RUM Group Inc. (RUM) Earnings Call Transcript

Greetings and welcome to RUM Group Incorporated's third quarter 2026 earnings conference call. [Operator Instructions] As a reminder, this conference is being recorded. I would now like to turn the conference over to your host, Shannon Devine. Thank you. You may begin.

Thank you, Rob. I'm here today with Chris Pawlowski, founder, chairman, and CEO of RUM Group, and Mike Masci, CFO. A press release detailing our third quarter 2026 results was released today and available on our investor relations website. Before we begin the formal presentation, I'd like to remind everyone that statements made on this call may include predictions, estimates, or other information that may be considered forward-looking. All forward-looking statements are made only as of the date of this call and should be considered in conjunction with the company's cautionary statements in our earnings release and the factors included in our filings with the SEC. Future company updates will be available via press release and the company's identified social media channels. I will now turn the call over to RUM Group's founder, chairman, and CEO, Chris Pawlowski.

Good morning, everyone, and thank you for joining us. Before we begin on this Veterans Day, we want to thank every man and woman who has served and their families, because the freedoms we build on every day, including the freedom to speak openly, were earned by their sacrifice. This was a very strong quarter for RUM Group. We posted record revenue, we beat guidance, and we were profitable on an adjusted EBITDA basis for the first time as a publicly traded company. Our company is ideally positioned for today's market. We have real assets running real AI workloads for real customers, generating real cash. We believe we have one of the least levered balance sheets among the larger public neoclouds, and we have a Quake AI team that is as experienced as anyone in our industry. I also want to be direct with you today about what's changed in the market, what it means for our projects, and how our management team is navigating it. Let me tell you how I view the Q3 results. Let's start with record revenue of $96.6 million, nearly four times what it was a year ago, but the story isn't just the size, it's the mix. In its very first full quarter as part of RUM Group, Quake AI contributed more than 70% of our revenue. That is a brand new profile for this company. A year ago, our revenue came almost entirely from video. Today, AI infrastructure is the core of our business, the engine of our growth. On guidance, Mike will walk you through the details. What matters more to me is that every quarter we operate as a combined company, we get a clear view of how this business runs, and we'll keep sharing that view with you as it sharpens. Profitability is the part I'm most proud of, and I want to put it in context. In the early years, I bootstrapped this company. Every dollar mattered because there wasn't another one coming. Then, as an early public company, we made a deliberate choice to invest in the platform, in our team, and in the assets that got us here. Those were the investment years, and we were open with you about that. Now, RUM Group is maturing into a profitable business, one that generates returns for you, our shareholders. That's the turn this quarter represents, and we're just getting started. The turn is being driven by Quake AI. We run approximately 22,000 GPUs today, one of the largest GPU estates in Europe. As of the third quarter, we are running at 85% utilization, which means we are nearly sold out. This is the cash engine of our business. And the market is moving in our favor. Supply is low, demand is high, and we don't see that changing for the foreseeable future. That manifests as increased pricing, which has been observed across the GPU market. With the majority of our contracts coming up for renewal by early 2027, management's plan is to optimize our contracts for the market we are in. To be clear, we and our customers expect prepayments and higher pricing. Now, as I promised at the top, let me turn to the market and our projects. I view AI infrastructure and in particular AI compute as similar to the oil industry. And as markets realize the value and returns, I believe capital will keep flowing into it. But over the past several weeks, as you've all seen in the news, the financing markets have become quite difficult for data center and AI buildouts. Execution challenges at some neocloud and data center builders on large-scale projects have put a spotlight on permitting, power, and team experience. Lenders and investors are doing more rigorous diligence on every customer and every operator. So let me be direct about what that means for us. First, the message from our customers has been very clear. The need for us to deliver on AI compute has never been greater, and we believe our customers want to find a way for us to deliver that compute. Second, regarding our Maysville site, we are in active discussions with our customer on an alternative path forward to deliver that AI compute. While this path is still in the early stages, we're optimistic it could represent a more financeable project despite current market conditions while still delivering attractive economics. Notwithstanding the recent market challenges, we remain optimistic about the assets we have, including our GPU estate, which I mentioned earlier. As the political dynamic changes and as AI compute supply continues to become more scarce, I believe our GPU estate assets will increase in value, and I believe we'll see this continue in '27. The only debt we hold today is a favorable note from Tether, and our existing estate of 22,000 GPUs generates meaningful cash flow. Now, let's turn to the Rumble video business, where we are applying that same discipline. In the third quarter, we began a cost reset in video, tied to a refocus strategy. We expect to start seeing those savings through 2027. That strategy comes down to two core elements. First, product-led automation, driving monetization centrally through Rumble Studio. Second, creator-led economics. In October, we announced the new Rumble partner program, which gives eligible creators up to a 90% rev share on video ads, including host-read sponsorships. Together, these make the video business leaner, sharper, and well-positioned for growth. And while we implement the new model, engagement on the platform remains strong. During the U.S. midterm elections this month, Rumble hosted three of the top five largest U.S. independent streams, according to StreamCharts, at around 9:00 p.m. on election night. We also launched a long-term distribution and content partnership with the WWE, which makes Rumble the new home of the weekly premium program WWE Main Event. As we grow, diversify, and lean into a more automated model, I am excited about the path ahead for our video business. I'll leave you with this. We know the market feels uncertain right now, but as a leadership team, we could not feel more confident about the AI compute market, and we are excited about our future. Our revenue is nearly four times what it was a year ago. We're profitable. Our GPUs are running, and our customers are asking for more. You'll keep seeing us focus, execute, and deliver. With that, I'll turn it over to Mike.

