Prepared remarks
Hello, everyone. Thank you for joining us and welcome to the Backblaze second quarter 2026 financial earnings call. After today's prepared remarks, we will host a question and answer session. I will now hand the conference over to Mimi Kong, Director of Investor Relations. Mimi, please go ahead.
Thank you, good afternoon, and welcome to Backblaze's second quarter 2026 earnings call. On the call with me today are Gleb Budman, co-founder, CEO, and chairperson of the board, and Marc Suidan, chief financial officer. Today, Backblaze will discuss the financial results that were distributed earlier. Statements on this call include forward-looking statements about our future financial results, the impact of our sales and marketing initiatives, cost savings initiatives, results from new features, the impact of price changes, supply volatility and pricing, our ability to compete effectively and manage our growth, and our strategy to acquire new customers, retain and expand our business with existing customers. These statements are subject to risks and uncertainties that could cause actual results to differ materially, including those described in our risk factors that are included in our most recent quarterly report on Form 10-Q and our other financial filings. All forward-looking statements that we make on this call are based on assumptions and beliefs as of today and we undertake no obligation to update them except as required by law. Our discussion today will include non-GAAP financial measures. These non-GAAP measures should be considered in addition to and not as a substitute for our GAAP results. Reconciliation of GAAP to non-GAAP results may be found in our earnings release, which was furnished with our Form 8-K filed today with the SEC. You can also find a slide presentation related to our comments in the webcast, which will also be posted to our Investor Relations page after the call. Please also see our press release or presentation for definitions of additional metrics such as NRR, gross customer retention rate, and adjusted free cash flows. And finally, we will be hosting an investor day on Wednesday, September 9th in New York City. Please reach out to IR at www.backblaze.com to RSVP for the in-person event. A live webcast will also be accessible from the Backblaze Investor Relations website. Thank you for joining us. And I will now turn the call over to Gleb.
Thank you, Mimi. And thank you, everyone, for joining us today. We had a fantastic second quarter. Revenue came in at $42.7 million, $2.5 million above the high end of our guidance range. Adjusted EBITDA margin was 30%, 700 basis points above the high end of our guidance range and B2 growth accelerated to 34% year over year. These results reflect broad momentum across the business. We also signed the largest contract in Backblaze's history, a $335 million multi-year agreement with CoreWeave, which I'll come back to in a moment. We continued to move upmarket and ended the quarter with 235 customers contributing more than $50,000 each in ARR, up 57% year-over-year. ARR from this cohort grew 67% year over year. We signed numerous AI companies, including a leading frontier model developer, introduced the ability to run our cloud storage in customer-owned data centers to support regional and sovereign workloads and expanded our AI startup outreach and agentic developer tooling. This quarter's results are proof that our AI strategy is working. The decision to lean into AI is translating directly into the financial performance you just heard and into the momentum I'll walk you through now. To understand our strategy, consider this. Every training data set, checkpoint, inference output, and Gen AI asset has to be stored and used. Customers consistently tell us they have three needs to support that. Number one, the ability to scale with fast growing data. Number two, architectural freedom to use their cloud of choice. And number three, storage performance that optimizes their AI workloads. And all of that needs to be affordable so that AI scales efficiently. The combination of those three requirements are why they choose Backblaze. Let's talk first about how that plays out with neoclouds and inferencing clouds. Many of these initially focused on GPUs as a service, but quickly recognized that their customers also needed storage. Some of them began by building flash-based storage tiers to support high-performance workloads. However, as data scaled and flash prices spiked, it became clear that flash storage should only be used where it's necessary. As neoclouds scale, they need a more complete storage stack, flash where maximum performance is required and a hard drive based capacity tier for everything else. With five exabytes of storage and almost two decades of technical optimization, we believe Backblaze has built the most efficient hard drive-based capacity storage platform available. And