Thanks a lot. You can hear us? We can see you. Okay. Hi everyone, thank you so much for joining us for SoundHound AI's fireside chat at the Global TMT Conference. Hope everyone's past few days have been going well. I know the energy has been amazing here with one-on-one meetings for SoundHound AI specifically, so we're excited to be here. Keyvan, thank you for joining us. Thank you for having me. Of course. We're going to start off really broad, and before we jump into the questions, we will definitely have time at the end for audience questions. Just hold those until the end, raise your hand, and we have a team member in the back who can help you with the mic. Definitely keep those top of mind. Keyvan, let's start at the beginning. Many of the crowd members here probably have not heard the story directly from you. I'd love to hear you tell them a little bit about how you founded SoundHound AI, why you founded it, and then we can go from there. I started three companies in undergrad when I was 19, 20, and 21, and then I just decided that I wanted to be a technical founder to a high-tech company that would make a big impact in the world, and I would spend decades of my life in it. I went to Stanford to get my master's PhD and turn my thesis into a business. When I started my thesis, I asked myself, what do they have in science fiction that we don't have in real life? I decided that voice AI is something that would happen for sure in my lifetime, and I wanted to be a part of that transformation. Voice AI is talking to computers and robots, and they would talk back to you. They would have a conversation with you. You can ask them questions. You can ask them to do things. That didn't exist 20 years ago. This is before Alexa and Siri, and those companies weren't even thinking about voice AI. We started in a dorm room at Stanford to make that happen. In fact, on Tuesday of this week, we celebrated our 20th anniversary of SoundHound AI being incorporated. Amazing. I remember the first VC pitch that I said, in 20 years, we're going to talk to computers. It finally is happening. That's amazing. Incredibly inspiring story. You touched on it a little bit, but AI is top of mind for everyone, all over the headlines. It's a huge landscape. There are so many places you can fit in. You envisioned 20 years ago that voice AI would be kind of central to what we can achieve today. How do you see SoundHound AI fitting into the landscape that we have? You can talk a little bit about that. Our business foundation is based on two predictions. One is people are going to prefer voice and natural language to interact with devices. The second thing is that AI customer service will be as necessary for every business as Wi-Fi and electricity. We provide solutions for both of those. Each can be a TAM of more than $100 billion. We power an AI assistant for devices from cars to TVs, but ultimately, tens of billions of IoT devices can and should have their own AI assistant. All the product creators need is a small, inexpensive microphone and a partner like SoundHound AI. We also provide AI customer service solutions for businesses of all sizes. We are in large enterprises like banks and insurance companies and healthcare, all the way to a single location, barbershop, or a plumber, or a restaurant can have its own AI customer service. That's awesome. It's great, especially to see that depth in terms of the target market, because transition from legacy systems to Agentic AI, as you pointed out, is happening everywhere. Few companies are doing it as well as you are with SoundHound AI that was just released in May. Can you talk about what that process looked like and how that uptake has been? Agentic is a step change in the quality and the capability of AI. There was a day where every business realized they needed a website. There was a day that every business realized they needed to have a mobile app or a mobile presence. I think today is the day that businesses are realizing they need an AI agent. The reason SoundHound AI is doing well is, I would say, two things. One is two decades of technology innovation is giving us speed and quality. We are really fast to make the stuff work and go into production, while others are still running science experiments after two years. We were the first to go with GenAI in production with cars. Within months of GenAI becoming available, we were live. We did a POC within weeks, and we were live within months. With Agentic, we have also incorporated that into our platform with Amelia 7. We are already migrating 15 of our top enterprise customers to Agentic. With Agentic, ultimately, AI used to do simple things. Then it escalated more complex things to humans. Yep. Now, with Agentic, AI can do more complex things. Yeah. Fewer escalations mean lower cost for our customers, more revenue for SoundHound AI, and better user satisfaction for the callers. In fact, we ran a survey of users who use AI agents regularly, and the satisfaction rate has gone up by 2x. Oh, wow. That's a really good validation for our business. Absolutely. I think you talked about two things that sort of differentiate SoundHound AI in that answer. You talked about the way that you're able to innovate faster and the way that you're able to