Thanks, Chris, and good morning, everyone. This is our first quarter reporting in two segments, Rumble Video and Quake AI, as we committed to you in our last call. I'll walk through the consolidated results, then the segments, and then our outlook for the fourth quarter. Revenue for the third quarter was $96.6 million, an increase of $71.8 million, or 290%, compared to $24.8 million in the third quarter of 2025. Above that increase, $68.2 million came from Quake AI and $3.7 million from higher Rumble revenue. Cost of services was $43.6 million, up from $25.2 million a year ago. The increase was driven by $16.4 million of incremental data center expenses from Northern Data and $3 million of higher programming and content costs, partially offset by a $1.1 million decrease in other costs of services. General and administrative expenses were $39.5 million, up from $10.5 million. Northern Data contributed $24.6 million of that, mainly lease costs, payroll, and other administrative costs. Excluding Northern Data, the increase primarily reflects $3 million of higher professional fees. Research and development expenses were $7.7 million, up from $4.5 million, and sales and marketing expenses were $7.7 million, up from $5.1 million, with Northern Data accounting for part of the increase in both. Adjusted EBITDA was a positive $4.4 million, our first positive quarter as a public company, as compared to a loss of $15.1 million in the third quarter of 2025, and a loss of $16.6 million in the second quarter of 2026. The net loss for the quarter was $91.7 million, or $81.5 million attributable to RUM Group, compared to a net loss of $16.3 million a year ago. The increase is primarily non-cash. Amortization and depreciation rose to $85.8 million, including $39 million related to the fair value step-up of the Northern Data assets in purchase accounting. Quake AI generated revenue of $68.2 million and adjusted EBITDA of $23.5 million, and an adjusted EBITDA margin of roughly 34%. Revenue was primarily from GPU cloud computing services with a smaller contribution from hosting and co-location. Rumble generated revenue of $28.4 million, up 15% year-over-year, driven by $5.6 million of higher advertising revenue. That was partially offset by a $1.2 million decline in subscription revenue and a $700,000 decline in licensing, platform hosting, and other revenue. Rumble's adjusted EBITDA loss was $12.1 million compared to a loss of $10.9 million a year ago. Corporate and other costs not allocated to the segments represented an adjusted EBITDA loss of $6.9 million. We ended the quarter with total liquidity of $142.3 million, consisting of $134.4 million of cash and cash equivalents, and 94.68 Bitcoin, valued at $7.9 million. Now let me turn to guidance. When we gave our first formal outlook in August, I told you building a credible track record was a priority for me. Part of that is being clear about what's in our numbers. We reported $96.6 million against a range of $87 million to $93 million. But Q3 included items that we don't expect to repeat. So it is not the right baseline for modeling. Q3 included certain items that we don't expect to repeat in the fourth quarter, including our work with one customer as a part of a barter agreement, which had more accelerated revenue than expected with a little over $3 million in the quarter, but with a larger drop-off in Q4 than expected. Excluding those items, Q3 revenue would have been roughly at the top end of our range. That's the run rate where we're really working from. For the fourth quarter of 2026, we expect revenue between $88 million and $94 million. This will be the last quarter where our existing fixed H100 contracts signed back in 2025 are in place, and we will see some small impacts on revenue as we begin to transition certain clusters for their renewal at hopefully significantly higher rates starting in the first half of '27. This is accounted for in our guidance. As Chris mentioned in his remarks, we continue to work with our customers on key projects. And as a seasoned operator with one of the least levered balance sheets among the large public neoclouds, and an estate of 22,000 GPUs generating significant positive cash flow, I believe our Quake AI team is well-positioned to take advantage of the extremely favorable supply-demand dynamics in the AI compute market. With the changes in the financing markets over the past few weeks, that gives us an even greater need to have a disciplined operating and financial model. First, with increased cost of capital, it is critical that our GPU per hour rates increase. Chris mentioned the pricing dynamics across the GPU market, and with a majority of our H100s coming up for renewal by the end of Q1 '27, I am excited for the tailwind this dynamic is expected to have to our bottom line. Second, many public neoclouds have used dilutive financial instruments to raise project equity in support of key deals. With the increased demand and scarcity of compute, we believe our potential customers will be more willing to fund substantial project equity needs via prepayments. The bottom line is, in our Quake AI business, we are taking a disciplined capital approach. Our GPU rates are increasing, our margins are improving, and we believe we have the opportunity to leverage customer prepayments for project equity. We are taking the same disciplined approach in our Rumble video business with significant CAPEX reductions taken in Q3, including a reduction in force and other recurring expenditures. Rumble Video has significant growth prospects ahead, and with a refocused strategy and reset cost basis, we will actively manage towards a more profitable trajectory. In summary, Q3 was our first full quarter with Quake AI, which is now more than 70% of our revenue, and our first quarter of positive adjusted EBITDA. We will continue to focus first on execution, renewing our existing 22,000 GPU estate at better economics, executing on our new operating model to deliver project financing and execution, and grow this business in the disciplined way we've committed to you. That concludes our prepared remarks. Operator, we are now ready to open the line for questions.