neoclouds, wanting to get performance scale efficiently, are choosing Backblaze. We estimate that this neocloud demand for capacity tier storage represents a $14 billion market opportunity by 2031. Our strategic agreement with CoreWeave, a more than five year multi-exabyte deal and the largest contract in Backblaze's history is the clearest proof this quarter that Backblaze can be the capacity tier for AI infrastructure. CoreWeave is recognized as the essential cloud for AI and runs many of the most demanding AI workloads in the world. As its platform expands, it is adding a variety of storage tiers that can scale rapidly and perform reliably and efficiently at massive scale. CoreWeave evaluated the available options and chose Backblaze for the software platform and operating expertise we have developed through years of managing large-scale hard drive-based storage infrastructure. CoreWeave is now the fourth major AI cloud infrastructure company to contract with Backblaze and we're in conversations with many of the leading other ones. As these AI infrastructure companies scale to broaden support of their workloads, we become increasingly relevant to them. Part of the CoreWeave agreement also introduces a new way for us to deliver that value. For Backblaze, this managed storage approach represents a capital light service model that brings our technology and operating expertise directly into a customer's infrastructure. This approach expands our opportunity to service customers in their regional data centers and sovereign cloud needs. Now, beyond AI infrastructure companies and to the broader AI market, we continue to see strong traction with AI native companies like HeyGen, Hume AI, Mirage, and many more. And this quarter, we continue to add to that list. AI companies are choosing Backblaze for our ability to scale fast. Last quarter, we highlighted a training data provider that signed a nearly $1 million deal in just 11 days. Less than a quarter later, as its business grew faster than expected, it added another $1 million commitment. AI customers also choose Backblaze for architectural freedom. It used to be that companies were okay just building inside one cloud, but AI technology is evolving rapidly. AI native builders are choosing from an increasingly fragmented set of cloud infrastructure. That requires the ability to use and move data to whichever hyperscaler, neocloud, inferencing cloud, or other AI infrastructure they need. Backblaze enables that through a combination of free egress, high performance throughput, and optimized networking between us and these clouds. The need for architectural freedom resulted in a six-figure deal with a customer building conversational AI models. They needed a cloud agnostic home for their training data. Expensive egress fees from their prior provider kept their data captive and limited what they could achieve. With Backblaze, they were then able to freely move their data to whatever cloud they wanted without the headache of calculating and worrying if egress fees will break them. And AI companies choose Backblaze for performance. We signed our largest B2 Overdrive deal to date, a seven-figure ARR deal with a frontier AI model developer. At the scale of data they work with, performance is critical, and B2 delivers high throughput at efficient price points. Together, scale, architectural freedom, and performance, all at an affordable price, are why AI companies are choosing Backblaze. And while AI is making this need especially urgent, it extends to nearly every company using storage at scale. In addition to making great strides moving upmarket, we also know the biggest companies start small and we're building for them, our technology advantage is one part of how we are strengthening our position. We're also working to make Backblaze the natural platform for developers and their AI agents to build on. This quarter, we shipped our SDK for TypeScript, the emerging language of choice for AI coding agents, released Genblaze, a generative media SDK, and built out a new set of tools after seeing developers turn to B2 to store their AI agent data. We also launched our multimodal focus generative media hackathon, which drove awareness of B2 as the storage layer for GenAI applications. In closing, we exceeded our financial expectations announced the largest agreement in our history, and delivered new wins and expansions across the AI market, including our largest overdrive deal to date. We also introduced a new managed storage approach that brings our software and operating expertise into customer owned infrastructure. But the bigger point is this, when AI scales, data grows, and that is good for us. When companies look to control AI costs, they come to us for that too. Growth or discipline, either way, we are well positioned to benefit and continue building a durable growth business. With that, I'll turn it over to Marc.
