target segment and value proposition. Bare metal is intended for those companies that are comfortable taking an LLM and fine-tuning it. They need the raw horsepower of bare metal. They need all the flexibility that a bare metal environment gives them, and they want none of the overhead associated with hypervisors or virtualized environments. Now, GPU droplets, on the other hand, start getting closer to our sweet spot of customers that want power but also want the flexibility and the abstraction that allows them to focus on the task on hand versus building and managing infrastructure. So GPU droplets, for example, abstracts out some of the housekeeping and the lifecycle management you need to do, like installing PyTorch, activating the right NVIDIA drivers, the NVLink driver infrastructure, and by the way, these drivers are not as stable as they are in the CPU environment. These things crash. They need to be babysat. It's a lot of work to manage this infrastructure. So GPU Droplets abstracts all of that out, and it gives you a plane where you can either have a reserved cluster or you can have GPUs on demand or even fractional GPUs. So that has become really popular. In the last earnings call, we talked about the fact that we are running out of capacity constantly on GPU Droplets, and we are reconfiguring our existing footprint into more GPU Droplets. So that's our base infrastructure layer. Four weeks ago, we announced our platform layer, which is our GenAI as a service. So the GenAI Platform is essentially a serverless LLM endpoints as a service. So what that means is for both closed-source and open-source models like Llama, Mistral, and DeepSeek, we expose them just as APIs. Customers can just use the APIs to build GenAI into their software applications using a serverless model, so they don't have to manage the headache of having to configure and operate an LLM. But at the same time, they can pay on a token-based usage model as they go. So it's give or take, let's say, $0.99 for a million tokens. So they can just focus on building an application that consumes LLMs using an API. And the topmost layer is our agentic layer, where we have started shipping AI agents to solve everyday cloud problems. So for example, the first agent we shipped is a site reliability engineering agent, which can analyze millions of lines of application/server/network logs to find out root cause analysis or perform root cause analysis on cloud applications. So that's the three-layer AI strategy we have defined, and we have started talking about proof points on each of those layers in terms of both revenue as well as the end customer benefits that we are delivering on all three layers. Perfect. Matt, maybe you could talk a little bit about starting with infrastructure and thinking about CapEx and investments. Is there any way to think about the mix of CapEx that's going toward GPU infrastructure, how many GPUs you have, or to kind of give that overview of revenue and growth around AI? Yeah. As we said in 2024, we started off the year with, I'd say, a balanced mix of GPU and CPU. We made a decision midway through the year that we were seeing pretty good demand, and we increased the amount that we were going to spend, which was going to be primarily in the GPU space. So we spent more on GPU in 2024 than we did on CPU. We expect that probably continues in 2025. But again, the quantums that we're talking about are very, very different than the quantums that you're talking about in some of the other kind of neoclouds or some of the pure-plays. We just raised our free cash flow guidance for next year, for this year, to 16% to 18%. You can kind of back in that the CapEx as a% of revenue is actually going to be a little bit down in 2025 versus 2024. And the reason that we're able to do that is, again, we're not building GPU farms to rent tons of servers out to model builders. We're able to mix and match technologies so that we can offer kind of cost-effective solutions for the different use cases. We've got the H100s installed already. We've got H200s that are already installed, some in Atlanta, our new data center. We still use some of the older technologies, the Ampere series and the L40S. And we've also started to take some AMD capacity, which is focused more on the inferencing workloads. So we're able to balance the different technologies across the use cases. And as Paddy said, we ran out of capacity for the GPU Droplet. We'll just reallocate that from the bare metal and just reconfigure things. So I think with the software-oriented and platform-oriented strategy, we can be more real-time in terms of the requirements for CapEx, and we can use customer demand to grow that. We're not having to build it and hope they will come. We're able to feather that in just within our normal capital program. Excellent, and I do want to ask one more, just thinking about where there's the potential for most differentiation on the platform layer and around AI agents. How are you thinking about the opportunity? What could that do for your model? What portion of your customer base do you think can benefit from those products? Do you mean from our platform and application layer? Yeah. I mean, the platform is early days, so is AI agents. And so just hoping you could kind of shine light on how you're thinking of sizing that opportunity and what impact it can have on the model. So this year, we have not factored much impact from a revenue or guidance perspective into our forecast for this year. But as we mentioned, even in the earnings call, the GenAI Platform, even though it is only four weeks old, we have seen very robust adoption and creation of agents on our platform. And what was really interesting is 90% of