SoundHound AI, Inc. (SOUN) Earnings Call Transcript & Summary

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 specifically. So we're excited to be here. Keyvan,, thank you for joining us.

Thank you for having us.

Of course. So 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 a mic. So definitely keep those top of mind. But Keyvan, let's start at the beginning. Many of the crowd members here probably have not heard the story directly from you. So 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.

So I started 3 companies in undergrad when it was '19, '20 and '21. And then I 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. So 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? And I decided that voice AI is something that would happen for sure. And in my lifetime, and I wanted to be a part of that transformation. So 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. And that didn't exist 20 years ago. This is before Alexa and Siri and those companies weren't even thinking about voice AI the way we were thinking about it. So we started in a dorm room at Stanford to make that happen. And in fact, on Tuesday of this week, we celebrated our 20th anniversary of SoundHound being incorporated. So I remember the first VC pitch that I said in 20 years, we're going to talk to computers. So finally it's happening.

That's amazing, incredibly inspiring story. And 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 fitting into the landscape that we have? Can you talk a little bit about that?

Our business foundation is based on 2 predictions. One is people are going to prefer voice and natural language to interact with devices. And the second thing is that AI customer service will be as necessary for every business as WiFi and electricity. And we provide solutions for both of those. Each can be a TAM of more than $100 billion. So 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. And all the product creators need is a small inexpensive microphone and a partner like SoundHound. And then we also provide AI customer service solutions for businesses of all sizes. We are in large enterprises like banks and insurance companies and health care 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. And few companies are doing it as well as you are with Amelia 7 that was just released in May. Can you talk about what that process looks like and how that uptake has been?

Yes, agentic is a step change in the quality and the capability of AI. And there was a day where every business realized they needed a website. And then there was a day that every business realized they need to have a mobile app or a mobile presence. I think today is the day that businesses are realizing they need an AI agent. And the reason SoundHound is doing well is, I would say, 2 things. One is 2 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 2 years. So we were the first to go with GenAI production with cars within months of GenAI becoming available, we were live. We did a POC within weeks, and we're live within months. And with agentic, we have also incorporated that into our platform with MDS7, we are already migrating 15 of our top enterprise customers to agentic. And with agentic, ultimately -- AI used to do simple things, then it escalated more complex things to humans. But now with agentic, AI can do more complex things. So it's fewer escalation means lower cost for our customers, more revenue for SoundHound and better user satisfaction for the colors. In fact, we ran a survey of users who use AI agents regularly and the satisfaction rate has gone up by 2x. It's a really good validation for our business.

Absolutely. And I think you talked about 2 things that sort of differentiate SoundHound 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.

Yes, it was a record quarter. It was a blowout quarter, I would say. Revenue was up 3x -- more than 3x and 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. And then we had new deals, we had upsells and we had cross-sells across the business units. And ultimately, what I -- the shift I'm seeing is that AI adoption is a mandate for these businesses. It used to be maybe 2 years ago, an innovation budget for a POC that you would get from a brand and they would just experiment. But now they're coming to us and they have an IT budget, and they have to choose their AI partner. So we feel very bullish about our near-term and long-term prospects.

That makes total sense. And I think part of that, too, is you talked a lot about your 3-pillar strategy and kind of 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 kind of within that framework?

So I should explain the 3 pillars. So our business is based on 3 pillars. In the first pillar, we power devices, as I mentioned. So we power an AI assistant for cars, TVs, IoT devices and so on. In the second pillar, we power AI customer service for businesses all the way from banks and insurance companies and health care to restaurants and retail and merchants. And a lot of these interactions are transactional. And that's our Pillar 2. So the third pillar, we connect them. We connect the users of our Pillar 1 customers to the transaction opportunities of our Pillar 2 customers. So 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. Now 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. Because we power those restaurants, too so 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. And it doesn't stop there. You can book appointments for the week and maybe buy groceries. And that's the third pillar of our business. It's the vision we've talked about for years. And finally, we unveiled that -- unveiled the experience at CES earlier this year in January, and we are working with multiple OEMs from Europe and North America. We have merchants in -- with national and global coverage. That's going really well. Everybody is running to be a part of it.

It'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?

So Polaris is SoundHound's own multimodal, multilingual foundation model. And we think it's the best of its kind. For what it does, it beats its competitors by a large margin. So including a big tech, we beat in accuracy by more than 35%, in latency by more than 4x, all of that running at a lower cost. And 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, and that gives us an advantage to build these models that are really good. And as our customers migrate to Polaris, they see an improvement in user experience, higher automation rates, ultimately, better user experience. So they migrate and we can renew them, we can upsell. And as we are pitching to new customers or in our pipeline, when they compare Polaris with their legacy models, they see a clear improvement, and 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 3 companies in the past year. How has that gone? Will you be doing more acquisitions? Talk to us about that strategy.

We bought 3 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. And then we gave them what they needed to thrive. First, we replaced their tech dependencies with SoundHound models like Polaris that improved the user experience, lowered the costs. And we give them resources to win. We did cloud migration. We did cost synergy exploration, revenue synergies, cross-sell, upsell. And all of that happened in 12 to 18 months. So it happened really fast. And we -- again, we think we can do it again. So we'll see.

