CoreWeave, Inc. (CRWV) Earnings Call Transcript

Thanks, everyone. Hello, and good afternoon — glad you could join us for CoreWeave's First Quarter 2026 Earnings Conference Call. Here with us to go over the results are our CEO, Mike Intrator, and our CFO, Nitin Agrawal. Ahead of diving in, I want to use this moment to point out that remarks made today will contain forward-looking statements. Those statements could turn out to be wrong, and what actually happens may be materially different from what they anticipate. The factors that might lead to materially different outcomes are laid out in today's earnings press release as well as in the quarterly report on Form 10-Q we will file with SEC. Every forward-looking statement offered on this call rests on assumptions held as of today, and we take on no duty to revise these statements because of new information or events that occur later. On this call, we will show both GAAP and certain non-GAAP financial measures. Today's earnings press release contains a reconciliation of GAAP to non-GAAP measures. You can find the earnings press release and an accompanying investor presentation on our website at investors.coreweave.com. There will also be a replay of this call on our Investor Relations website. At this point, I'll hand the call over to Mike.

Hello, everybody, and thanks for being here this afternoon. CoreWeave experienced a quarter of major change in Q1. We secured our largest customer order to date — more than $40 billion in new commitments — bringing our contracted revenue backlog up to nearly $100 billion. Top-line revenue reached about $2.1 billion, up 32% from the prior quarter and 112% compared with the same period last year, and we surpassed 1 gigawatt of active power. Only a handful of cloud providers have ever hit that mark, and it reflects our progress in converting contracted capacity into revenue-generating cloud services; we ended Q1 stronger than at any previous point. AI adoption is accelerating, and our total addressable market, our customer roster, and our platform are all expanding rapidly. CoreWeave remains at the forefront of this generational change. Today I'd like to focus on 4 themes: first, demand is ramping up even more, fueled by the explosive growth of our current clients and the fast maturing of new enterprise verticals. Second, our platform's capabilities have been expanded. to handle any customer use case — from training to inference to agentic workloads — which puts CoreWeave in position to gain from durable, margin-accretive expansion. Third, we've achieved hyperscale status with over 3.5 gigawatts of contracted power, an increase of more than 400 megawatts in this quarter by itself, and the substantial majority is anticipated to be online by the close of 2027. Fourth, our financing approach has advanced considerably, opening up fresh capital sources in various markets at a reduced weighted average cost, which allowed CoreWeave to obtain over $20 billion in debt and equity so far this year. Turning to the demand side. The market CoreWeave can serve keeps growing, and clients are committing to our platform for the long haul. AI workloads are shifting away from training toward inference, agents and enterprise production spanning many industries, and that trend requires ever more compute. Because of this, our main customers — traditionally hyperscalers and foundation labs — are strengthening their ties with us. At the same time, a whole new group of enterprises is showing up and asking for large-scale access to CoreWeave's platform. So because of that pattern, we booked more new business in Q1 than ever before—our first virarubin contracts got signed, and at the same time we kept earning money off our Blackwell, hopper, and Ampere capacity. Most of those fresh customer commitments count toward what we're aiming for in 2027. And here's the key part: we anticipate they'll carry very strong contribution margins, in line with the returns we've traditionally assumed when underwriting new deployments. During Q1, Anthropic came on board as a client to help build and roll out the Claude family of AI models. On top of that, we closed several additional deals with Meta, among them the $21 billion agreement we announced in early April. As it stands, all 4 of the top AI model developers globally are now on CoreWeave Cloud, which means 9 out of the 10 leading AI players outside China are with us. Meanwhile, fresh verticals are appearing, and they've already hit a scale above $1 billion. In the Financial Services space, companies powered by technology are growing their primary machine learning operations through CoreWeave. These aren't AI research organizations — they're enterprise clients who recognize the real monetary gains and appealing returns that result from using CoreWeave Cloud. Revenue backlog for this specific portion of our business is already approaching $10 billion. That expansion