improve customer outcomes. You just had an amazing quarter. I'm sure many of you in the audience have seen growing 200% year over year. That's essentially unheard of. Talk about what contributed to that growth. Yeah, it was a record quarter. It was a blowout quarter, I would say. Revenue was up three times, more than three times. Every business unit delivered success and growth. We had huge growth in AI for automotive, huge growth in AI for enterprise, and huge growth in AI and automation for restaurants. We had new deals, upsells, and cross-sells across the business units. Ultimately, what the shift I'm seeing is that AI adoption is a mandate for these businesses. It used to be, maybe two years ago, an innovation budget for a POC that you would get from a brand, and they would just experiment. Now they're coming to us, and they have an IT budget, and they have to choose their AI partner. We feel very bullish about our near-term and long-term prospects. That makes total sense. I think part of that, too, is you talk a lot about your three-pillar strategy and how that encompasses essentially increasingly more complex frontiers. Can you talk about where we've been, where we're going, and where you see yourself within that framework? I should explain the three pillars. Our business is based on three pillars. In the first pillar, we power devices, as I mentioned. We power an AI assistant for cars, TVs, IoT devices, and so on. In the second pillar, we power an AI customer service for businesses, all the way from banks and insurance companies and healthcare to restaurants and retail and merchants. A lot of these interactions are transactional. That's our pillar two. In the third pillar, we connect them. We connect the users of our pillar one customers to the transaction opportunities of our pillar two customers. Imagine you're driving your car, and you're talking to your car to control the air conditioning, do navigation, maybe ask for weather, historically limited use cases, and we power those experiences. What if you could do more? What if while you're driving home, you can find a restaurant that your family might like, and you can order for pickup? We power those restaurants, too. You can order from your car instead of going to a drive-thru or going inside or using a mobile app and breaking the law. It doesn't stop there. You can book appointments for the week and maybe buy groceries. That's the third pillar of our business. It's a vision we've talked about for years. We finally unveiled the experience at CES earlier this year in January. We are working with multiple OEMs from Europe and North America. We have merchants with national and global coverage. It's going really well. Everybody's writing to be a part of it. That's amazing. Super, super exciting. Part of what contributes to, again, your success has been really you being at the forefront of technology. Polaris is central to that. Can you talk a little bit about Polaris, what it is, and what is new about that technology for the context of the group? Polaris is SoundHound AI's own multimodal, multilingual foundation model. We think it's the best of its kind. For what it does, it beats its competitors by a large margin. Including big tech, we beat in accuracy by more than 35%, in latency by more than four times, all of that running at a lower cost. We can achieve that because of 20 years of innovation and iteration and the massive amount of data that we have from real users, billions of interactions from millions of users in dozens of languages, in noisy environments. That gives us an advantage to build these models that are really good. As our customers migrate to Polaris, they see an improvement in user experience, higher automation rates, ultimately better user experience. They migrate, and we can renew them. We can upsell. As we are pitching to new customers or in our pipeline, when they compare Polaris with their legacy models, they see a clear improvement. That has improved our closing rates. That's awesome. I would say you also, we haven't talked yet about your track record of successful M&A. You have acquired three companies in the past year. How has that gone? Will you be doing more acquisitions? Talk to us about that strategy. We bought three companies in 2024, and they have gone really well. In fact, we think we have demonstrated a repeatable success formula. First, we bought companies that were really great on their own, amazing teams, great solutions, loyal customers that they spent 10, 20 years accumulating. We gave them what they needed to thrive. First, we replaced their tech dependencies with SoundHound AI models like Polaris that improved the user experience, lowered the costs, and we gave them resources to win. We did cloud migration, cost synergy exploration, revenue synergies, cross-sell, upsell, and all of that happened in 12 to 18 months. It happened really fast. We think we can do it again. We'll see. That's great. I have a fun speed round of questions. Before we do that, I'd love to hear about questions from the audience. Yeah. Where's the competitive landscape? Is it just you guys are running away with it, or are you meeting like the big tech people, you know, Alexa and these kind of guys? Where's the differentiation that will enable