Thanks, Gleb, and good afternoon, everyone. Q2 was a pivotal quarter reflecting both strong operating results and the significance of our strategic relationship with CoreWeave. Revenue was $42.7 million up 18% year-over-year and representing our strongest growth in six quarters. Adjusted EBITDA nearly doubled year-over-year to $13 million, with the margin expanding by 1,200 basis points to 30%. Both results exceeded the high end of our guidance. These results demonstrate the benefits of our strategy. Based on our Q2 performance and outlook for the remainder of the year, we are raising full year guidance again. Turning to revenue, B2 accelerated to 34% year over year, our strongest growth rate in seven quarters. B2's strong performance was broad-based with almost every route to market and GTM lever overperforming, with strong performance in direct sales bookings, continued self-service momentum, and increased usage from larger customers. We also signed larger and longer duration commitments, increasing RPO. The price increase implemented on May 1 contributed about 8 percentage points in B2 growth. While churn from the price increase was anticipated, it did not materialize. Excluding that impact, underlying growth continues to show strength. Sequential B2 ARR increased by $20 million, of which the price increase drove $9 million of the $20 million. B2 ARR reached $113 million, an increase of 39% year over year. B2 net revenue retention was 113% compared to 114% last year. We also continue to make progress upmarket. Customers contributing more than $50,000 in ARR increased 57% year-over-year to 235, and we closed four deals valued at over $500,000 this quarter, including three AI-related wins. That progress is also showing up in the size and duration of customer commitments. We added approximately $320 million in RPO during the quarter, including $313 million from CoreWeave, net of the $22 million in warrant values. This increase provides greater visibility into contracted demand as more customers enter into multi-year agreements with committed minimum spend. We retain additional upside as usage above those minimums is built on a consumptive basis. Computer backup revenue declined 2% year-over-year, better than expected, as churn initiatives and targeted customer acquisitions help stabilize performance. The business continues to generate recurring revenue and cash flow, and our focus remains on retention, operating efficiency, and margin improvement over time. Moving on to total company gross margin, it was 63% in Q2, benefiting from the B2 price increase and continued operating efficiency, partially offset by higher hardware infrastructure costs. Operating expenses increased 6% year-over-year, well below revenue growth. As a percentage of revenue, operating expenses improved by 800 basis points to 73%, demonstrating continued operating leverage. We expect to make targeted investments in R&D while continuing to reduce operating expenses as a percentage of revenue. That operating leverage also translated into an adjusted free cash flow margin of 8% for the quarter, even as we continue to invest in infrastructure for 2027's committed demand. We ended the quarter with $50 million in cash and marketable securities, up from $45 million in the prior quarter. We also increased our available and unused capital lease lines to over $150 million. While our new lease lines are generally at lower interest rates, we will continue to look for ways to optimize our cost of capital. Before turning to guidance, I want to note that we filed an S-3 today to register the warrants issued to CoreWeave in connection with our agreement. The warrants reflect the strategic and mutually beneficial nature of the relationship and align both companies around the long-term success of the agreement. Moving on to guidance. For Q3, we expect revenue to be in the range of $44.4 million to $44.8 million. We expect adjusted EBITDA margin to be in the range of 27 to 29%. For the full year, we are raising revenue guidance to a range of $172 million to $174 million, up more than $10 million from our prior range of $161.5 million to $163.5 million. At the midpoint, this revised guidance represents approximately 19% in overall year-over-year growth, up from the previous 11%. Consistent with our guidance philosophy, the raised outlook reflects Q2 actuals and greater visibility from contracted demand. Our guidance excludes variable usage above contracted minimums and potential deals greater than $500,000. We are also raising our full year adjusted EBITDA margin outlook to 27% to 29% from 23% to 25%. Looking ahead to 2027, based on the B2 underlying business fundamentals, CoreWeave's minimum RAB, and the previously announced $15 million plus TCV deal, we expect B2 revenue to grow over 40% year over year. This is early directional commentary, not formal guidance. We will provide our full 2027 outlook in February. We wanted to give investors visibility into the contracted demand already supporting growth beyond this year. Turning to capital investments, we are accelerating CapEx in the second half of 2026 and into 2027 to build the required capacity to support signed customer commitments. Using capital leases, we expect to be adjusted free cash flow neutral for the full year despite the increase in CapEx. Our CapEx track record demonstrates how we make use of these assets for well over six years and deliver healthy gross margins. Our CapEx break-even is less than 24 months. Moreover, the managed storage portion of the CoreWeave agreement is delivered on customer-owned hardware, so we have no CapEx requirement for the managed service. B2 revenue growth accelerated, larger customers continued to expand, and contracted demand increased our visibility. On a rule of 40 basis, we are proud of achieving a combined B2 revenue growth and adjusted free cash flow margin of approximately 42%, up from 18% a year ago. We entered 2026 with a clear objective, demonstrate that our business can grow efficiently and generate profitable operating leverage. Q2 showed clear progress against that objective. With that, operator, please open it up for questions.
