Blaize Holdings - Earnings Call - Q2 2025
August 14, 2025
Transcript
Speaker 4
Thank you for standing by and welcome to Blaize Holdings' second quarter 2025 earnings conference call. At this time, all participants are in a listen-only mode. After the speaker presentation, there will be a question and answer session. To ask a question during the session, you will need to press *11 on your telephone. To remove yourself from the queue, you may press *11 again. I would now like to hand the call over to Vernice Pozynski, Investor Relations. Please go ahead.
Before we begin the prepared remarks, we would like to remind you that earlier today, Blaze Holdings issued a press release announcing its second quarter 2025 results. A corporate overview presentation was published and is available on the investor relations section of Blaze Holdings' website. Today's earnings call and press release reflect management's views as of today only and will include statements related to our competitive position, anticipated industry trends, our business and strategic priorities, our financial outlook, and our revenue guidance for the third quarter of 2025 and full fiscal year 2025, all of which constitute forward-looking statements under the federal securities laws. Actual results may differ materially from those contained in or implied by these forward-looking statements due to risks and uncertainties associated with our business.
For discussion of the material risks and other important factors that could impact our actual results, please refer to the company's SEC filings and today's press release, both of which can be found on our investor relations website. Any forward-looking statements that we make on this call are based on assumptions as of today and, other than as may be required by law, we undertake no obligation to update these statements as a result of new information or future events. Information discussed on this call concerning Blaze Holdings' industry, competitive position, and the markets in which it operates is based on information from independent industry and research organizations, other third-party sources, and management's estimates. These estimates are derived from publicly available information released by independent industry analysts and other third-party sources, as well as data from Blaze Holdings' internal research.
These estimates are based on reasonable assumptions and computations made upon reviewing such data and Blaze Holdings' experience in and knowledge of such industry and markets. By definition, assumptions are subject to uncertainty and risk, which could cause results to differ materially from those expressed in the estimates. During the call, we will discuss certain non-GAAP financial measures. These non-GAAP financial measures should be considered as a supplement to and not a substitute for measures prepared in accordance with GAAP. For reconciliation of non-GAAP financial measures discussed during this call to the most directly comparable GAAP measures, please refer to today's press release.
Speaker 3
Good afternoon, everyone, and thank you for joining us today. Blaize Holdings' second quarter of 2025 marks a clear inflection point from building technology to putting it to work. Our hybrid AI strategy has now moved from pilot validation to early-stage deployment, with contracted programs underway across industries and regions. For the last 12 months, we've been focused on validation. Now, we're shifting to execution at scale. Blaize is now deploying its technologies to advance sovereign AI strategies and power public safety networks. We're not only delivering chips, we're also enabling hybrid AI infrastructure that complements GPU-based systems powered by our programmable, power-efficient Blaze AI Platform. In the last few months, we've locked in two major contracts together worth up to $176 million to be fulfilled through 2026. These contracts send a strong signal that our product-market fit and our platform approach complement the global demand for hybrid AI.
The first is a $120 million contract with Starshine, building hybrid AI systems across Asia. The second is a $56 million purchase order rolling out sovereign-ready smart infrastructure in South Asia, serving as many as 250,000 cameras for smart traffic and public safety. Both rollouts complement existing GPU-based systems, with Starshine deployments starting in the third quarter and South Asia systems continuing to ship through the third and the fourth quarters of this year. These two significant wins are just the start. In addition to these contracts, we have a robust pipeline of over $725 million in active opportunities through 2027. Today's AI deployments are fundamentally heterogeneous, powered by a mix of hardware types across edge and cloud. That complexity often creates bottlenecks, especially as organizations run multimodal workloads at the edge. Blaize's hybrid AI approach answers that need.
Our purpose-built platform is built to complement GPU-based systems and to unlock better performance per watt and more responsive inference. While GPUs handle training jobs and complex AI in the cloud, Blaize handles fast, efficient tasks like processing live video, small language models, or sensor data right where it happens, delivering a compelling total cost of ownership benefit to customers. Customers aren't replacing infrastructure, they're augmenting it, and Blaize helps them do exactly that. To meet the growing demand for hybrid AI, we're introducing the Blaze AI Platform, a plug-and-play software and hardware stack that makes deployment faster and easier. At its heart is our programmable graph streaming processor, the Blaze GST, designed for low-power inference where data is created. We combine that with full-stack verticalized software, a low-code software development kit, and a growing partner network to make AI applications deployable out of the box.