the traction we saw on our GenAI Platform came from our customers. So 90% of the adoption came from our customers, which is a phenomenal validation of what we've been saying all along, which is the everyday software development company does not have the expertise or the budget to take GPUs and play with LLMs and build a solution from ground up. There's a huge pent-up demand for offering serverless endpoints as a service to enable software shops to build GenAI into their applications. And the necessary platform components that they need is going to grow exponentially over the next few months. So already out of the box, we have a knowledge-based service. We have guardrails. We have so many different building blocks to be able to quickly take GenAI or LLMs out of the box and build it into your application. So we are super bullish about the fact that not only are we making it super easy for our customers, as the first four weeks have shown, it is also quite encouraging for us because for every LLM use case that drives GenAI adoption on our platform, it also drives three or four other services from our platform. Because to be able to take advantage of GenAI, you need to pump in data for which you need storage and a database. You need to do pre- and post-processing for which you need compute and networking. So we are super optimistic that as we get the flywheel of GenAI adoption going with our type of target customers, it is not only going to be great for our AI business, but it is also going to require the cloud primitives to come along with it, which is a huge differentiation for us versus the NeoClouds because we have just spent more than a dozen years perfecting how to build and run and operate a massive cloud infrastructure, which is all essential when you're trying to do inferencing on the cloud. That makes a lot of sense. Matt, you mentioned the Atlanta data center. Is that online yet? It's still on track to come online this month. And then with regard to the comments on running out of capacity, once that comes online, should we expect sort of an immediate bump once that capacity opens up? So from a timing standpoint, we're on track. In fact, we're a little bit ahead of what we had originally anticipated. And that's part of the reason why we indicated that we'll be a little heavier on CapEx in the first quarter. It'll be front-loaded because we're taking advantage of bringing some capacity online. Like I said, we already have H200s up and operational in that facility. And that'll give us a substantial amount of capacity so we can build larger clusters. It'll be easier for us to move between bare metal and GPU droplet and reallocate resource. And it also is a huge step forward for us in our long-term data center strategy. We think there's a lot of cost optimization in getting some more consolidation, but also getting into some lower-cost areas. So we'll move some of the compute out of expensive locations in New York, both on the core cloud side and on the GPU side. From a revenue standpoint, the growth that you're going to see, a lot of what Paddy's been talking about, where we're most excited is in the GPU Droplet and in the GenAI products. Both of those are usage-based and consumption-based. So you'll see revenue growth, but it's not like we have a bunch of Bare Metal contracts that are stacked up that we're going to deploy. That's not our target. We're doing some of that, and we're certainly able to serve customers that are doing a lot of that model fine-tuning. And that'll be a big element of our growth. But what I'm more excited about is the longer-term, more durable inferencing revenue and GenAI revenue, which fits more in our competitive advantage. Excellent. And a follow-up on the optimization on the cost side, how should we think about the evolution of that? Is that like a multi-year once certain contracts or leases expire? You can move workloads over? Yeah, that's the right way to think about it. It's multi-year. We're in 16 or so data centers. We've got multi-year commitments with some large providers, and we'll migrate over time, staying in some of those facilities and maybe just moving some of the workloads out, reducing our footprint. But there's a lot of potential there. I think there's a lot of juice on the gross margin side to drive over the coming years. But that's not a quick quarter-to-quarter exercise. That's something the market and you will be able to see because you'll hear us announcing new facilities. Great. I'm going to ask a few financial questions, and then we'll open up for questions from the audience. Thinking about your FY25 guidance, in 2024, you added close to $100 million in ARR. In 2025, based on my estimates, the guidance implies a little bit more than that. But we just talked through Atlanta data center coming online, new AI products hitting the market that are growing very rapidly already, so more contribution there. And then the shift in focus up market and the stronger growth from your larger customers. My takeaway is that when you think about all of those things that will contribute more this year versus last year, the guidance seems conservative. Is that the right takeaway? Can you talk a little bit about some of your assumptions behind your outlook for this year? Yeah. I would say our guidance is appropriate. And when I say appropriate, our philosophy is we're going to give you guidance that's based on our baseline growth rate, which is what we're delivering. And we're certainly working to exceed that. We'd like to see NDR get noticeably above 100. We'd like to see AI take off, and we're doing a lot to drive both those things. But what we're telling you is we're on the trajectory to deliver the 13-ish% growth that