Questions and answers
Your first question comes from the line of Mike Cikos with Needham. Mike, your line is now open. Please go ahead.
Great. Thank you to the team for the question here. Congratulations on the quarter. Marc, I was hoping to start with you on this outlook here. I'm really just trying to get a better sense. Obviously, you have this large deal with CoreWeave that you had announced, and congratulations again on getting that over the finish line and the sheer size of it. Could you help us think about to what degree the improved calendar '26 outlook is tied to the ramp for the minimum commitments from that contract or anything on the calendar '27 to support that 40% plus outlook we're putting out there for B2 Cloud?
Yes, sure, Mike. Can you hear me fine?
Yes, I can.
Okay. Yes. As it relates to the second half of 2026, the $10.5 million increase is benefiting from a broad mix of factors: the business performing better, the Q2 beat of $2.7 million, the price increase, and the CoreWeave ramp. None of those factors dominate; it’s a healthy combination. And just a reminder, our guidance approach remains consistent — we stick to contracted minimum spend by customers and do not project above that. Even though we just won a deal greater than $0.5 million, we are not projecting more per quarter so we remain consistent with that approach. That continues through 2027 as well. With that, I would say B2 should be well on track to grow at 40% or higher for the rest of this year and in 2027. Regarding the ramp, the CoreWeave ramp happens over the coming year and reaches their minimum around mid-2027, and that assumption is included in these numbers.
I see. Thank you. Thank you for that. And maybe a question for Gleb. Gleb, obviously, you guys have made some pretty significant changes to the go-to-market in the last year. And I know we're starting to see that specifically in B2 on the NRR front or I remember last quarter, you guys were talking about pipeline from existing customers. You also have the new CRO in place now for, call it, a quarter or so. Can you just give us a status update on where we are with that go-to-market transformation? I guess, what are the findings for today versus 90 days ago.
Yes, thanks, Mike. So as you mentioned, we've been undergoing the GTM transformation. Anuj joined as our new CRO. He's been doing a great job as he's coming on board. We've also had a broader people, processes, systems rework. So in addition to Anuj, we have a new sales development leader and a head of RevOps and some of the other leadership functions ahead of an operational strategy under him that he's worked with before. So we've brought up, I think, the team in terms of the GTM site. We also, as you know, we're undergoing a big systems effort. A lot of that is done. We are still continuing to invest, especially with some of the AI technologies that are out there. So we're leaning in on some of those. I think one thing that we look at is, obviously we're excited about the CoreWeave deal, but for as far as the GTM side of things, we have almost 50 more customers that are in the $50,000 plus ARR group than we did last quarter. And that's almost as many or about as many as we add in a year historically. And so I think that more broad-based repeatability is a good sign that the GTM is working. Now obviously I expected that'll fluctuate up and down but the general direction I think of that execution is showing up.
Your next question comes from the line of Ittai Kidron with Oppenheimer & Co.