That's great. So I have a fun speed round of questions. But before we do that, I'd love to hear about questions from the audience.

Where is the competitive landscape? Is it just you guys are running away with it? Or are you meeting like the big tech people or Alexa and these kind of guys? And where is the differentiation that will enable us to keep winning the market, whatever you do?

Yes. So in our -- in Pillar 1, we have a limited number of players in automotive, there's 1 or 2. But in enterprise, the TAM is so large. The field has been disrupted. The legacy players, a lot of them haven't -- have been disrupted and haven't been able to reinvent themselves. So 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. So that gives SoundHound a unique opportunity to really win in this space.

And what's the revenue model? Like if you have an agreement with a car company, is it on a per unit basis? Is it just over 5 years for your technology?

So in Pillar 1, our revenue model is royalty-based, and that has 2 versions. The version of our technology that runs on the edge is kind of a onetime 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. So even if it might be lower, if they sign up for a usage-based model, they are willing to pay a higher fee so that it's predictable. In Pillar 2, it's based on usage. Sometimes it's a SaaS-based, sometimes it's a fixed monthly fee, sometimes it's based on interactions. So when they get 10 million interactions, 100 million interactions per year, there's a fee associated with each interaction.

Awesome. Other questions? Okay. That's okay. We'll move on to our speed round. But if new questions pop up, let us know. So Keyvan, any good stories about encountering your products in real life in the car, at a restaurant, anything like that?

So I'm a new proud owner of Lucid Air and just bought it a few weeks ago, and we -- SoundHound AI powers the AI system in Lucid. So I went on a drive with my 5-year-old daughter. So I use my voice to set the navigation and then also ask for weather just to -- and then she started talking to it. And she had this nonstop interaction, asked for stories, asked for -- asked information about some animals, played some games with it, and she started trying to make phone calls. And I had to actually intervene and stop her from making phone calls.

That is so funny. She made a new friend that day. 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 is actually bringing that to real life. And I think it's really going to transform the way businesses interact with their customers.

We just talked about it funny enough just before, but what do you consider SoundHound's competitive edge?

I would highlight 2 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. Our customers need a partner. They don't want a vendor, right? So they don't want a big tech API and documentation who says good luck, right? We partner with our customers. They know that success requires a partnership. We sit with them. We listen to them. We pick their challenges. We help them innovate, and that's why they choose us.

Absolutely makes sense. In one sentence, give 2020 you a piece of advice.

So 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 and I would say we did it. We beat the odds when most others didn't survive, we thrived. So 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. And then finally, in one sentence, describe what you think SoundHound looks like in another 5 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. Yes, go for it. I think the mic is coming.

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? And what are the gating factors today?

So voice commerce is a vision we had back in 2019, and we've been talking about it publicly since 2021. So we've had this vision for a long time. But what we needed was scale, right? So if you're not -- 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 there's a bit of a chicken and egg problem, right? So we finally reached that scale. And sometime last year, we are in millions of cars on the road, and we have national coverage of prominent brands that are willing to participate in this program. So in every state, you can find some brands that are willing to sell you something. So we power them, we power those cars, and we can bring it together. Scale was one. Implementation, we accelerated that through -- we had a small acquisition of a company called Allset that spent years building a commerce platform. So they came on board and built the voice commerce for us, which has those building blocks that you mentioned. So 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.

There's another question in the back.

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. And I think this is probably what industry has to look and solve in the next 5 years if you actually want to give it even more agentic powers. Is SoundHound in any way looking at mechanistic interpretability and believing this is a future?

Yes, a lot of the models that we run are made and maintained by SoundHound, but we also integrate models, third-party models. And 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, not to be a threat. So when they reduce their cost, it's good for us. When their models improve, it's good for us. But a lot of -- but we don't have 100% dependency on them because a lot of models are made by us. And the solutions that we go after are mostly around customer service. So 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 health care appointment for yourself and for your daughter on the same day, but back to back with different physicians. That's a complex request. And those used to get escalated, but now agentic AI can do it, right? So -- or even in voice commerce, you could say, check the weather in San Francisco and if it's more than 70 degrees, give me a hot latte, and if it's -- or cold latte and if it's less than 70 degrees, give me a cold latte. And 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. So we can deliver that kind of experience. So we are focused on experiences that are very right for automation. And as AI becomes more capable, it's more of a benefit than the risk. And there are opportunities that it's very important to make the right decision and do the right thing. And if -- the downside of doing the wrong thing is very expensive, but SoundHound is focusing on the more right opportunities.

Sorry about that. Yes, I understand that you're trying to build powerful solutions, and I don't question that this is possible. But even the models that you build yourself, do you understand your own models? I mean if you have a secretary and she is basically scheduling you in the best possible and optimal way, how your day will look like. You can always ask her, why should we do it in this way? And that will be very helpful for you. She will understand what your schedule looks like well enough and she can be making intelligent solutions. Now 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? So it's really about opening the black box of the tools that you're developing and if there is an option there.