stems from larger commitments by existing partners, for example Jane Street, which added $6 billion more in capacity in Q1, as well as new clients like Hudson River trading. Similarly, revenue backlog tied to Physical AI and spatial computing has now surpassed $1 billion. Companies pushing the boundaries in world models, robotics, self-driving technology, and scientific research choose CoreWeave because we offer a unique combination of speed, purpose-built infrastructure, and developer resources that accelerate training and deployment. World Labs, Physics X, and Sunday Robotics are among those we've brought on board lately. Along with our intense dedication to getting things done, these developments are driving more customer diversification at CoreWeave. Right now, 10 clients have pledged to spend a minimum of $1 billion with CoreWeave. Supporting such a broad spectrum of customers and workloads requires us to use both recent and prior generations of NVIDIA GPUs, particularly with inference demand surging and agents entering the workforce. So as a result, we're seeing momentum build in all of our categories. When you look at average pricing, it went up from one quarter to the next for the A100, the H100, the H200s, and the L40s, and on a fleet-wide basis, we remain largely sold out for capacity in the near term. That validates a view we've held for quite a while. The need for inference-ready compute that stretches across several GPU generations just keeps growing. We expect that to hold up over time and to contribute to our margins and earnings power over the long run. The size and the quality of that demand deserve a bit of framing. Inference is where AI turns into revenue, and as it speeds up, it's driving real productivity improvements in the actual world, which supports putting more capital in and brings more enterprises on board. Because of that, we brought in more backlog during one quarter than most AI cloud platforms have accumulated over their entire history. Turning to CoreWeave's cloud platform; we are intentionally reinforcing our integrated stack so we can provide the most capable AI cloud. CoreWeave enables researchers, developers and platform engineers to iterate quickly across training, inference and agentic workloads, speeding the path from experimentation to production at scale. In doing so, we keep rolling out new capabilities so enterprise customers can build on CoreWeave. For instance, during Q1, we launched the CoreWeave Trust Center so enterprises can productize AI rapidly and efficiently while keeping their security or compliance standards intact. What draws customers to CoreWeave isn't just GPUs — it's our complete, integrated suite of AI cloud capabilities. Getting the most out of accelerated computing means having CPUs, storage, networking, software solutions, and developer tools all working in concert across every layer. Right now, more than 90% of those with reserved instances use at least two of our offerings, while upward of 75% use three or more. Our storage segment continues to grow at a fast clip. We're seeing similar momentum in software, CPU, and networking, and we anticipate that by year-end, each of those will have exceeded $100 million in ARR. Watching customers put these products to work in innovative ways is truly inspiring. To give you one example, the next generation of inference workloads from Perplexity is going to be run on CoreWeave's platform, and weights and biases will be used by them as well to assist with training, fine tuning, and managing their models. Meanwhile, Advaita Bio is using CoreWeave to accelerate pathway and single-cell analysis over enormous biological datasets, cutting workflows that previously took weeks down to only minutes. As we scale up, there are more purchasing options for compute available to customers. Flex reservation plus spot pricing was introduced by us in Q1, which changes how clients deal with demand spikes and sudden bursts, allowing better budget planning while taking advantage of the larger preemption windows CoreWeave provides. Right away, demand for both of these offerings was greater than supply. We're additionally simplifying cross-cloud AI for clients, all while making sure CoreWeave turns into their essential cloud partner. Following the success of CoreWeave's AI object storage along with our 0 egress offerings. We just unveiled CoreWeave Interconnect together with Google Cloud; alongside Sunk Anywhere and < > Crosscloud, the goal is to eliminate the hassle of overseeing a multi-cloud presence, helping organizations run workloads anywhere with less complexity and more speed, and these are already delivering strong results in capturing a larger share of customer spending. And to reach customers wherever they happen to be, we're starting to provide CoreWeave Omni, which lets us install and run our complete cloud stack inside customers' personal data centers using their GPUs.