[Operator Instructions] Our first question comes from Jacob Stephan with Lake Street Capital Markets. Your line is now live.

Hey guys, appreciate you taking the questions. Congrats on a nice quarter here. I guess just first I want to touch on the 250 megawatts that you said may slip beyond 2027. I guess your prepared remarks kind of implied that it would be on the Maysville side, but I wondered if you could just clear that up for me.

Yeah, hey, Jacob, and first just welcome, so glad to have you on the call. So we've been very disciplined about not publishing specific timelines on projects until we got to places of firm certainty, because we're a company, frankly, that executes what we say we're going to deliver to. So we mentioned in our disclosure, and you mentioned it before, we're not going to do project-specific guidance. But overall, we do expect a portion of the 250 megawatts to move beyond 2027. But we are still targeting a portion of that to land in 2027. As Chris mentioned in his remarks, we're working closely with our key customer on a new potential construct that we're very excited about. And so we'll provide you further details as we go along. But just going back to our philosophy, we are very much a we don't put out information until we know that we can hit it, and we are a high execution firm.

Okay, helpful. Maybe just on the H100 repricing, I think your numbers imply that it was like $1.40 or so this quarter. But maybe if you could kind of help us think through where you expect those rates can be repriced at and maybe how certain clusters kind of roll off into higher-price contracts over the next quarter or so.

Yes, sure, great question. So I'm not going to go into the specifics on pricing, but I can speak to the market and some of the dynamics of: you have a choice between often the duration or the length of the contract, and you have a choice between how much you choose to do in terms of prepayment and also then the rates. But what I can tell you is the prevailing rates of H100s today by any market metric that you see are firmly with a two-handle on them, and we definitely are seeing market rates that are similar in our discussions with our customers. So we'll provide more specific guidance later as we get commitments toward it, but we're definitely in a really good spot and we're liking where the market is today. We're going to make optimized portfolio decisions based on duration, prepayment, et cetera. And then to your second question, the vast majority of our H100s will have their contracts roll off by the end of Q1 '27.