So we've been at this for 20 years, and the technology has -- that we have been working on has seen a lot of disruptions. Some of those disruptions were done by us actually. Like we were the first company that showcased compound complex queries in a voice assistant. You can see our demos from 2015 that handled those. So we were a pioneer of compound complex questions. But also there have been advancements outside of our companies that we've incorporated. So we've seen the evolution of this. It used to be all deterministic, right? So a lot of software engineering, really good software engineering, but deterministic. And then when the machine learning capabilities became more proven, we introduced this [indiscernible] configuration where you have machine learning and you have deterministic working together. Initially, the bigger semi-circle was deterministic or software engineering and the smaller semi-circle was machine learning. But we are seeing that over time, machine learning becoming a bigger portion. What's unique to SoundHound is that we've seen them both. We have made them both, and we can utilize them both at the right scale, whereas teams and companies that are just relying on one, they have to wait for it to be ready. So people have tried to do machine learning approach, 100% machine learning to solve problems. And they've been waiting, and it's just not good enough. You can make demos, you can go live, right? So even with OpenAI, they give these great demos. And it might be 70% right and when it's right, it's so amazing. And then when it's wrong, the 30%, their audience is forgiving because their audience is seeing what the future looks like. But you can't be 70% right and be the AI customer service for a business, right? Our audience is not forgiving. So we have to be closer to 100%. So we combine the machine learning models with deterministic models that we are very good at, and we've made over 20 years to go the last mile. And over time, I expect the reliance on those will go down. In the meantime, it's giving SoundHound the advantage to go live faster than others because it's an asset that we have. And that's how we solve those -- the problems that I think might be on your radar.

Looks like we have another question here.

I have a few questions. The first one regarding the Pillar 1. I mean, like for the Pillar 1, I think like most of our customers still like the automotive, right? I'm wondering like what's the runway there? For example, like -- I mean, like how many brands like we can break into? And for the brands like we established partnership, I mean, like what's the penetration rate now? And how is like the runway looks like? And by the way, like for the pricing there, like it's just like royalty, right, royalty by volume is not like recurring, but like will we consider the recurring pricing there? So that's my first question.

So our Pillar 1 is mostly automotive today, but it's not just automotive, the power TVs. And ultimately, we think every IoT device, 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, and that's good news because our customers are still scaling with us, right? So in automotive, it takes time to win brands, takes time to go live with them, then takes time to scale with them, and we haven't even reached that level of scale with the existing brands that we have. So we have room to grow with our existing customers, and we have room to add new logos in automotive. In terms of royalty, there's -- the edge component of our solution is a onetime fee and then 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 for 3 years and those get renewed. So the recurring aspect will continue.

Yes. I mean for the brands we enter into, I mean, like we think like how much of their cars like can really like adopt our like products? I mean like you mentioned like maybe 20% or like it could be much higher because like going forward, like everybody needs to like talk in the car and like.

Yes. I think cars were one of the first industries we went after because you didn't need to convince carmakers they need voice. We just need to convince them to work with us. And usually -- and I think that 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. So the way we scale with them is, one, we just get on more and more next-generation models. So usually, they have the legacy model and then the new model, the new model grows and then the old model shrinks, and that's how we scale with them. The other way to scale with them is adding new territories. So we might start in North America, then we add Europe, then we add Asia. So as we add new languages for them and new territories, that's how we expand with them.

Yes. And also, I noticed like we have some cooperation with like Korean brands, right? But like 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. So there could be an impact ultimately, but we really haven't seen that yet. So I don't want to downplay it. But if anything, we've seen it was like people rush to buy new cars because there was the uncertainty, and so we saw some increase in volume.

Yes. And sorry, like for the Pillar 2, I'm wondering like what's the implementation process? Because like, for example, if we want to enter into a restaurant or enterprise, I think like we not only need to sell the software, right? They need to have the whole system. They need to like have some hardware to support. So I'm not sure like how like implementation looks like. And I mean, like -- and also, we need to somehow teach the -- to educate like the customers how to use, and I'm not sure like how is the process there.

Yes. So that's one of the benefits of that SoundHound delivers. So we are not just an API, we are a whole solution. So in restaurants, for example, we integrate with their POS. And they don't need to get a new hardware. They don't need to get a new POS or the new tablet. We integrate -- they don't need to -- they do data entry to our system. We read from the POS, we write to the POS. They already have the POS. So we do the integration with their existing system, same with retail, same with hospitality, same with health care. We do integration with their infrastructure. And that's something that SoundHound brings to the table, something we've accelerated through some acquisitions and something that gives us customer loyalty because we are deep in the organization.

Yes. So you mean like we offer the product. We offer the hardware product, right? I mean like we integrate our software into the hardware product, right?

Yes. But we don't provide the hardware. They already have their infrastructure. We provide the -- we integrated with their infrastructure through API integration and other.

I see. Sorry, last question, if I may. I'm curious about like our profitability trajectory, I mean, like in the future, what are the thoughts there?

So 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. Well, then thank you guys for joining us, and thank you.

Thank you so much.

Thank you [indiscernible] conference.

Thank you.