Congrats again. Great numbers. Great to see the acceleration. Gleb, I guess I have a little bit more of a bigger picture here. The announcement of CoreWeave clearly is quite unique, I guess, in its size and its messaging. I was kind of wondering, like I hear your point on the growth in the $50,000 accounts. But if you try to look at the AI cohort specifically, does the deal with CoreWeave, does it generate more interest from customers or less? I'm kind of wondering if customers view that relationship as something that potentially ties you perhaps too closely to CoreWeave for people to do business with you? How do you think about that?
Yes, thanks, Ittai. So the announcement with CoreWeave has been great. I mean, first of all, I would say the CoreWeave team has been great to work with. But also the announcement has helped, I think, elevate Backblaze as a key player in the AI infrastructure stack. I went to this AI infrastructure conference in Europe, I think about 1.5 months ago. And I'll just say that the conversations that I had were consistently, hey, we're excited by this deal that you did. We may not be as big as CoreWeave, but this seems very relevant to us, very applicable. Can we talk to you about it? How can we leverage your technology to do the same. We're starting to see storage as a key need for us. So I think, obviously, we provide a platform for a variety of customers, both in the neocloud and AI infrastructure space and also to the AI natives themselves. So we have lots of start-ups and developers and larger sized direct AI companies that see the CoreWeave deal and see that it's a stamp of validation. On the AI infrastructure side, there are certainly competitors to them, but it's a big market. It's a growing market, and I think a lot of people are trying to figure out how best to solve it. The other thing I would just mention is we introduced this managed service approach or managed storage approach, right? So we do that for CoreWeave, but we're also doing that as an offering for others. And a number of the conversations that we're now having with these other AI infrastructure companies is them being interested, not only in us providing infrastructure ourselves for them, but also providing this managed storage in their data centers and their sovereign cloud environments.
That's great, great to hear. And you kind of set me up for the next one, I guess. On this topic of managed storage, when you look at your pipeline, I know clearly part of CoreWeave, this was part of the CoreWeave transaction as well. But when you look at your pipeline, we look at the conversations that you're having with customers, is this common that people are looking for this? Or this is going to be more the exception rather than the rule. And then also when you talk about the 2 wins, for example, you have this quarter that you highlighted the conversational AI company with 6 figures and the frontier model company with 7 figures. Is there a way for you to have insights into their business to understand how much more opportunity you have within those organizations to kind of expand your use cases with them?
Yes, so in terms of the managed storage side, it's obviously early, right? That's the newer approach that we're offering. What I will say is that in the past, we've had prospects come to us and express interest in us doing that, and in the past we haven't done it. But with CoreWeave, we're doing this in partnership with them and offering it out to the broader market, so we're having conversations now. I would say there are probably half a dozen of these managed storage conversations that were fairly active. This is not going to be every customer doing this, in part because it requires a fair level of sophistication on the customer side and a fair amount of scale to make it worth doing. But for the larger AI infrastructure organizations and, frankly, not just AI but anybody who needs large-scale capacity storage, I think managed storage is a good approach. So I think we'll see a number of those. It won't be the predominant number on a volume basis, but I think those will be larger opportunities.
I'll add to, I mean, if you want to jump in, Gleb, on a second question, whether we have visibility. I'll let Gleb answer on how much visibility we see into the large type of customers. But generally speaking, what we notice in Q2 is a lot of our, we said 4 deals greater than $0.5 million, a lot of those were expansions. So we're seeing their appetite and needs. We said that before that AI companies grow a lot faster in the data appetite, and so we're seeing that profess itself.
Yes, and maybe just the one thing to that is, I mean, in the world that we were in at IPO, we were almost entirely self-serve. So we had very little forward insight into what was happening with the customers because they would just sign up and pay on a consumption basis. Now, as we're heavily investing in the sales-led side of the business and working with these larger opportunities, the team is actually actively in discussions with them, working around what their needs are, what their plans are, and kind of co-planning together. So we do have better visibility as well as the commitments themselves.
Great stuff. Congrats.
