SoundHound AI, Inc. (SOUN)

SoundHound AI is a company that builds voice-recognition and conversational AI software. It started out making music-identification technology — you hum or sing a song, and the software figures out what it is — but over time shifted into building voice assistants for cars, restaurants, and smart devices. The company makes money primarily by licensing its technology to car manufacturers, restaurant chains, and device makers. Its shares trade on NASDAQ (SOUN) and are useful to understand as a play on the shift from button-based interfaces (you push a button, menu appears) to voice-based interfaces (you talk to the car, the car talks back).

From music ID to voice AI

The company started in 2005 as SoundHound, making an app that listened to a song playing and told you what it was. For years that was the main thing anyone knew about the company — it was the humming app, competing against Shazam, which did the same thing. The humming feature was clever but ultimately a niche; Shazam dominated the space and got acquired by Apple for over a billion dollars. SoundHound, meanwhile, realized the humming app was never going to make big money, so the team pivoted.

The pivot was smart. The core skill the company had built — teaching software to understand human speech and intent — was useful for way more than identifying songs. Any place where people currently use buttons or touch screens could potentially use voice instead. If you are driving a car, you cannot look at a screen, so voice makes sense. If you are working the register at a restaurant, you cannot stop and type, so voice makes sense. If you have a home device or a speaker, voice is often the most natural way to interact.

Starting around 2014 or so, SoundHound began building conversational AI software aimed at these use cases. The company licensed its technology to car companies, restaurant chains, and consumer-device makers. This is a classic enterprise software model: the company builds the software once, and then charges customers for using it or licensing it.

The three market segments: automotive, restaurant, consumer

Automotive. SoundHound’s push into cars makes sense. Modern cars are increasingly full of software and connectivity. Instead of having to press buttons or look at a dashboard screen while driving, you can talk: “Call mom,” “Navigate to the nearest gas station,” “Set the temperature to 72.” The company has partnerships with multiple car manufacturers to provide the voice AI that powers these interactions. As cars become more connected and software-first, the addressable market for voice-AI-in-cars grows.

The catch is that the car business is notoriously difficult to break into. Car manufacturers move slowly, have long development cycles, and demand extremely high reliability from any software they embed. It takes years from the time you sign a deal with a car company to the time the software actually ships in a vehicle that reaches customers. And once a car company picks a voice-AI vendor, switching is expensive and complex, so customer lock-in is real — but it takes a very long time to lock in.

Restaurant. SoundHound has also built tools for restaurants — systems that listen to customer orders at the counter and help the staff process them faster. The idea is that the software reduces error and speeds up service. There are tens of thousands of quick-service restaurants in the United States, so the addressable market is large. But restaurants operate on thin margins and are slow to adopt new technology. A restaurant is willing to invest in voice ordering if it saves labor or increases speed, but the company has to prove the value, and adoption is gradual.

Consumer devices. The company also sells or licenses to consumer-device makers — smart speakers, smart displays, and the like. This market is growing as more devices become internet-connected, but the competition is intense. Amazon’s Alexa, Google Assistant, and Apple’s Siri dominate the space. For SoundHound to win, it has to offer something those giants do not — maybe better accuracy in specific contexts, or privacy advantages, or integration with a particular use case.

The unit economics puzzle

The big question with SoundHound is whether the unit economics make sense. Say the company licenses its software to a car manufacturer. The car company pays a one-time license fee and then a per-vehicle fee for each car shipped with the software. SoundHound’s cost to serve that customer is mostly zero — the software is built, the infrastructure is in place. So the gross margins should be very high, maybe 70% or more. And that would be fine — a high-margin software business is what investors dream about.

But the trouble is the sales cycle. Getting a car company to sign a contract and then ship the feature in production vehicles can take five, ten, or even fifteen years. During all that time, SoundHound is spending money on R&D, sales, and operations but not yet earning revenue from that customer. The company has to survive on its balance sheet or on capital raises. That is fine if the company can raise capital and if there is conviction that these long-term partnerships will eventually pay off. But if investors lose patience, or if the company runs low on cash, the story changes.