you to keep winning the market, whatever you do? Yeah, so in our, in pillar one, we have a limited number of players. In automotive, there's one or two. In enterprise, the TAM is so large, the field has been disrupted. The legacy players, a lot of them have been disrupted and haven't been able to reinvent themselves. I think we have a really good advantage as a company that has scale, but also has the latest and greatest technology offering. The smaller players don't have the scale, don't have the resources, don't have the experience behind them. The larger players don't have the partnership mentality that SoundHound has, where we partner with our customers and we help them innovate. That gives SoundHound a unique opportunity to really win in the space. What’s the revenue model? If you have an agreement with a car company, is it on a per unit basis? Is it just over five years for your technology? In pillar one, our revenue model is royalty-based. That has two versions. The version of our technology that runs on the edge is going to have a one-time royalty, and the version that runs in the cloud is a recurring royalty, but a fixed fee because device makers want to understand their cost and they don't want to be surprised. Even if it might be lower, if they sign up for a usage-based model, they're willing to pay a higher fee so that it's predictable. In pillar two, it's based on usage. Sometimes it's SaaS-based, sometimes it's a fixed monthly fee, sometimes it's based on interactions. When they get 10 million interactions, 100 million interactions per year, there's a fee associated with each interaction. Awesome. Other questions? OK, that's OK. We'll move on to our speed round. If new questions pop up, let us know. Keyvan, any good stories about encountering your products in real life, in the car, at a restaurant, anything like that? I'm a new product owner for Lucid Air and just bought it a few weeks ago. SoundHound AI powers the AI assistant in Lucid. I went on a drive with my five-year-old daughter. I used my voice to set the navigation and then also ask for weather, just to, and then she started talking to it. Oh my God. She had this nonstop interaction, asked for stories, asked for information about some animals, played some games with it, and started trying to make phone calls. I had to actually intervene and stop her from making phone calls. That is so funny. She made a new friend that day. Yeah, yeah, loved it. What do you consider your most exciting new product? I would say voice commerce because it's a real-life implementation of Agentic AI. People talk about Agentic AI as a vision of the future, but SoundHound AI is actually bringing that to real life. I think it's really going to transform the way businesses interact with their customers. Makes sense. We just talked about it, funny enough, just before, but what do you consider SoundHound AI's competitive edge? I would highlight two things. One is technology. Again, 20 years of innovation. Our customers test us, and then they choose us. The second thing, as I mentioned, is the partnership versus vendor mentality. Yep. Our customers need a partner. They don't want a vendor. Right. They don't want a big tech API and documentation who says, "Good luck." Mm-hmm. Right? We partner with our customers. They know that success requires a partnership. We sit with them, we listen to them, we fix their challenges, and we help them innovate. That's why they choose us. Absolutely makes sense. In one sentence, give 2020 you a piece of advice. At the risk of not sounding humble, and we are very humble as a company, and we are our own biggest critics, I would pat myself on the back. Mm-hmm. I would say we did it. We beat the odds when most others didn't survive. We thrived. The advice would be, if you believe in yourself and your vision and your capabilities, the potential is unlimited. That's awesome and well deserved. Finally, in one sentence, describe what you think SoundHound looks like in another five years. Our ambition is to be a $100 billion-plus company, and I think our ambitions are justified. Perfect. We believe it too. Any final questions from the audience? Perfect. Yeah, go for it. I think the mic's coming. Great. Thank you. You touched on the opportunity around voice commerce. Can you maybe talk about some of the building blocks that are required to kind of get users to transact? What are the gating factors today? Voice commerce is a vision we had back in 2019, and we've been talking about it publicly since 2021. We've had this vision for a long time. What we needed was scale, right? If you don't have merchants with national coverage that are willing to provide the transaction opportunities, and if you're not in millions of endpoints, then you have a chicken and egg problem, right? We finally reached that scale. Sometime last year, we were in millions of cars on the road, and we have national coverage of prominent brands that are willing to participate in this program. In every state, you can find some brands that are willing to sell you something. We power them. We power those cars, and we can bring it together. Scale was one. Implementation, we accelerated that through a small acquisition of a company called Allset that spent years building a