Okay, got it. And maybe just last one for me, touching on the video business, you guys talked about the cost reset there. Wondering if you have some sort of annualized savings target or maybe any update on the video data licensing that we've talked about.

Yeah, let me do the first one and then I'll hand it off to Chris for the second one, because I know he's excited about that and that's a great opportunity. So to answer your question, we haven't released the exact numbers on what some of those actions will ultimately manifest, but let's just call them significant. Think of them in excess on a run-rate basis of 20% plus run rate on operations, and we felt like we can do that in a way that was still going to drive growth for the business through the automation and some of the constructs that Chris mentioned before. So a significant reduction getting us into an OpEx space where we can be positioned for growth.

Hey, this is Chris. I'll also add on the things we've introduced to the platform like Boosted and Premium, we see that we can push a lot of growth there with the 90% that we're giving on the other items. But the most important part, which will come in the second half of 2027 that we're gearing up for, is the potential for the creators to tap into data licensing on their IRL content and their video content. So we're in the process of designing and moving forward a marketplace that will allow AI robotic companies or AI companies in general to tap into that and grab content that creators may want to license to them on the spatial-temporal data front. So we're super excited about that. That's something that we haven't seen in the market. And we obviously have an enormous community that can generate an enormous amount of IRL content and video content across the spectrum. So our target for that is late 2027 to roll that out, and hopefully we'll see some more meaningful updates in the first half of 2027 as well.

Got it. I appreciate all the color. Thanks, guys.

Our next question comes from Param Singh with Oppenheimer. Your line is now live.

Yes, hi. Thank you so much for taking my questions. Firstly, if I could kind of dive into the cost of the 250-megawatt build-out and the financing on it. I wanted to understand how are you thinking about it? Because by my math, it should cost about $11 to $12 billion for, say, a Vera Rubin to build out that capacity. And how much are you thinking is going to come from customer payments versus, say, secure financing or equipment financing and other ways you're thinking about funding that growth? Thank you. And I have a follow-up.

Yeah, no problem. And hey, welcome, Param. So we haven't given specific CapEx guidance, and we don't plan to today. But to your question and to our overall operating model and how we're thinking of it moving forward, I think that customer prepayments are really going to be a large portion of the financing for projects that is necessary for the project equity. And then I do think that the debt markets, in spite of the fact that there is a potentially higher cost of capital, I think that as they become more discerning about the operational execution of the companies to whom they are issuing the debt, I think that's going to be really beneficial to folks like us who are experienced and have a trust and belief in terms of folks within the market. The other thing I would mention to you is that the 250 megawatts and some of the opportunities we're looking at, there are different flavors and styles of customers that we have. So not every deal we would paint with the exact same brush in terms of the operating model. I think the discipline will hold regardless, but I think that the circumstances may be different in terms of the mix between how much is needed to be prepayment, how much is financeable via debt, et cetera. So we'll continue to monitor and we'll release to the shareholders and investors more information as it becomes available.

Understood. And then let me take the other side of the equation then. Look, obviously you'll get a chance to reprice higher on the H100s, H200s, I think it's 30% to 50% higher, and you said it's a two-handle, so that makes sense. But as you think about the new capacity, that's predominantly going to be Vera Rubin or Grace Blackwell, if I'm not mistaken. So whether it's a short-duration contract or a long-duration contract, those are going for significantly higher than the Hoppers. So is this a $6 billion revenue run rate versus the $3 billion before because of the pricing in the market? Or how are you thinking about the longer-term opportunity of a consistent 250-megawatt build-up?

Yeah, Parm, to your point, and I think you know obviously we're not going to give the long-term run rate or guidance at this point, but to your point, if you look at revenue per megawatt at this point on next-generation NVIDIA GPUs—I know that space has given indications, but we have seen nothing but increases. I think that depending on how we work between the short-term duration versus the long-term duration mix that we have within our portfolio, we'll have a fairly material impact on what we think the run rates will look like. But we are certainly seeing significantly higher revenue per megawatt on Vera Rubin and next-generation GPUs compared to even the increased rates that we're seeing in our existing Hoppers from a revenue per megawatt perspective. Thank you so much, I'll get back in line.

Our next question comes from Jason Helfstein with Oppenheimer. Your line is now live.