According to Gartner's 2024 AI services forecast, the combined market across defense, smart cities, retail, industrial, and energy is over $112 billion today. In its 2024 AI server forecast, Gartner notes that inference systems will outnumber training systems by as much as six to one. That is why the Blaze AI Platform is positioned to deliver better performance, lower total cost of ownership, and more deployable solutions at scale. That is how we take customer demand and turn it into deployments quickly and at scale. It is what gives us confidence in the quarters ahead. From day one, my co-founders and I set out to create a programmable architecture to power the physical world, not just through the cloud, but with localized intelligence systems. Our vision is to be the trusted AI platform that helps people and machines act on real-time intelligence in every critical industry.
Our mission is to deliver energy-efficient, scalable AI infrastructure at the edge and in the cloud. That vision and mission are now coming to life. Our platform is already in the field, validated through the Starshine project and the South Asia rollout, where hybrid AI is moving from signed contract to real deployment. Whether it's smart infrastructure, autonomous zones, or next-generation defense, the common thread is clear: Blaze is enabling real-time intelligence wherever it's needed. Momentum is growing as projects move forward and new opportunities open in Asia, the Gulf, and the Americas. The hybrid AI platform today runs on our current generation chip, and we're already developing our next-generation chip to keep our strong edge AI position while deepening our reach into cloud-native enterprise and data center environments.
We believe Blaze is becoming a core part of the AI infrastructure stack, complementing GPU-based systems, enabling heterogeneous, multimodal AI workloads, and optimizing for power, latency, and cost from edge to cloud. This second quarter was a milestone for us. We secured $176 million in contracts, launched the Blaze AI Platform, and proved our go-to-market strategy across smart cities, defense, and sovereign AI. Hybrid AI is no longer just a roadmap item. It's in the field, helping to advance national infrastructure and delivering real outcomes. The product is ready, our partners are aligned, and customer momentum is real. With that, I'll turn it over to Harminder to walk through the financial highlights and updated guidance.
Speaker 1
Thank you, Dinakar, and good afternoon, everyone. I'll take you through our second quarter financial performance, what we've been getting done, and where we're headed next. As you heard from Dinakar, in the last six weeks alone, we signed $176 million in customer commitments. That's two deals: a $56 million purchase order for server and software deliveries to a South Asia company, and a $120 million minimum revenue contract for service to Starshine, covering markets across Asia Pacific. We booked $1.6 million of the South Asia order in the second quarter, net of partner commission, and there is about $4 million in backlog for the remainder of this year. Starshine shipments are planned to begin in the third quarter, with up to 25% of the total order anticipated to be fulfilled this year.
Cash collections should come in steadily, and most of our deliveries in the third and fourth quarters of 2025 are expected to be paid within the year. We believe these bookings alone largely de-risk our revenue outlook for fiscal years 2025 and 2026. Our pipeline growth is robust, now over $725 million, with $300 million of that in advanced discussions. We expect conversion to accelerate as we move into 2026 and plan to share news as contracts and purchase orders close. Now let's look at the second quarter by the numbers. I'm pleased to report that revenue came in at $2 million, net of around $200,000 in related party sales commissions. That's almost double the revenue reported last quarter and above the high end of our guidance range.
The South Asia purchase order includes around 15% of perpetual software licenses shipped with each server, and we also recognized $300,000 in AI Studio license revenue from another customer. Excluding inventory cost adjustments made in prior periods, our underlying blended gross margin for the second quarter was 64%. Gross margins are expected to dip for the next two or three quarters as we ramp up the Starshine contract because of third-party hardware. Research and development expense of $9.6 million in the second quarter included a non-cash stock-based charge of $3.2 million. The underlying cost of $6.4 million was $700,000 lower than the prior quarter's underlying cost of $7.1 million. This reduction was largely due to the uneven nature of third-party costs related to the development of our next-generation chip. We continue to actively manage our labor costs by optimizing resources in lower-cost geographies.
Selling, general, and administrative expenses, excluding depreciation and amortization, and stock-based compensation, grew slightly to $8.6 million in the second quarter, up from $8.3 million in the prior quarter. The primary drivers were higher legal and financial advisory fees and external marketing costs, offset by savings in labor costs. We plan to selectively invest in our go-to-market capability in the regions where we're experiencing highest demand for the hybrid AI platform. Net loss for the second quarter of $29.6 million was over $118 million lower than the prior quarter, which included significant one-time and non-cash accounting adjustments related to the merger. Cash ended the quarter at $29.1 million, including funds in escrow. In July, we entered into a common stock purchase agreement with B. Riley. This agreement gives us the flexibility to raise equity when we want to.
