SoundHound AI has launched OASYS Edge, an embedded architecture that brings large language model-powered voice AI agents directly to edge hardware. The technology is set for commercial deployment in late 2026, targeting vehicles and smart devices.
TL;DR
- SoundHound AI introduces OASYS Edge, enabling LLM-powered voice AI agents to run directly on edge hardware.
- The technology supports multi-step requests and multi-system orchestration entirely on the device.
- OASYS Edge reduces dependency on cloud connectivity, lowering costs and enhancing data privacy.
What happened
SoundHound AI, Inc. (NASDAQ: SOUN) has unveiled OASYS Edge, a new embedded architecture designed to bring large language model-powered voice AI agents directly to edge hardware. This technology is primarily aimed at vehicles and smart devices, with commercial deployment scheduled for late 2026. Unlike traditional voice systems that handle only individual commands, OASYS Edge enables multi-step requests and multi-system orchestration processed entirely on the device itself.
AI agents built on the platform can understand vague or complex user requests, resolve intent, and coordinate multiple functions using optimized on-device language models. Developers can build an AI agent once on SoundHound's OASYS platform and deploy it across cloud, edge, or hybrid environments without rebuilding for each configuration. This deployment flexibility is expected to make the technology attractive to manufacturers with varying hardware capabilities, connectivity constraints, and cost structures.
Why it matters
Local processing reduces the need for constant cloud connectivity, lowering recurring API costs and bandwidth requirements that manufacturers would otherwise pass on to consumers. OASYS Edge can also support voice requests in connectivity dead zones, including tunnels, underground garages, and remote locations where cloud-dependent systems would otherwise fail.
User interactions, voice processing, and telemetry can optionally remain stored on the local device, addressing growing manufacturer and consumer concerns around data privacy. Customizable model sizes and configurations allow the technology to run on lower-cost chipsets, broadening the range of devices and price points that manufacturers can target.
Key facts
- OASYS Edge is designed for vehicles and smart devices, with commercial deployment scheduled for late 2026.
- The technology enables multi-step requests and multi-system orchestration entirely on the device.
- Developers can build an AI agent once and deploy it across cloud, edge, or hybrid environments.
- Local processing reduces dependency on cloud connectivity, lowering costs and enhancing data privacy.
- OASYS Edge can support voice requests in connectivity dead zones.
- User interactions and telemetry can optionally remain stored on the local device.
- Customizable model sizes allow the technology to run on lower-cost chipsets.
- SoundHound AI competes with Five9, Inc. and NICE Ltd. in the enterprise conversational AI market.
Context
SoundHound differentiates OASYS through self-building and self-optimizing agents, faster movement from pilot programs to production-scale implementations, and omnichannel orchestration capabilities. Despite the product momentum, SOUN shares have lost 38.8% year to date, underperforming the broader industry by a significant margin through September 2026.
From a valuation standpoint, SOUN trades at a forward price-to-sales multiple of 10.28, sitting slightly below the industry average on that same basis. The Zacks Consensus Estimate for SoundHound's 2026 and 2027 loss per share has narrowed over the past 60 days to 19 cents and 14 cents, respectively. The projected 2026 loss nonetheless remains wider than the prior year's loss of 13 cents per share, reflecting ongoing investment in platform expansion and go-to-market efforts.