Your next question comes from the line of Jason Ader with William Blair.
First question, just on the CoreWeave deal, could you give us a sense of what the gross margins are going to look like relative to your traditional B2 business?
Yes, Jason, this is Marc. For our gross margin, I mean, for the time being, we're pricing everything on all deals to keep it in and around where it is. So even with CoreWeave, despite a lot of scale, there shouldn't be that much detriment to the current 63%. What I would say is between the CoreWeave deal, the previous deal, the $15 million plus TCV deal, all these committed contracts we're signing up, it does increase our CapEx needs. And so our CapEx for the year will be between 55% and 65% of revenue. And the reason why I mentioned that is we're putting a lot of CapEx out, so depreciation will start and then we'll start ramping up the customers in terms of that revenue spend. So that 2 quarters or so before they get ramped on that capacity, you'll have more depreciation. So there could be a few hundred basis point setback to our gross margin, but then it should recover after that.
Got you. And then if over time you end up doing more of the managed storage option with CoreWeave, is that, I'd imagine that would be a significant boost.
Yes. I think that both of those are currently similar-ish in gross margin, although as we think about our capital-light approach with the managed storage overall, we think that, that certainly has the possibility of being a higher-margin offering over the longer term.
Okay. Maybe I'm not clear. I thought you were just selling software there. What are you selling there in the managed storage approach of theirs if it's delivered on their hardware?
Yes, good question. It's a managed storage offering. The way it works is they provide the data center space and specify how much storage they want, we provide the bill of materials, they buy the equipment and hand it to us, but that part of the data center is cordoned off for our purposes and it's our people working there. Our team racks, stacks, and manages that part of the storage stack using our software. With CoreWeave, we'll manage multiple parts of their storage stack: we'll manage certain components, they will own the equipment, and we will provide the people.
Got you. Okay. So the main cost for you is people in that scenario.
Exactly.
Okay. Got you. All right. And then one last question for me, for Marc. And for Gleb, I guess just as we think about going forward and the CoreWeave situation and the CapEx needs that you're going to have over the next few years. What are you contemplating in terms of capital needs? Do you have enough capital today to be able to meet the needs of this build-out? Or are you going to have to raise more capital?
Yes, Jason. I'd say between our cash balance, over $150 million of available capital lease lines, and our operating cash flows, which have become really healthy due to operating leverage, we're well set to proceed the way we're doing it via capital lease lines. We'll always evaluate all options and do what's best for shareholders, but we're set to proceed as is now.
Your next question comes from the line of Jeff Van Rhee with Craig-Hallum Capital Group.
Congrats guys, just breadth, really impressive to hear what you guys did. Maybe spend a second on the Frontier model and the win there. I'd love to hear a bit more color, competition, maybe a little more particulars around use case, duration of deal. And then I think you just said 7 figures. I mean, any sense you can dial that in a bit? Are we talking mid-single-digit to some figures, upper or lower? Yes, color around the frontier model would be great.
Yes. Thanks, Jeff. Good to chat with you. What I'll say is that this one, similar to some of the others, follows the same pattern. They have storage and data sets they use to build their models and were having a couple of issues. One was that as the size of the data increased, they were actually hitting quota ceilings. The cloud provider they were working with had trouble providing the amount of storage they needed, so they were hitting certain ceilings. They also had performance requirements — they needed to move that data at high throughput to places where they would build the models. Because of those two needs, they required both the ability to scale and the performance of B2 Overdrive. They were on another cloud provider previously, so they were familiar with the model and switched to us. Like the other use case we've always described, they're not running GPU model training directly off the data on Backblaze. They're using Backblaze to store the big capacity-tier data set and then moving it to the flash tier next to the GPUs when they're ready to train, while keeping it long term and at scale on Backblaze. The pricing is the Overdrive pricing, so it's higher than the $6.95 self-serve price on the website. Finally, it was a large opening commitment and a building block to start from, one they expect to expand significantly.