commerce platform. They came on board and built the voice commerce for us, which has those building blocks that you mentioned. At this moment, it's just a matter of a little bit of time to bring it all together and go live in production, which we think is going to happen very soon. Great. Thank you. Looks like there's another question in the back. Hi. No, I really believe that automating as much as you can is a good step forward so that you can do more interesting things. The only thing that I'm often wondering is to what extent you can still understand the decision-making process in these AI systems in a way. I think this is probably what industry has to look and solve in the next five years if you actually want to give it even more Agentic powers. Is SoundHound AI in any way looking at mechanistic interpretability and believing this is a future? Yeah, a lot of the models that we run are made and maintained by SoundHound AI. We also integrate models, third-party models. The reason we do that is because we always want to deliver the best and the latest and greatest to our customers. We positioned ourselves, for example, that if a company like OpenAI delivers a big model change that improves things, we want that to benefit SoundHound AI, not to be a threat. When they reduce their cost, it's good for us. When their models improve, it's good for us. We don't have 100% dependency on them because a lot of models are made by us. The solutions that we go after are mostly around customer service. That is, I think, the most ripe opportunity for GenAI and Agentic AI because a lot of those requests, even though they're very complex, have a solution, right? If somebody wants to make a very complicated appointment, let's say you want to make a healthcare appointment for yourself and for your daughter on the same day, but back to back with different physicians, that's a complex request. Those used to get escalated. Now Agentic AI can do it, right? Even in voice commerce, you could say, check the weather in San Francisco, and if it's more than 70 degrees, get me a hot latte. If it's less than 70 degrees, give me a cold latte. Agentic can do that because we have a tool to get weather. We have a tool to find restaurants. We have a tool to place an order. We know your destination. We can deliver that kind of experience. We are focused on experiences that are very ripe for automation. As AI becomes more capable, it's more of a benefit than the risk. There are opportunities that it's very important to make the right decision and do the right thing. If the downside of doing the wrong thing is very expensive, SoundHound AI is focusing on more ripe opportunities. Sorry about that. Yeah, I understand that you're trying to build powerful solutions. I don't question that this is possible. Even the models that you build yourself, do you understand your own models? I mean, if you have a secretary and she's basically scheduling you in the best possible and optimal way, you know how your day will look like. You can always ask her, you know, why should we do it in this way? That will be very helpful for you. She will understand what your schedule looks like well enough, and she can be making intelligent solutions. If you replace that with an AI tool, can you also ask that question? Can you understand the decision-making process? How important is that? How much should we look at that? It's really about opening the black box of the tools that you're developing and if there's an option there. We have been at this for 20 years, and the technology that we have been working on has seen a lot of disruptions. Some of those disruptions were done by us, actually. We were the first company that showcased compound complex queries in a voice assistant. You can see our demos from 2015 that handle those. We were a pioneer of compound complex questions. There have been advancements outside of our company that we have incorporated. We have seen the evolution of this. It used to be all deterministic, right? A lot of software engineering, really good software engineering, but deterministic. When the machine learning capabilities became more proven, we introduced this yin-yang configuration where you have machine learning and you have deterministic working together. Initially, the bigger semicircle was deterministic or software engineering, and the smaller semicircle was machine learning. Over time, machine learning becomes a bigger portion. What's unique to SoundHound AI is that we have seen them both. We have made them both. We can utilize them both at the right scale. Teams and companies that are just relying on one have to wait for it to be ready. People have tried to do a machine learning approach, 100% machine learning to solve problems. They have been waiting, and it's just not good enough. You can make demos. You can go live, right? Even with OpenAI, they give these great demos, and it might be 70% right. When it's right, it's so amazing. When it's wrong, the 30%, their audience is forgiving because their audience is seeing what the future looks like. You can't be 70% right and be the AI customer service for a business, right? Our audience is not forgiving. We have to be closer to 100%. We combine the machine learning models with deterministic models