There's already significant demand building among prospective cloud, enterprise, and sovereign clients, and we're eager to provide updates once it's available on the market. When you put all of this together, what these capabilities represent is a straightforward yet substantial transformation. People choose CoreWeave so they can build and deploy alongside us. Now on execution — we take limited AI infrastructure and turn it into AI cloud capacity that generates revenue, doing so swiftly, dependably, and with good margins. During this quarter, CoreWeave went past 1 gigawatt of active power. Our goal of exceeding 1.7 gigawatts before 2026 wraps up remains something we are solidly positioned to hit or surpass. Among cloud platforms throughout history, only a small number have ever achieved this kind of scale, and ours is the sole one designed specifically for AI. Every fresh build-out involves 5 complicated phases: power, cooling, networking, servers, and the software orchestration layer. Across roughly 50 data centers, we've handled this intricacy, reliably achieving best-in-class dock to live times for our customers. CoreWeave treats these build-outs as investments in the communities where they're located, and we strive to earn our standing in each region we serve. Our facilities act as anchors for regional economic activity that builds over time — grid upgrades, workforce development, sustained local investment. Every project brings its own distinct set of opportunities and challenges. That's the mindset we bring to each market we enter, helping our team and partners expand their expertise and execution capabilities for whatever project comes next; during Q1 we brought on more than 400 megawatts of contracted power, pushing our total contracted power beyond 3.5 gigawatts. By the end of 2027, we anticipate the substantial majority of that contracted power will be online; all of this capacity was added through long-term leases, as data center partners and the financing markets identified a unique blend of execution, technology and contracted demand that characterizes our business. Going forward, we intend to keep growing our contracted power footprint via leases while simultaneously speeding up development of self-build sites, giving us better operational control and long-term financial upside. Later this year, we expect our first self-build site to go live. Our lease and self-build strategies gain additional support from our strategic relationship with NVIDIA, through which we keep assessing ways to speed up expanding our footprint together. This multifaceted approach gives CoreWeave a unique position to grow within a highly competitive market. Last, a word on our financing approach. During Q1, we hit a transformational milestone that will push CoreWeave's weighted average cost of capital downward, closing our $8.5 billion delayed draw term loan 4.0 facility. Nitin will cover our broader capital markets success, but I wanted to call out a few aspects of this particular facility; it's the first ever investment-grade delayed draw term loan backed by HPC infrastructure, earning an A- equivalent rating from Moody's, Fitch, and DBRS. The facility was nonrecourse to the parent. The transaction, which saw strong oversubscription, was priced at a level implying a cost of less than 6%. Through this financing, we've moved significantly closer to hitting the target we set out: pushing our debt costs down to investment grade. That isn't a small, step-by-step gain. It represents a fundamental change in the way we intend to fund investment-grade customer contracts moving forward, in one of the capital markets' deepest sectors. Keep in mind, between 2023 and 2025 we already brought our weighted average cost of debt down by roughly 600 basis points. Right now, as of today, we've squeezed that weighted average cost of debt even lower by about 80 basis points so far this year, all while locking in over $20 billion in debt and equity capital. That lowers the risk in our execution plan and gets CoreWeave ready for ongoing hyper growth. Before handing things to Nitin, I wanted to restate that this company has been built around a clear, defined set of objectives: first, keep providing the most technically advanced cloud platform for AI workloads, letting our customers innovate, build, and deploy; second, broaden and expand our customer base; third, achieve best-in-class execution at hyperscale; and last, align our capital structure so it scales alongside the opportunity. Our backlog now sits at close to $100 billion, every bit of it tied to contracts either already online today or expected to come online during 2026 and 2027; we've assembled a diversified customer base spanning each of the world's leading model platforms and reaching large enterprises across industries representing tens of billions in market opportunity and growth. We've gone past just GPUs to offer an integrated AI cloud platform to our customers, who are quickly embracing our CPU, storage, networking, and software solutions. Active power has now passed 1 gigawatt, and contracted power stands above 3.5 gigawatts, which puts us in a strong spot to hit our target of more than 8 gigawatts of active power by 2030. We also keep innovating in the capital markets, having completed the first investment-grade-rated financing ever backed by HPC infrastructure and creating a new asset class in the process. Every one of these things makes the business sturdier, every one of them builds on the last, and every one of them is a foundation for our next phase of growth. The bottleneck in AI isn't whether enterprises and AI labs are willing to deploy anymore. It's how fast high-performance, reliable AI cloud capacity can actually be delivered. That's where CoreWeave excels; with that, I'll hand it over to Nitin.