Yes, I would think so. And then maybe just a follow-up back to the managed storage offering. So when you're selling to a neocloud or one of these larger AI players, what is the delta between when they want the managed offering versus a white label offering? I mean I understand there's sort of some geo data sovereignty issues. A lot of things probably come into the play, but like why one over the other, traditional B2 versus managed storage?
Yes, the main reasons why most people want us to take care of it on our own infrastructure. And the reason for that is because it's fully taken care of, right? They don't have to worry about it. They don't have to think about it. And even for neoclouds, the neoclouds obviously range in the level of sophistication and their level ability to operate a full platform. You have CoreWeave on one extreme of a company that is very, very good at managing the whole infrastructure and technology and stack and everything else. And you have others who are just brand new. They have data centers, they have GPUs, but they're still building out all the other pieces, right? And so for many of them, they prefer to just have us fully take care of it. The managed storage side comes into play when they want to have the physical data in their own data centers and that has sometimes the conversation has come up because they have data centers in regions that we aren't and so they would like the data there sometimes because they have a sense of they would like a more sovereign experience with their data. And then for some of them, the conversation has simply been that they would like to actually own the assets on their balance sheet. So those have been the reasons why they go one way or the other.
Your next question comes from the line of Erik Suppiger with B. Riley Securities.
Congrats. Great quarter. Couple questions. One just on the go to market. Have you hired most of the executives across the go to market team that you need at this point and then secondly I think you talked about the CoreWeave business reaching a minimum level, meaning meeting the minimum commitment level in mid '27. Does that mean that we can assume that one you're kind of at the one-fifth of the $335 million, which is about $65 million run rate. Does that imply that you're reaching about a $65 million run rate by mid-'27 on that CoreWeave agreement?
Yes. Thanks, Erik. So on the GTM side, yes, we've hired the Chief Revenue Officer. We've hired the Head of Development, Head of RevOps, the Head of Customer Success, and the Head of GTM Strategy Operations. So we are obviously there. There may always be additional folks, but in practice I would say we've got the team in place. And one thing I'll mention too is, I think when I was reflecting earlier on our journey, you were with us when we went public. When we went public, our average customer was a self-serve customer that paid us less than $500 a year. We said our goal is to move upmarket, become more of this core infrastructure for startups, companies, and enterprises. We started signing companies that were paying us tens of thousands, $50,000. At some point, we signed our first $1 million deal, then we started highlighting roughly a $1 million deal per quarter. Then we had a $15 million deal that we announced in February and then this $335 million deal that we just announced. So obviously, it was quite a journey to go from a primarily self-serve company doing mostly less than $500-a-year deals to a company that is able to service $1 million, $10 million and multi-hundred million dollar deals. But I think we're now in a great place with a great team and processes and systems to go and execute against this opportunity.
Yes, and on the second question, Erik, let me dive into the second question. So the CoreWeave deal has two components. As you know, there's working off of our platform. That's one component, and that was all disclosed in the June 23 deal. And then there's the managed service component. So roughly speaking, it's like almost a 70-30 split. So when I said we would reach the minimum it's relates to that 70%, not the 30%. The 30% would come afterwards because the managed service has a different kind of ramp. And then the other thing to keep in mind is the warrants are a contra revenue. So that's why you shouldn't take the just a $335 million times 70. You also have to deduct the $22 million value of a warrant.
Can you just expand on that? You deduct the $22 million for the warrants. Is that across the five years? Is that a straight up division?
Yes, exactly. So the warrants just they follow the revenue. So there's five years for both components of the deal. So if you take that 70% of the $313 million, which is net of the warrants. That one ramps up over the first 12 months, and then once it's up to 12 months, then it starts operating at that minimum.
Thank you. Your next question comes from the line of Eric Martinuzzi with Lake Street Capital Markets.
