SoundHound Stock: An AI Voice Assistant for Everything

Let’s talk about AI voice assistants. So far, Siri, Alexa, and company represent decades of development of conversational intelligence technologies, particularly in what’s called natural language processing (NLP) and natural language understanding (NLU). Yet it’s all still extremely crude, and the algorithms remain a poor substitute for human grey matter. In fact, giving them the name of “assistant” is probably a bit generous. The reality is that these interactive voice systems still struggle with simple queries – and also tell pretty awful jokes. When they do understand a command or question, smart voice assistants can do things more quickly than typing, calling, or other forms of communication. It’s technology for the lazy, so undoubtedly it will become ubiquitous, as everything including the kitchen sink becomes voice interactive. 

That’s what Silicon Valley-based SoundHound Inc is trying to sell to investors, with a proposed merger with a special purpose acquisition company (SPAC).

About SoundHound Stock

We first covered SoundHound more than three years ago as one of the most well-funded smart voice assistant private companies at that time. The 16-year-old Santa Clara, California company says it has raised about $280 million from a pretty impressive list of investors that includes major automakers (Hyundai, Daimler, and AG/Mercedes-Benz), electronics giant Samsung; Chinese tech behemoth Tencent; and AI chip maker NVIDIA, along with some sizable telecommunication companies (Orange and Korea Telecom), venture capital firms like Kleiner Perkins, and more. 

Despite its pedigree, the company has opted to take the back door to the public markets by combining with a SPAC called Archimedes Tech SPAC Partners Co (ATSPT). Additional private investors joining the party include Oracle, Qatar First Bank, and Koch Industries, sweetening the deal with another $111 million and netting SoundHound a total of $244 million from the reverse merger, with a valuation of about $2.1 billion. That’s assuming that no one decides to cash out, which seems to be becoming more common with SPACs, including another one we just covered called REE Automotive that saw its net proceeds drop by more than $200 million after some investors got cold feet.

In addition, retail investors seem to be finally paying heed, as the current SPAC stock price barely budged when news broke of the impending merger last month, still trading just below the $10 default share price assigned to all of these blank check companies. Is SoundHound stock yet another SPAC to avoid or simply flying under the radar because it has nothing to do with electric vehicles?

The Technology

Investing in natural language processing certainly isn’t the most sexy side of AI technologies, but it’s definitely one of the most important ones if we want machines to understand and respond to us in useful and meaningful ways. SoundHound contends that it has developed a conversational AI assistant capable of complex questions. Founded in 2015 by a group of Stanford graduates, SoundHound originally created a Shazam-like app that recognized a song from a snippet of music but was not bought by Microsoft for $400 million like the actual Shazam app. A decade later, the company unveiled an AI-powered interactive voice platform that consists of two core technologies – Speech-to-Meaning and Deep Meaning Understanding.

The former is the flagship product and refers to the platform’s ability to convert speech to meaning simultaneously and in real time. Most traditional approaches first convert speech to text, then convert text to meaning. SoundHound contends its technology results in faster, more accurate responses because it performs both speech recognition and language understanding. The latter enables the voice assistant to address multiple, complex questions and filter results simultaneously. 

For example, the AI voice assistant can actually handle a question like: “Show me all restaurants within half a mile of the Space Needle that are open past 9 pm on Wednesdays and have outdoor seating,” and follow-on qualifications such as, “Okay, don’t show me anything with less than three stars or fast food.” Try that with Siri. A free Hound voice search and assistant is available on both Android and Apple devices, if you want to give it a spin yourself. A cursory test run found it was indeed fast but also stumbled and struggled with queries that also stumped Siri and her brethren.

On the enterprise client side, there’s also a developer platform called Houndify that gives the company’s clients access to dashboards and other tools to analyze customer usage and behavior in order to optimize the voice experience of their products. The open-access platform also provides access to a library of more than 100 content domains, such as weather, flight status, sports, and more. The company’s Collective AI architecture connects these different knowledge domains, ensuring the platform grows smarter over time. The technology can also work without a cloud (i.e., internet) connection, as well as speak more than 20 languages, so the platform isn’t just U.S.-centric.