We will continue to explore capital formation strategies to fund capital needs for future growth opportunities. Coupled with anticipated receipts from customers, we believe that our cash runway supports the commercialization of the two announced contracts and engagement of third-party design partners to begin developing our next-generation silicon. Since our last earnings call, here's what I'd highlight. First, we secured $176 million in contracts and purchase orders. South Asia deliveries are underway, and we anticipate the first shipments for Starshine to start in the third quarter of 2025. Second, we launched our hybrid AI platform, which is resonating strongly with customers serving multiple use cases. This is no longer a roadmap item. It's being deployed in national and enterprise infrastructures and shaping real-world outcomes.
Next, our qualified pipeline now exceeds $725 million, with $300 million in higher confidence deals expected to contribute towards more predictable revenue growth in 2026 and beyond. Finally, we continue to maintain cost discipline, investing where demand is strongest, and have capital formation strategies in place to fund growth. Thank you, and with that, we'll now open the line for questions.
Speaker 4
As a reminder, to ask a question, you will need to press *11 on your telephone. To remove yourself from the queue, you may press *11 again. Please limit yourself to one question and one follow-up. Please stand by while we compile the Q&A roster.
Speaker 3
Good afternoon, everybody.
Speaker 4
Afternoon.
Speaker 3
Sorry, Speaker, we just needed to pass it back to Dinakar for some closing remarks.
Speaker 4
Okay, thanks for saying.
Speaker 3
Good afternoon, everybody. As we get started, I wanted to share something we're excited about. Our client, Starshine, recently published a case study featuring Blaze Holdings, along with a great video from their CEO, Matt. It's always energizing to see our work and vision showcased. Matt shared the video link with us, and we've posted it on the Blaze Holdings website and our blog for anyone who would like to check it out. With that, I'd like to hand it over back to you for Q&A.
Speaker 4
Thank you, sir. Our first question comes from the line of Gail Lauria of D.A. Davidson. Please go ahead, Gail.
Speaker 5
Thank you. Good afternoon. First question is just to check some math here. If we have $176 million, it will be delivered by the end of 2026. Based on this year's guidance, that would imply that there's $140 million that should still go into 2026 before any additional wins from any other contracts that could be secured between now and the end of next year. Is that right?
Speaker 3
That would be correct, yeah.
Speaker 5
Okay. The second part is more about the architecture and the market opportunity. It sounds like the architecture for Starshine is, I think you referred to it as a hybrid architecture where you're putting your product side by side in a server with NVIDIA GPUs because your inference is so much more efficient. That goes into the data center. Unlike other projects that you were going to deliver on the edge, this architecture puts you in the data center. Does that mean that you now have a bigger incremental opportunity within data centers as opposed to the opportunity at the edge?
Speaker 3
Absolutely, Gail. It is a combination of, for example, the South Asia contract that we announced is also with a sovereign AI provider, and they are doing traffic management use cases where they have a box right behind the camera and also a server in an on-prem cloud that can do analytics for traffic. What we're seeing in the bigger picture is customers primarily care about ROI for a project, and this is where they're looking at hybrid as the right strategy, where they do GPU-based systems for part of the problem, and then they need part of the AI stack. Blaze runs more efficiently for a TCO advantage, for a cost advantage, power advantage, and this is where they complement GPUs with Blaze, and that's the momentum that we're seeing, and that's what we're referring to as hybrid AI.
You are right that it's also extending our reach into the data center.
Speaker 5
That's great. Thank you.
Speaker 3
Thank you.
Speaker 5
Our next question comes from the line of Richard Shannon of Craig Hallum Capital Group LLC. Please go ahead, Richard.
Speaker 2
Thank you. Thanks, Dinakar and Harminder, for taking a couple of my questions here. Maybe let's just ask more a tactical one here, following up the guide as well as a couple of the past questions here. I guess the first part of this, you were talking about a dip in gross margins here in, I think, the next couple of quarters. Maybe you could nail this down a little bit better, help us out what kind of levels you're thinking of, and then how much of the sales here come from these two large contracts from Starshine and from the Southeast Asian customer?
Speaker 3
In your second question, for clarification, Richard, do you mean in 2025? How much of those contracts are in 2025?