Thank you, Mike, and hello to everyone this afternoon. CoreWeave's first quarter was yet another landmark period. Unprecedented levels of customer commitments pushed our backlog to approximately $100 billion, alongside over $2 billion in early revenue and greater than 1 gigawatt of live power, all while we opened up deeper and more efficient financing avenues. What we're delivering matches exactly the plan we shared during our previous earnings call — broadening and deepening customer relationships, securing contracts that carry appealing and steady margins, quickly bringing additional capacity into service, and extending our purpose-built AI cloud platform. Interest in CoreWeave cloud keeps climbing, and our 2026 capacity is still mostly spoken for, with pricing rising throughout every tier from Ampere through Hopper and on to Blackwell. This trend is carrying into 2027 too, since we've started assigning capacity that we anticipate will go live next year. Now for the first quarter results. Revenue reached $2.1 billion during Q1, representing 112% growth versus a year earlier and 32% growth quarter-over-quarter, powered by our ongoing strong execution in bringing capacity online. Appetite for CoreWeave cloud keeps growing more intense. The quarter closed with revenue backlog at $99.4 billion, roughly 50% higher sequentially and almost 4x greater than the prior year; this backlog skews toward the near term, with 36% anticipated to be recognized over the coming 24 months and 75% within the coming 4 years. As enterprise adoption grows stronger and our client mix becomes more varied, commitments originating from noninvestment-grade AI-native firms and foundation labs now account for under 30% of total backlog. Clients keep dedicating their foundational AI workloads to CoreWeave, which means the weighted average contract duration for newly added capacity stays at roughly 5 years. First quarter operating expenses came to $2.2 billion, which includes $153 million in stock-based compensation expense. The rise in operating expenses stemmed directly from our ongoing effort to scale live power capacity and turn backlog into revenue. That in turn produced matching growth in cost of revenue along with technology and infrastructure spending. On top of that, sales and marketing rose because of greater investment in our go-to-market organization as we keep diversifying our client base and push into additional products and markets. G&A went up due to personnel costs needed to support our expansion, though it moderated relative to revenue growth on a quarter-over-quarter basis. Adjusted EBITDA for Q1 hit $1.2 billion, versus $606 million in Q1 of 2025, a 91% increase year-over-year. Our adjusted EBITDA margin came in at 56%. For the first quarter, adjusted operating income came in at $21 million. That compares to $163 million in Q1 of 2025, and it put us slightly above the middle of the range we had guided to. Our adjusted operating margin in Q1 was 1%, and we continue to expect that figure will mark its low point, since cloud capacity will keep coming online across the coming quarters. We posted a net loss of $740 million in Q1, against a net loss of $315 million in Q1 of 2025. Interest expense in Q1 was $536 million, up from $264 million in Q1 of 2025, which reflects the extra debt we took on to finance the continued build-out of our infrastructure and to meet the customer commitments we have under contract. We booked an income tax provision even though we had a net loss, because of a valuation allowance against net deferred tax assets. Assuming there aren't any major discrete items or a shift in circumstances, our tax rate ought to stay broadly steady through 2026. Adjusted net loss came to $589 million for Q1, versus $150 million in Q1 of 2025. Now on to capital expenditures. CapEx reached $6.8 billion in Q1, with execution continuing on schedule. Construction in progress, or CIP, was essentially flat quarter over quarter. Keep in mind that CIP reflects infrastructure that isn't in service yet and isn't being depreciated yet. Once those assets enter service, they generate incremental revenue along with matching depreciation. The way we've set up our financing is meant to match how we roll out these projects. Most of our term debt comes through delayed draw facilities, which means we only pull capital once a data center is up and running. Global supply chains remain challenging, but we continue to navigate those issues through operational discipline, leaning on our partnerships to secure the inputs we require in a strategic manner. Turning to the balance sheet and our strong liquidity, at the close of March 31, we had more than $3.3 billion across cash, cash equivalents, restricted cash and marketable securities. From the start of the year, we've achieved meaningful strides in strengthening our balance sheet and expanding the depth as well as the breadth of our access to capital. What we've accomplished is rooted in the principle that the money we bring in is connected to what customers actually need. From day one, that philosophy has steered this company, and it's been crucial for our ability to grow access to capital in an efficient way. In the first quarter, we brought in $2 billion of equity associated with deepening our partnership with NVIDIA. In addition, we secured about $8.5 billion in investment-grade debt by way of our fourth delayed broad term loan, a landmark deal for CoreWeave. As Mike pointed out, DDTL 4.0 represented the first HPC infrastructure-backed debt facility ever to earn an investment-grade rating, with an A- equivalent grade given by 3 separate ratings firms. The facility came in at an implied rate below 6%, which is a significant drop compared to what we had with earlier facilities. On top of that, we added an ABS-style draw feature, which frees up another $1 billion in drawable capital once the underlying contract reaches stabilization. That extra capital can go toward future investments, helping us deliver subsequent capacity at a very attractive price. In addition, DDTL 4.0 was set up as nonrecourse, which let us build a facility with greater capacity and better pricing while leaving CoreWeave Inc and its lenders unaffected. Broadly speaking, we see this approach becoming the standard way CoreWeave refinances capacity build-out for investment-grade customers. This marks a major leap in what we can do when it comes to raising money — enormous amounts of it, on terms that stack up against what the hyperscalers get. Coming into Q2, that momentum kept building. We closed several separate deals, pulling in over $10 billion in combined debt and equity. Demand blew past what was on offer in every one of those transactions, which shows just how badly investors want a piece of CoreWeave's hypergrowth. In fact, the response from investors pushed both the convertible deal and the high-yield deal we did most recently to noticeably bigger sizes. On top of that, Jane Street's $1 billion strategic investment makes clear how much worth and distinctiveness customers see in what our platform offers. In addition to those rate increases, S&P revised its view of our corporate credit rating, moving it from stable up to positive. We set the pricing yesterday for our fifth DDTL facility — the first one syndicated in the public loan markets that funds contracts with Open AI and Cohere. Investor demand proved substantial once again, allowing us to set pricing 50 bps below where we had initially marketed it. With this deal complete, we now have every piece of financing in place needed to fulfill all our current commitments with Open AI. This deal also highlights how investors are backing CoreWeave, whether we're financing investment grade customers or AI lab customers. Taken together, these deals mean we have secured over $20 billion in debt and equity capital so far this year. For CoreWeave, expanding our ability to raise money at a reduced blended cost will remain a key tool as we turn backlog into revenue and operating cash flow while actively overseeing our capital structure. Because of this, aside from self-amortizing contract back debt and OEM vendor financing, we face no debt maturities until 2029.