The Market

SoundHound claims it is addressing a market that it expects to hit $160 billion over the next five years. Solutions touch a number of industries, as illustrated below:

Some of the companies that use SoundHound’s interactive voice technology include Hyundai, Mercedes-Benz, Pandora, Mastercard, Deutsche Telekom, and Snap, among others. For example, the deal with Mastercard actually involves developing voice-enabled drive-through services at fast-food restaurants like White Castle (yes, those soggy little hamburgers with the one sad little pickle are still around). Given the labor market shakeup of the last couple of years, fast-food companies are scrambling to automate through AI, including a little outfit called McDonald’s. In fact, the fast-food chain is working with Denver-based Valyant AI, a startup that has created an interactive AI voice system to take customer orders, which are then merged into point-of-sales systems. So there’s definitely interest and competition in the retail sector.

The Internet of Things (IoT) is another target market, but certainly the competition is much tougher. Amazon’s Alexa already owns an estimated 70% share of smart home devices, followed by Google at about 25%. SoundHound potentially has a stronger foothold in automotive thanks to its existing partnerships and a rapidly growing customer base. Last year, total in-car voice assistant users in the United States hit nearly 130 million, with 83.8 million active monthly users, according to Voicebot.ai, a market research firm. That’s almost double the number of people who own smart speaker devices.

SoundHound claims it processes more than 100 million queries per month across the company’s ecosystem of product partnerships – twice the number compared to 2020 – and expects to exceed one billion by the end of the year. 

How Does SoundHound Make Money?

So how does all that convert to actual revenues? SoundHound makes money in three ways:

• Royalties: SoundHound collects royalties based on volume, usage, or duration when Houndify is placed in a car, smart speaker, appliance, or some other device.
• Subscriptions: The company charges a subscription when it enables an ongoing service, such as food ordering in the case of those nasty-tasting White Castle burgers.
• Monetization: This is what the company appears to be betting on for its long-term success. It basically combines the first two streams of revenue, so that when the users of a voice-enabled product access voice-enabled services, SoundHound gets a cut of the revenue from the latter for generating leads and transactions and then shares that with the manufacturer of the voice-enabled product.

An example might be helpful here. So, imagine that a driver of a voice-enabled car places an order at a White Castle because he’s drunk and it’s the only place open at 2 am. The restaurant pays SoundHound for enabling the seamless transaction between its business and the vehicle, which SoundHound shares with the car manufacturer. In theory, everyone is happy: The restaurant makes an easy sale, the car manufacturer gets a little something, something (an enticement to install SoundHound in its products), and the customer isn’t quite as hungover in the morning. 

Should You Buy SoundHound Stock?

It all makes for a good story, but we don’t invest in fairy tales but facts. After more than 15 years in business, SoundHound expects to clear about $20 million in revenue this year, so the company is starting to finally generate significant revenues. Still, based on our simple valuation ratio (market cap/annualized revenues), the stock is way overvalued at 100 versus our upper limit of 40.  A share price of about $4 would be at the high end based on the current numbers.

While SoundHound is trying to diversify its revenue streams, nearly 90% currently comes from royalties, with the other types of revenue growing over time:

There are a few problems with this picture. For instance, why does royalty revenue decline as a total share of revenue? Is it because royalty revenue in real dollars declines or because the other two are projected to generate so much more money? While the company has managed to eke out a bit of monetization revenue, the long-term numbers assume that SoundHound will become the default voice assistant in millions of connected devices. In the real world, David rarely beats Google Goliath. SoundHound has established a few key partnerships, but will need to expand its network exponentially to challenge the big tech players.

There’s absolutely no insight into these numbers that should give us confidence that SoundHound can realize annual revenue of $1.2 billion in just five years, as it claims on another shiny investor deck slide. Another black box: Existing customers will account for an estimated 96% of revenues next year, but only about a quarter by 2026. How does the company plan to make that happen?

We also don’t know the cost of these revenues. SoundHound can crawl the web for free to access basic information – like what’s the population of China? – but relies on content partners to access their data and APIs to fulfill user requests such as weather and sports scores. What does that cost? 

Some pretty big companies have put their faith and money on SoundHound. We like the concept and international diversification, but until we can better understand the company’s financials and somewhat convoluted revenue model, we’ll sit this one out.  

Conclusion

A few years ago, investors didn’t have many options when it came to AI stocks. That’s no longer the case, though there are few options for those who are specifically bullish on NLP technology. Microsoft is acquiring the biggest public company in the AI voice market with Nuance, which largely pivoted toward healthcare solutions in recent years. Then there’s legaltech stock Docusign that uses NLP and machine learning to help find, filter, and analyze contract agreements. Soon, NLP will just become another tool in every company’s toolbox.

If the SPAC merger goes according to plan – which is becoming less of a sure thing over time with SPACs – SoundHound will trade under the ticker symbol SOUN on the Nasdaq.