Speaker 2
I meant in the third quarter, but if you'd like to go for the whole second half, that'd be great too.
Speaker 3
Yeah. It's going to be a combination. If I take your second question first, between the $56 million and the $120 million, there'll be more of Starshine this year in proportion, and of course, it'll be proportionate as we go into 2026. The gross margin dip will depend on the mix of not just the deliveries of these two contracts. Both have third-party hardware components in them, but the margins on the Starshine third-party components are slightly lower. I can't quantify it just yet. It'll all depend on the mix, but what we do know, what we can say is that it will be a little bit lower than what we are seeing today. What will also and can also offset that is the more software licenses and so on that we deploy in 2026. That should counter some of that reduction from the Starshine contract specifically.
Speaker 2
Okay, perfect. That's a very helpful detail here. My second question is regarding the MOU with the UAE entity they announced last year. I ask this partly, Dinakar, because a couple of times in your prepared remarks, you talked about defense applications being proven out here, and it smells something like what's going on there. Maybe if you can give us an update of where that sits, if it's still in the pipeline, when you expect that to come across the finish line.
Speaker 3
Sure. Our defense pipeline is actually growing, but let me specifically address your question with regards to the MOD status. The delivery and revenue recognition of that particular MOD project will follow the customer's deployment schedule, which currently targets 2026. Given the size and immediacy of the Starshine and the South Asia orders, we're actually prioritizing these programs to deliver a recognized revenue in Q2, Q3, and Q4, and that's where we are. On top of that, we are engaged with defense industry in various, this is part of our pipeline, larger pipeline, including use cases like drones, as well as video security, perimeter security. As these further materialize, we'll be sure to announce them.
Speaker 2
Okay, great. One last question. I will jump out of line here. The pipeline, $725 million, I think the last quarter you talked about, I can't remember what it was, $400 million or something like that. A nice increase, and I would assume we take out the $176 million that the contracts you've announced since then here. Maybe you can tell us about where the new elements of the pipeline sit, and also with these, I think, $300 million in late stage here, how many different contracts are we talking about, and would it be reasonably expected to see those close sometime this year?
Speaker 3
Yes, you're right. The $176 million is outside of the $725 million. We've always maintained a very deliberate nature of how we qualify our opportunities. It's really playing to the advantages of Blaze, the programmable device, the low latency, high performance, and low power consumption. There are probably around 20 to 40 different applications or customer engagements in the $725 million, so they vary in size. As we would have discussed in earlier conversations, we approach what we call a beachhead customer. We take a beachhead customer approach with our ISVs. When we deal with one particular customer, one ISV in one specific industry, as that gets deployed, we expect that ISV's pipeline becomes available to us. The $300 million specifically, that one, we've gone through POCs, we've done some pilots, we've identified the ISV, we're working with those customers, and it's about when they wish to start deploying solutions.
We expect most of those to begin at scale in 2026.
Speaker 2
Okay, great. Thanks all for—that's all from you guys.
Speaker 3
Thank you.
Speaker 4
Thank you. Once again, to ask a question, please press *11 on your telephone. Again, that's *11 to ask a question. Our next question comes from the line of Kevin Cassidy of Rosenblatt Securities. Please go ahead, Kevin.
Speaker 0
Thank you for taking my question, and congratulations on the great results and good outlook. One question I have is your guidance for the year for 2025. It tightens it on the low end and also on the top end. I'm just wondering, on the high end, is it more of your supply chain or what brought that number down?
Speaker 3
No, it's not a supply chain issue, Kevin. It's really, we're now working very closely with those customers on what their deployment needs are. We know that they had requirements for 2025, which we are fulfilling, and we've got schedules for 2026 that we're working through.
Speaker 0
Okay, great. Your outlook for 2026 remains the same?
Speaker 3
Yes.
Speaker 0
The revenue guidance you had given before? Okay, great.
Speaker 3
Yes. A while ago, we had updated the lower end from $105 million to $130 million, but the upper end still remains the same.
Speaker 0
Okay. Can you talk about the supply chain? Is there any problems getting wafer starts or any other, I guess, what is the long pole in the tent for manufacturing your products?
Speaker 3
No, we're fortunate in the sense that our first generation of chip is at 14 nanometers. It doesn't cause us any capacity sort of challenges with the Samsung Foundry. We maintain extremely close relationships with them and our contract manufacturer. As these contracts were being discussed and we knew what the rollouts were going to be, we'd already placed orders for not just chips, but also the end, you know, the cards and so on. Dinakar, did you want to add? Okay. I'll say that, no, right, our foundry is here too, but it's also beneficial.