Now for our outlook. Given how strong the first quarter came in, how tightly we're running the business, and the momentum that keeps building across the customers we serve, we feel good about standing behind our guidance for the whole year: revenue between $12 billion and $13 billion, and adjusted operating income between $900 million and $1.1 billion. Looking at Q2, we're projecting revenue of $2.45 billion to $2.6 billion, and adjusted operating income of $30 million to $90 million, since margins should begin widening off the lows we posted in Q1, which lines up with what we told you previously. That margin pattern comes down to timing, not economics. Let me give you a bit more color on this. Once we take possession of a powered shell, we're paying lease and power costs and depreciating servers and other data center equipment while the fit-out work happens, and that usually runs about 1 to 2 months. Through that stretch, we book costs with no revenue coming in, so those freshly deployed sites operate at negative contribution margins. Come month 3, though, we're normally producing revenue, with contribution margins settling into the mid cities. Since the start of 2025, active power has nearly tripled, now above 1 gigawatt, and we're closing in fast on escape velocity; we still expect adjusted operating margin to climb quarter over quarter through the rest of the year, getting back to low double digits by Q4. We're looking for Q2 interest expense of $650 million to $730 million, which mirrors how our debt balance has grown to fund deployments that are speeding up; CapEx should land at $7 billion to $9 billion as we keep bringing substantial capacity online to serve the contracted revenue backlog we've signed. For the full year, CapEx is now projected at $31 billion to $35 billion. That higher low end versus our prior guidance traces to rising component pricing. Because our contracted revenue backlog stretches so far out, we keep getting clear visibility into 2026 and beyond, and our confidence in the revenue and margin targets we've laid out is unchanged. We now anticipate finishing 2026 at $18 billion to $19 billion in annualized run rate revenue, lifting the low end of our expectations by $1 billion. We still expect annualized run rate revenue to climb past $30 billion when we exit 2027. Over 75% of that is under contract already, leaving out any upside from customer renewals that haven't been exercised yet. We've locked in enough power capacity to hit our 2027 target, and we plan to keep adding capacity and customer commitments to build on this remarkably strong foundation; with every quarter we execute, every megawatt we deliver, every contract we sign, and every new customer we bring on, we're steadily working through our long-term plan and gaining more conviction that we can meet or beat our long-term revenue and margin targets. Q1 marked a quarter of tangible progress. Revenue and backlog both rose significantly as we kept adding new blue-chip customers while expanding alongside the partners we already have. Every contract we take on still turns a profit. Q1 marked the lowest point in our margin path, but we're still set to see margins grow quarter over quarter for the rest of the year, and both revenue and margin should begin climbing as we move from Q2 into Q3. Just as crucially, we've made real headway on our capital structure, securing over $20 billion in debt and equity, entirely through meaningfully or subscribe deals. Any one of these stands on its own as meaningful. Put together, though, they build on each other, since being able to fund an opportunity of this size at a falling cost is among the strongest tools we have; our contracted revenue backlog gives us visibility across multiple years, and our capital structure is now deeper and more cost efficient than it has ever been; we're eager to share updates on our progress over the rest of the year. Thank you. With that, let's take questions.