I wanted to revisit the upward revision to the 2026 guidance. The way I understood it, Marc, you talked about three reasons for the upward revision, and they were all equally weighted. The business outperformance to date and the price increase and the CoreWeave ramp, the business outperformance, is that primarily going to be those greater than $500,000 ARR transactions?
Yes, Eric, it includes that, but it also includes the self-serve product-led growth. We did that price increase on May 1, so we anticipated some churn. We really haven't seen any churn. In fact, we've seen an acceleration of people signing up and the ARPU per sign-up is higher. So yes, I think we're seeing good momentum across almost all routes to market and all go-to-market levers. It's pretty broad-based. And Gleb mentioned the 50 customers that went to over $50,000 in ARR. You can see our RPO increases every quarter. We're getting customers committing to one-year or multiyear contracts. Overall, it's broad-based and reflects general business health. We're seeing really healthy acceleration.
And then on the CBU, I think you said last quarter that you were expecting it down, what was it 3% or so or low single digits? Is there any change to that expectation for the year in the new forecast?
Yes, that business is kind of like we talked about, Eric. It's a good business; customers like the experience. It's cash-flow-generating and helps fund some of the B2 growth. We are spending time and investing in it, but it still faces overall market headwinds, so we still expect a single-digit decline. It performed a bit better this quarter than expected, partly because of efforts on churn mitigation and customer acquisition, but it remains a single-digit declining business.
Thank you. Your next question comes from the line of Rustam Kanga with Citizens.
Great. Thanks, Marc, and glad for taking the question. Great to see the sustaining momentum here. My question is just around CoreWeave. Can you help frame the extent to which that recent win is helping accelerate discussions with other neocloud providers evaluating HDD-based storage tiers? And are you finding that the best conversations are those who have already experienced challenges with the costly flash storage approach or is it better or more effective to cut them off at the pass and approach those who are even yet to begin a DIY approach?
That's a great question. About a year and a half ago, when we launched B2 Overdrive and B2 Neo and leaned into the idea that neoclouds and AI infrastructure companies would need a capacity layer for storage, it actually felt counterintuitive. We assumed the best approach would be to target providers that didn't yet have storage. What we've found instead is that the most engaged conversations come from those that already have storage. They feel the pain of relying solely on flash-based storage or trying to build solutions themselves — their customers expect them to service these workflows, but they struggle with the scale and cost. Providers that don't yet have storage tend to have their customers go elsewhere for those workflows, so they aren't feeling the need yet. As they begin broader conversations with customers about servicing their needs, that demand becomes clearer. In short, neoclouds are moving toward needing both a capacity tier and a flash-based tier of storage. It's not one or the other; they need both to serve customers, and we're a strong solution for the capacity tier.
Perfect. Thanks. And then, Marc, just you gave the caller on the CapEx of 55% to 65% of revenues. Was that comment more for the back half of this year, or do you expect that to hold through 2027? Just help us think about if it would ramp from that level, or if that comment was applying to this year and next year.
Yes, Rus, that's for this year. That's for 2026. Too early to give 2027, frankly, mainly because the prices of this hardware changes pretty quickly and obviously our growth outlook keeps accelerating. So that number will for both those reasons will change the '27 number. So for the time being that 55% to 65% is for the revenue as a percentage of revenue for 2026.
We have reached the end of the Q&A session. I will now turn the call back to Gleb Budman for closing remarks.
Thank you. So AI is reshaping the entire infrastructure market and Backblaze has built exactly what this moment demands. Storage is that durable layer beneath AI. I'm really pleased with how our team has stepped up to capture this generational opportunity. I want to thank all our Backblazers for leaning in and to our customers and partners and investors for joining us on this journey. Finally, we look forward to seeing many of you at our Investor Day in New York City and on our live webcast on September 9th. Please RSVP to ir@backblaze.com if you'd like to join the in-person event. Thank you all for joining today's earnings call. Operator, you may now end the call.
This concludes today's call. Thank you for attending. You may now disconnect.