Speaker 0
Okay, thank you.
Speaker 4
Thank you. Our next question comes from the line of Scott Cyril of Roth Capital. Please go ahead, Scott.
Speaker 2
Hey, good afternoon. Thanks for taking my questions. Hey, Dinakar, I was hoping to dive in a little bit on the hybridization commentary. You know, there's certainly been the evolution within the data center, which was a huge upside opportunity as we look out over the next several years, but it sounds like there's a little bit of hybridization going on at the edge as well. I'm wondering a couple of things. Coexistence at the edge with GPUs as opposed to complete displacement, how do you see the evolution over the next year or so? I'm wondering, within the data center itself, it sounds like some of those early opportunities are starting to crop up and present themselves. I'm wondering about the timeline of when we start to see that materialize in a little bit more of a meaningful way.
As we look at that $725 million pipeline, how is that spread across traditional data center versus edge applications? Thanks.
Speaker 3
Sure, sure. Scott, thank you. First of all, I guess the technology part behind it, customers clearly, especially in these real-world projects, care about ROI, how much is the spend, and they have their budgets established. Within that, they have to show results and a return, whether it's a government or city or what have you. That is what is driving. If they did the entire project on GPU only, the costs would be prohibitive. At the same time, there are parts of the AI stack that Blaze is more efficient and more power efficient, more cost efficient. One of the uniquenesses of Blaze's architecture is full programmability so that it can adopt these workloads. The software toolchain to make it seamless and easy to adopt, right, so that it also helps with the customer's time to market and IT spend.
These are all coming to play, and complementing a GPU-based design, a system with a Blaze server alongside is what they're witnessing significant TCO advantage. That's what is driving the momentum. The second part, and let Harminder jump in as well.
Speaker 0
You asked about a mixture of the split of the pipeline, and actually, it is a blend. I'll give you one extreme, which is where in certain defense applications, drones, for example, it's deployment of one of our cards on the drone itself. That is the solution that's being provided. At the other end, you've got the kind of things that you've heard from the South Asia deal and the Starshine deal where we are delivering server systems powered by Blaze, and they coexist with GPU-based systems. In the middle, you've got, again, another defense type of application where you've got a mixture of the two. On the one hand, you will have sensors, perimeter security, for example. You have sensors on the perimeter, which is fusing visual spectrum, near-infrared, infrared radar, et cetera. There are certain alerts being delivered from the algorithms that are being run there and then.
They're then backed up by a command center, which has a server system. When we look at that pipeline, it's got combinations of all of those, and we just focus on where customers want to deploy the fastest, and that's how we react.
Speaker 2
Hey, if I could just quickly follow up, I think you indicated in your opening remarks about inference driving about 6x the opportunity versus traditional training. I'm just wondering if you're seeing those types of numbers in any of these early deployments, in terms of the wallet share or market opportunity for you guys, or is that the evolution that you're going to expect over the next several years? Thanks.
Speaker 3
This is actually quite clear, right? The initial phase of AI was all about AI training and creating models in the cloud. Now, AI does need to come outside the data center into real use cases. It could be smart city, it could be smart agriculture, industrial automation, all of these use cases. This is where the majority of the workload is inference. As the world starts adopting AI, increasingly inference is the main use case. We're actually finding this, even as we discuss with our customers about the hybridization strategy, pretty common to see them have more inference demands than anything else. Short answer is yes.
Speaker 2
Great, thank you.
Speaker 3
Sure.
Speaker 4
Thank you. I would now like to turn the conference back to Dinakar Munagala for closing remarks. Sir?
Speaker 3
Thank you. As we wrap up, we started the year as a young public company, and we had a pipeline, and our focus was to convert the pipeline. I want to emphasize that Blaze Holdings is now entering this next chapter of hybrid AI with focus, momentum, and the right partnerships in place. We have a clear path to execute on our commitments at scale with our Blaze AI Platform and capture the opportunities ahead with our pipeline. Our technology is proven. Our go-to-market is working. Our team is fully aligned. I'm excited about what's ahead for Blaze Holdings, and I want to thank our employees, our partners, customers, and shareholders for their continued trust and support. Thank you.
Speaker 4
This concludes today's conference call. Thank you for participating. You may now disconnect.