Fantastic, and nice work on a genuinely impressive quarter across basically every one of the main approaches you folks are working to bring to this market, and the pace at which the business is moving is simply remarkable. A few items that are more or less clarifying ones inside this, since there are many figures and plenty of new material coming at us. Perhaps beginning with the CapEx part of things and the higher component pricing. I believe that is one thing putting some pressure on the stock somewhat after hours. Could you walk us through how the higher component pricing flows through the contract and how that in the end impacts your profitability? And then — and perhaps we could discuss the NVIDIA relationship and the broadening of that relationship, what essentially shifted in that relationship during this quarter? And how ought we to think about that 5 gigawatts inside the roughly 8 gigawatt active power target for 2030. I will save my remaining questions for the call back.

Thanks, Keith. That was quite a loaded question there. Let me attempt to sort of unpack them and sort of go through them one by one; so the first item is a question about the CapEx budget and the higher component pricing; and this has been a recurring reality for the cloud and for artificial intelligence infrastructure throughout the space. Look, we have sort of constructed this company in an environment that has always been strained on the supply chain side; so we are truly skilled at it. We think about it frequently. We have built the company from an efficiency standpoint. And so we are accustomed to functioning in a difficult environment regarding components or inputs into our product. That said, over the last 6, 9 months, there has been a severe shortage of certain components that have risen. And as I said, you have heard about that throughout the space. When you are considering that for us, the way to consider it is that we are a success-based company, which means that we construct our contracts to incorporate the cost of all of the components that are necessary to deliver infrastructure. And so generally speaking, we are protected from the price inflation on some of the components because we include that in our pricing that we ultimately bring to clients in order to target the margins that Nitin spoke to, right, is up in the mid-20s is how we think about it. on a unit basis; and so it is an issue, it is a problem, but we have an incredible capacity to navigate the supply chain. We have great partners, and we include the pricing that is required in order to end up delivering the infrastructure that is required, but also ensuring that we are able to secure the economics that we are targeting. The next question you asked was about NVIDIA. And it has been a very exciting period for us with NVIDIA. They rolled out a bunch of initiatives, but the two that mattered most were getting our software solution qualified as one of their reference architectures — which was an unbelievable endorsement of how good the software we bring to market really is. We've said this previously: we're committed to giving AI infrastructure customers the finest solution anywhere in the world, and we think we're doing exactly that, and we think NVIDIA backs that up, given where they stand on treating our software as a reference architecture, which is terrific. On the 5 gigawatts of infrastructure, I'd like to begin by noting that over the past 12 months, CoreWeave has locked in 2 gigawatts of infrastructure. During the most recent quarter, we locked in 400 megawatts of infrastructure. We can secure infrastructure at enormous scale, without relying on support from NVIDIA or anybody else. What the 5 gigawatts gives us is the chance, when opportunities arise, to speed up our capacity to go secure more infrastructure at a genuinely remarkable scale, while our clients are seeking to lock in our solution for delivering computing infrastructure. And I'd say those are truly the standout points regarding our relationship over the last quarter with NVIDIA, past the ordinary work of us integrating with them with engineering as the first priority, which has been simply terrific.