The technology challenge and the competitive landscape

Voice recognition has gotten dramatically better in recent years, especially with the rise of large language models like GPT. A large language model can understand natural language and context in ways that older voice-AI systems could not. SoundHound has incorporated these advances into its technology, but so have all the competitors.

The problem is that the biggest companies in the world — Amazon, Google, Apple, Microsoft — are all heavily invested in voice AI. They have vast resources, huge amounts of training data, and deeply embedded products. Amazon’s Alexa is in tens of millions of homes. Google Assistant is on Android phones and ChromeOS devices. Apple’s Siri is on every iPhone and iPad. These companies can afford to invest in voice AI, to absorb losses in the near term, and to wait for the space to mature. SoundHound is a much smaller company with limited resources. The only way SoundHound survives and wins in this competitive landscape is by being better in a specific niche or by partnering with strong customers who need its specific capabilities.

Revenue recognition and the licensing model

SoundHound’s business model creates a mismatch between when the company signs contracts (which can happen years before revenue is recognized) and when revenue actually appears on the financial statements. There are upfront fees for licensing, and then recurring fees as customers use the service. Some contracts include milestone payments (triggered when a customer reaches certain usage levels or ships a certain number of devices).

This means you have to read the 10-K carefully to understand what commitments the company has in place. Just because SoundHound has a deal with a major car company does not mean revenue is imminent; it might mean revenue starts flowing several years from now. Similarly, the company’s quarterly revenue might look flat for a while, then spike up when a major customer launch happens, then grow steadily as usage scales.

Cash burn and the capital raise treadmill

Like many software companies that have not yet reached profitability, SoundHound burns cash. The company raises capital, spends it on R&D, sales, and operations, and then when the cash gets low, the company raises more capital. The stock issued in each raise dilutes existing shareholders. If the company takes many years to reach profitability, there can be a lot of dilution. Conversely, if the company reaches profitability soon and starts generating positive cash flow, the dilution stops and the balance sheet becomes a source of financial strength.

SoundHound has raised multiple rounds of capital, and the terms have not always been favorable. In 2023, the company raised capital at prices that were below where the stock had traded in previous years, indicating that investor sentiment had turned negative. That happens to unprofitable tech companies when investors lose conviction or when the broader market environment turns against risk assets.

Boom and bust implications

In a strong tech environment when investors are optimistic about AI and software, SoundHound can raise capital at good valuations and the stock trades at a premium. Customers (car companies, restaurants, device makers) are also more willing to invest in new technology. The company’s near-term prospects look good, and the stock can run higher.

In a weak tech environment — recession, rising interest rates, skepticism about AI, or an earnings miss from a large tech company — investors flee risky, unproven tech companies. SoundHound becomes hard to fund. The stock trades at a deep discount, and the company might struggle to raise capital at favorable terms.

This means SoundHound’s stock is a leveraged play on AI sentiment and on venture-capital availability. In a boom, it can be a great performer. In a bust, it can be a terrible performer. An investor holding SOUN has to believe in the long-term vision (voice AI is the future, car companies will adopt it) and be able to survive the near-term volatility, including the possibility that the company needs to raise capital at a low price and heavily dilutes shareholders.

How to research SoundHound

Start with the latest 10-K and 10-Q filings (SEC CIK 0001840856). Pay close attention to the revenue section — it will break down revenue by customer type (automotive, restaurant, consumer) and will describe major customer relationships and the timing of expected revenue from those customers. Look at the balance sheet to understand how much cash the company has and how long that cash will last at the current burn rate.

Read the risk factors section; it will list the things management thinks could go wrong — and management’s risk factors are usually accurate and honest.

Track the quarterly earnings reports and look at the trajectory of gross margin and operating costs. Is the company moving toward profitability or away from it? Is gross margin improving (suggesting that the licensing model is working) or declining (suggesting pricing pressure or mix shift toward lower-margin business)?

Finally, understand that this is a speculative investment. The company has not yet proven that its voice-AI technology can win meaningful market share in its target markets. The long sales cycles mean it could be several years before the revenue story becomes clear. If you cannot afford to lose your entire investment or if you need the money in the next few years, this stock is not appropriate for your portfolio.