SoundHound AI, Inc. Q3 FY2024 earnings call

• Keyvan Mohajer highlighted the acceleration of the mission to voice enable the world with conversational intelligence, with revenue up 89% YOY to over $25M, run rate of over 6 billion queries annualized. - Extended AI engine customer service offering to hundreds of brands, with AI agent customer service now over half the business and growing. - Automotive portfolio growing, with wins in EV space and continued traction with Stellantis. - Restaurant segment expanding with new deals like Panda Express, Church's Texas Chicken, and White Castle, and surpassing 100 million AI-handled interactions. - Polaris foundation model powering a third of AI interactions for restaurant customers and expected to power all customer experiences. - Will have a prominent presence at CES 2025 showcasing automotive and AI solutions.

Q3 revenue was $25.1 million, up 89% year-over-year. Automotive saw double-digit unit growth and double-digit unit price expansion driven by Generative AI solutions. Restaurants continued to scale with new logos and diversified product offerings, counting seven of the top 20 quick service restaurants as customers. The automotive segment had wins like a large deal with an EV OEM in the Middle East and continued traction with Stellantis. The restaurant segment expanded with new deals and saw strong demand for phone ordering and drive-thru solutions, having surpassed handling 100 million inbound customer calls exclusively with AI.

• 2024 revenue expected to be in the range of $82 million to $85 million. - 2025 revenue expected to be in the range of $155 million to $175 million. - Aim to achieve adjusted EBITDA profitability by the end of 2025.

• Past risk of high customer concentration, now diversified with top five customers comprising less than a third of business. - Integration challenges with acquisitions, including determining which parts of acquired businesses to retain and discontinue. - Seasonality in business and need for careful integration and calibration of acquisitions for growth and profitability.

Q: Three months ago, when you announced the Amelia acquisition, you talked about looking at various parts of the business to decide which ones to incorporate. Can you give sense for how much of the Amelia business you're going to be retaining, etc.?

A: Nitesh Sharan said they're still in the process, but they're excited about accelerating product cross sell, upsell, and technology stack integration. They will keep high margin, deep integration parts, double down on voice enablement with proprietary tech, and accelerate investment in AIOps. Professional services and escalation support are still under contemplation.

Q: Could you help us understand how you benchmark the Polaris foundation model?

A: Keyvan Mohajer said they use standardized tests and real-life interaction tests. Speech recognition benchmark is word error rate, latency, and speed; they beat big tech in accuracy and speed. Natural language understanding now measures hallucination, and they focus on reducing it. They've been building models for decades and spend on GPUs to iterate.

Q: What's your M&A appetite and comment on energy and retail sectors?

A: Nitesh Sharan said programmatic M&A is imperative, they've had successful acquisitions like Amelia, and they'll be judicious in future M&A. They see opportunities in energy and retail sectors as they have horizontal solutions and large TAMs there.

Q: What's your plan collaborating with large AI LLM companies?

A: Keyvan Mohajer said they use OpenAI as a vendor, arbitrate across multiple LLMs, and are building their own Polaris foundation model, which is LLM agnostic.