And Keith, to wrap up what you were saying regarding the CapEx fees. For us, CapEx appears ahead of revenue and cash flow. So you're seeing it show up first. The portion concerning the P&L impact from that is already baked into the guidance we've provided you today. So with that, I understand this is your final quarter earnings covering us. We'd like you to — we want to thank you for your partnership and the support you've demonstrated for us since our IPO today, take today. Thank you very much, Keith. You will be missed.

What you all have put together, or what has genuinely been amazing and great to see up close, is how fast you have expanded this.

Fantastic team.

I want to give Keith an excellent and say congrats. Nitin, you posted $81 million of EBITDA during the first half, yet your guidance for the bottom line in the second half sits at $919 million. What makes you confident about that bottom line build across the back half of the year?

Sure. Looking at Q2, things landed precisely where we had anticipated. Keep in mind that active power and capacity ramp schedules drive everything in our business, and those schedules don't always move in a linear fashion. What matters here is that both our revenue growth and our margin trajectory are unfolding just as we laid out during our Q4 earnings. I believe your question was likely about EBIT rather than EBITDA, so let me clear that up for everyone; on the EBITDA side, we delivered $1.2 billion during this quarter. If you consider things from a Q1 standpoint, Q1 represented the lowest point where our margin began. We stay on course for sequential expansion throughout the rest of the year, exactly as we had outlined. Revenue and margin growth should inflect as we move from Q2 into Q3. We are reaffirming our guidance for the full year, covering full year revenue and full year adjusted operating income, along with the fact that we will finish the year at a low double-digit adjusted operating income margin. We are reaffirming our active power, and we are lifting the floor on our 2026 exit ARR guidance. Each of these signals reflects the conviction we hold regarding our execution plan for the rest of the year. You will watch us carry out this plan, and we expect our adjusted operating income to speed up more quickly than revenue growth during the second half of the year.

Brent, just a few more points. You wanted to know what's behind our confidence, and a few things I'd like to mention as you observe a business sort of reaching escape velocity: we've moved forward and worked through our supply chains so that we can reach those figures; we're in roughly 50 data centers; no single data center provider accounts for over 17% of our active infrastructure. We have several OEMs and ODMs, and we're now truly structured to lean into a resilient supply chain spanning every component we need to be able to deliver in order to generate that revenue. And the reason we're reaffirming it is the degree of confidence we have that, whatever the cost of that resiliency, we will reach those figures.

Let me also congratulate Keith. And nice work on the solid first-quarter results. If you've got a business you previously booked that isn't completely operational yet, and component prices have gone up — I'd imagine energy costs are higher too, though perhaps some of that is locked in — is the margin picture on those earlier signed contracts a bit different from what was initially planned? Or is there some way for you to pass along higher pricing or perhaps rework a few of those agreements? Then I have one quick follow-up.

Yeah, that's a fair question, thank you. Honestly — I truly want to state that this has been a remarkable quarter for the company, and we couldn't be more thrilled. Regarding components, you'll notice we're making a minor tweak to the CapEx to account for smaller CapEx items that could see some price inflation. However, when we set prices on our deals, we do so with purchase orders already secured for the infrastructure needed to fulfill them. We know what the power cost will be now, in the coming year, and throughout the contract's duration, since that's contractually provided to us; so we've managed quite well to understand future component and electricity expenses, and we've set it up so that it essentially flows through at the time we sign the contract, again, to make sure we achieve our unit-level target margins.

Okay, got it. And considering how well bookings went in Q1, the excellent execution on data center builds, your diversification among enterprises, the successful capital raise, plus the noticeable revenue upside in Q1 — what's stopping you from flowing that revenue upside into the full-year revenue guidance? Because, in a way, that would technically lower our remaining-year revenue forecast slightly. Is there another factor at play? Perhaps weather, or infrastructure shortages, or maybe labor shortages, creating a mild headwind?

Yes. As we see it, we brought this up during our previous earnings call too. We're essentially still sold out when it comes to our 2026 capacity. That remains the case for us right now, and this kind of index for us centers on 2 vectors; we've lifted the guidance for the floor of exit ARR. So we'll close out the year in a far stronger position than we anticipated a few months back, when we had our last quarter. At the same time, where it's also showing up is in 75% of our 2027 ARR guidance that we gave, which is $30-plus billion already booked right now, excluding any potential renewals better than 75% of that. So that's where you see it reflected. From a 2026 standpoint, we're essentially still sold out of our capacity.