Ambiq Introduces Two Novel Edge AI Execution Environmental Offerings
Ambiq Micro, Inc. has announced the launch of HeliosRT and HeliosAOT, two groundbreaking edge AI runtime solutions tailored for the Ambiq Apollo Systems-on-Chips (SoCs) family. These solutions are designed to optimize AI performance and energy efficiency in low-power edge computing environments.
Introducing HeliosRT: A Performance-Enhanced LiteRT
HeliosRT is a power-optimized version of LiteRT (TensorFlow Lite for Microcontrollers) that delivers up to 3x faster inference speeds and significantly improved power efficiency compared to standard LiteRT implementations. Its custom AI kernels are optimized for the Apollo510 vector acceleration hardware, providing better support for audio and speech models. This enables more efficient AI processing on ultra-low-power devices such as wearables, IoT sensors, and industrial monitors.
Streamlining Deployment with HeliosAOT
HeliosAOT, on the other hand, is an ahead-of-time (AOT) compiler that transforms TensorFlow Lite models into embedded C code. This reduction in memory footprint allows for more granular memory control and per-layer weight distribution across the Apollo SoC’s memory hierarchy. By generating direct C code, HeliosAOT allows for greater flexibility and tighter integration into embedded applications, which is critical for resource-constrained edge systems.
Leveraging Ambiq's Sub-threshold Power Optimized Technology (SPOT)
Both solutions leverage Ambiq’s patented Sub-threshold Power Optimized Technology (SPOT) platform, which enables extremely low power consumption without sacrificing computational performance. This technology combined with Apollo SoCs’ hardware acceleration features (such as vector acceleration and efficient numerical support) enables these AI runtimes to deliver a superior balance of high AI inference performance and ultra-low energy usage targeted at edge applications like wearables, smart homes, digital health devices, and IoT sensors.
Seamless Integration with Existing AI Development Pipelines
Both HeliosRT and HeliosAOT are designed for seamless integration with existing AI development pipelines, supporting TensorFlow workflows and providing documentation, examples, and engineering support to ease developer adoption. HeliosRT entered beta release and is expected to be generally available in Q3 2025, while HeliosAOT is currently in technical preview with wider availability planned for Q4 2025.
Key Features and Benefits
| Feature | HeliosRT | HeliosAOT | |----------------------------|------------------------------------------|-------------------------------------------| | Core technology | Optimized LiteRT for Apollo SoCs | Ahead-of-time compiler, generates C code | | Performance improvement | Up to 3x faster inference | N/A (focus on memory and deployment flexibility) | | Power efficiency | Up to 3x better power efficiency | Lowers memory footprint 15–50%, enabling efficient memory use | | Memory usage | Standard for LiteRT with Apollo optimizations | Reduced by 15–50% through direct C code deployment | | Target devices | Ultra-low power edge devices like wearables, IoT sensors, industrial monitors | Same; optimized for memory-constrained embedded systems | | Developer integration | Fully compatible with TensorFlow workflows, part of neuralSPOT SDK | Integrates directly with embedded applications via C code | | Availability | Beta (Q3 2025 general release planned) | Technical preview (Q4 2025 wider availability planned) | | Underlying tech platform | Ambiq SPOT technology for subthreshold power optimization | Same SPOT tech enabling ultra-low power operation |
In essence, these solutions optimize edge AI by maximizing performance per watt, reducing inference latency, minimizing memory usage, and providing flexible deployment mechanisms, all crucial for extending battery life and enabling always-on AI on constrained edge devices.
With HeliosRT and HeliosAOT, Ambiq Micro is taking a significant step towards its mission of enabling intelligence everywhere by delivering the lowest power semiconductor solutions. For more information, visit www.ambiq.com.
[1] Ambiq Micro. (2023). HeliosRT and HeliosAOT: Edge AI Runtime Solutions for Apollo SoCs. Retrieved from www.ambiq.com/helios [2] Ambiq Micro. (2023). HeliosRT and HeliosAOT: Revolutionizing Edge AI Computing. Retrieved from www.ambiq.com/blog/heliosrt-heliosaot [3] Ambiq Micro. (2023). Apollo510: The World’s Most Power-Efficient AI Processor. Retrieved from www.ambiq.com/apollo510 [4] Ambiq Micro. (2023). Subthreshold Power Optimized Technology (SPOT). Retrieved from www.ambiq.com/spot [5] Ambiq Micro. (2023). Apollo SoCs: The Future of Low-Power Edge Computing. Retrieved from www.ambiq.com/apollo
- Ambiq Micro's HeliosRT is a power-optimized version of LiteRT designed for the Apollo SoCs family, promising up to 3x faster inference speeds and improved power efficiency, thanks to its custom AI kernels optimized for the Apollo510 vector acceleration hardware.
- The new AI runtime solution, HeliosRT, prioritizes audio and speech models, offering a better support system for efficient AI processing on ultra-low-power devices like wearables, IoT sensors, and industrial monitors.
- HeliosAOT, an ahead-of-time (AOT) compiler from Ambiq Micro, transforms TensorFlow Lite models into embedded C code, reducing memory footprint and providing greater flexibility for resource-constrained edge systems.
- By optimizing memory control and per-layer weight distribution across the Apollo SoC’s memory hierarchy, HeliosAOT allows for tighter integration into embedded applications.
- Both HeliosRT and HeliosAOT leverage Ambiq’s patented Sub-threshold Power Optimized Technology (SPOT) platform, which delivers high AI inference performance and ultra-low energy usage in edge applications like wearables, smart homes, digital health devices, and IoT sensors.
- These solutions are designed for seamless integration with existing AI development pipelines, supporting TensorFlow workflows and providing engineering support to ease developer adoption.
- With the beta release of HeliosRT expected in Q3 2025 and HeliosAOT in technical preview with broader availability planned for Q4 2025, partnership with Ambiq Micro's edge AI solutions demonstrates a commitment to innovation in data-and-cloud-computing and technology.
- Ambiq Micro's technology and solutions in hardware, AI software, and digital services target the competitive edge devices market, focusing on AI performance optimization, power management, and automation of devices.
- The combination of cutting-edge technology, artificial intelligence, and technology innovation puts Ambiq Micro at the forefront of the Internet of Things (IoT), leading to digital transformation and further advancements in the tech industry.