that we are very good at, and we have made over 20 years to go the last mile. Over time, I expect the reliance on those will go down. In the meantime, it's giving SoundHound AI the advantage to go live faster than others because it's an asset that we have. That's how we solve the problems that I think might be on your radar. Looks like we have another question here. Thank you. Hi. Thank you. I have a few questions. The first one regarding the pillar one. For the pillar one, I think most of our customers do like the automotive, right? I'm wondering what's the runway there? For example, how many brands we can break into? For the brands they establish partnership, what's the penetration rate now and how does the runway look? By the way, for the pricing there, it's just royalty, right? Royalty by volume. It's not recurring, but will we consider the recurring pricing there? That's my first question. Thank you. Our pillar one is mostly automotive today, but it's not just automotive. We power TVs, and ultimately, we think every IoT device, you know, tens of billions of IoT devices, can benefit from their own AI assistant. In automotive, we are in over 20 brands. I would say logo-wise, our market share is more than 20%, but unit-wise, it's lower than that. That's good news because our customers are still scaling with us, right? In automotive, it takes time to win brands, takes time to go live with them, then takes time to scale with them. We haven't even reached that level of scale with the existing brands that we have. We have room to grow with our existing customers, and we have room to add new logos in automotive. In terms of royalty, the edge component of our solution is a one-time fee, and the recurring aspect of it is that they ship new devices every year. The cloud portion of our AI is a recurring fee because it's a royalty per year, royalty per three years, and those get renewed. The recurring aspect will continue. Yeah. I mean, for the brands we're entering into, we think how much of their cars can really adopt our products. You mentioned maybe 20%, or it could be much higher because going forward, everybody needs to talk in the car. Yeah, I think cars were one of the first industries we went after because you didn't need to convince car makers they needed voice. We just needed to convince them to work with us. I think the way the world is going, every car is going to be enabled. It's going to be cloud-enabled. It's going to be connected, and it's going to be enabled with AI. The way we scale with them is, one, we just get on more and more next-generation models. Usually, they have the legacy model, and then the new model, the new model grows, and then the old model shrinks. That's how we scale with them. The other way to scale with them is adding new territories. We might start in North America, then we add Europe, then we add Asia. As we add new languages for them and new territories, that's how we expand with them. Yeah. I noticed we have some cooperation with Korean brands, right? The tariff has no impact to us for the tariff. The cloud portion is something that usually the regional, the local companies pay for it. The edge portion usually is part of the manufacturing. There could be an impact ultimately, but we really haven't seen that yet. I don't want to downplay it. If anything we've seen was like people rush to buy new cars because there was the uncertainty, and we saw some increase in volume. Yeah. For pillar two, I'm wondering what's the implementation process? For example, if we want to enter into a restaurant or enterprise, I think we not only need to sell the software, right? They need to have the whole system. They need to have some hardware to support. I'm not sure how the implementation looks like. Also, we need to somehow teach or educate the customers how to use it. I'm not sure how the process is there. Yeah, that's one of the benefits of what SoundHound AI delivers. We are not just an API. We are a whole solution. In restaurants, for example, we integrate with their POS. They don't need to get new hardware. They don't need to get a new POS or a new tablet. They don't need to do data entry to our system. We read from the POS and write to the POS. They already have the POS. We do the integration with their existing system. Same with retail, same with hospitality, same with healthcare. We do integration with their infrastructure. That's something that SoundHound AI brings to the table, something we've accelerated through some acquisitions, and something that gives us customer loyalty because we are deep in their organization. Yeah, you mean we offer the product or we offer the hardware product, right? I mean, we integrate our software into the hardware part, right? Yes, we don't provide the hardware. They already have their infrastructure. We integrate with their infrastructure through API integration and others. I see. Last question, if I may. I'm curious about our profitability trajectory. In the future, what's the thought there? Thank you. I can tell you what we have guided, which is we expect to exit this year on an adjusted EBITDA positive basis. Any final questions? Awesome. Thank you guys for joining us, and thank you. Thank you so much. Thank you for our conference. Thank you.