we said. When you look at ARR, it's heavily, particularly when you're looking at an annual basis, it's heavily dependent on what you do in the fourth quarter. And so we want to make sure that we're setting appropriate expectations, that this is growth that we can count on and we believe in. It's Paddy's job, my job, and Bratin's job, and the rest of the executive team to beat those numbers. But we feel very good about the guidance that we have. Like you, we're very encouraged by how we entered the year with 99% growth, sorry, NDR, 100% in the core cloud. We feel pretty good about how we entered the year, and we're going to work to make sure that we can beat those numbers. Excellent. And then on the margin side, I mean, we talked a little bit about some opportunities to optimize on the infrastructure side. And thinking about this past year, you guided margins initially a little bit more conservative given all the leadership transitions there. But hoping you could unpack a little bit about just what to expect from sort of margin expansion philosophy going forward and really where operating leverage comes from. Yeah. So you've seen this from us. We're very fixated on providing super tight guidance around free cash flow. And clearly, free cash flow per share is a critical metric for us. And we feel very good about the free cash flow margins for this year. And we were able to increase that to 16%-18%. As you said, we've been a little bit wider in our guidance around EBITDA and operating expenses effectively. Gross margin is not something that changes on a monthly basis. You take down capacity, you grow into it. You see those changes when we make it. Atlanta data center is a good example. And so with the new team, with basically a brand new executive team, we've wanted to give ourselves a little bit of wiggle room around operating expenses. If we want to accelerate some of the product roadmap by bringing on contractors or Bratin wants to surge in a specific area, we want to give him the flexibility to do that because it's in service of growth for our larger customers, but from a long-term leverage perspective, we continue to leverage a global workforce. We're ramping up our hiring in India. We got a country leader in India. We've had a presence there for a long time, but we're taking down an office. We're building a bigger footprint there. We continue to get really good leverage around G&A. Our G&A margins are definitely improving, and we've done, I think, a very, very good job of controlling stock-based comp. That's come down quite a bit, so I think from an OpEx leverage and an operating leverage, we still have room to go. But we've demonstrated we can do a pretty good job of managing our expenses and still delivering growth. Two more topics that we've seen in the past: pricing increases, both for the core and then subsequently Cloudways and M&A with Cloudways, Paperspace. Wanted to ask you about what should investors assume around your potential to take price or increase prices down the road and also to get an update on your M&A strategy. Yeah. So I'll take the pricing, and you can take the M&A strategy. So from a pricing perspective, I think more about the packaging than pricing. So a good example of this is the Premium Droplets that we have been launching over the last several weeks. And we have been very pleasantly and positively surprised by the uptick rate of customers choosing. We don't make it a default. Customers choose into these Premium Droplets, which are obviously significantly more expensive than the base Droplets. But what it tells us is customers are willing to pay extra for a different type of performance. And they are willing to make the trade-off themselves. And it, of course, cannibalizes our base Droplets, but we are happy to make the trade-off any day if it helps our customers. There are a lot of other packaging innovations that we are doing, and we'll continue to do that. That's just a regular part of our business. Pricing at a base level is not a huge part of our strategy. We'll absolutely be very creative on the packaging front. And then from an M&A standpoint, I'd say with the team being relatively new and the huge focus on nailing our core customers and getting that growth going, we haven't had a huge focus on it since the Paperspace acquisition. I'd say going forward, if we were going to be active in anything, it would be more likely small-tuck-in technology acquisition, something talent that accelerates our product roadmap. We believe there's a lot of room from an organic runway standpoint. That's where we're putting the bulk of our capital. But if we saw an opportunity to accelerate something, we certainly would take a look at it. Excellent. Do we have a question? We do have one question out there. Can you repeat the question, please? Away from the hyperscalers, who do you see as the competition and your points of differentiation against? Yeah. So it's hard to go away from the hyperscalers because they have such a dominant presence in the market. And as I explained, we're really customer-obsessed, not competitor-obsessed. We are absolutely competitor-aware, but we are obsessing on what our customers need. And that's what we are focused on. And the reality of today's software world is that most companies are multi-cloud, which means surely there is some hyperscaler presence in most of our large customers anyway. So it's a long-winded way of saying we are really focused on what our customers need and delivering that value to them and taking care of them. And we don't really obsess over what competitors are doing. Great. We are out of time. Paddy, Matt, thank you very much. Really appreciate your insights. Thanks, Jeff. Thank you, Doug.