The popularity of distillation models sparked by DeepSeek has drawn increased attention to deploying large AI models on edge devices. While TinyML has long been a focus in the MCU domain, edge AI is now emerging as the next market frontier, prompting major semiconductor manufacturers to accelerate their strategic deployments. In a recent development, STMicroelectronics, a leading Microcontroller manufacturer, acquired Deeplite to strengthen its edge AI capabilities. This move follows similar strategic acquisitions by industry giants like Qualcomm and NXP Semiconductors, underscoring the growing competition in edge computing AI solutions.
STMicroelectronics Acquired Deeplite
Recently, STMicroelectronics acquired Deeplite, an AI startup in Toronto, which has attracted industry attention. Deeplite is known as DeepSeek for edge AI. It has unique technologies in model optimization, quantization, and compression, which can make large AI models run faster, smaller, and more energy-efficient on edge devices. These technologies can also help accelerate the adoption of ST's first high-performance STM32N6.
Deeplite's software makes it easier to run AI applications on chips in devices such as mobile phones and robots. Its customer base is mainly semiconductor companies, which is highly consistent with STMicroelectronics' business, which focuses on designing and manufacturing chips for various customers.
STMicroelectronics acquired Deeplite
A significant challenge when deploying deep learning models commercially is how large, processor-intensive, and power-intensive they are to run. Deeplite solves this problem by providing an automated software engine to optimize DNN (deep neural network) models and enable AI for edge computing on any device.
Deeplite's product is a bit like an AI nesting doll. Using AI to automatically make other AI models faster, smaller, and more energy-efficient, it creates highly compact, high-performance deep neural networks for deployment on edge devices such as security cameras, sensors, drones, mobile phones, and vehicles.
Deeplite has been recognized as a top-edge AI innovator by Gartner, Forbes, Inside AI, and ARM AI. It launched in 2017 through the TandemLaunch incubator and became an independent technology company in 2019 when co-founders Dr. Ehsan Saboori, Davis Sawyer, and Nick Romano united to democratize AI implementation in everyday applications. The company gained significant traction following its 2020 mid-year release of the Neutrino™ software platform, which has since seen strong adoption by major OEMs, semiconductor manufacturers, and application developers. This automated optimization engine integrates with industry-standard MLOps frameworks like PyTorch, ONNX, and TensorFlow, enabling AI engineers to develop ultra-compact, energy-efficient models that reduce cloud infrastructure costs while empowering new battery-operated edge computing applications. According to PitchBook data, the Canadian startup has secured $6.47 million in funding to date, backed by prominent investors including BDC Capital, Desjardins Capital, and Somel Ventures.
Although the specific terms of the acquisition were not disclosed, Deeplite CEO Nick Romano announced on LinkedIn that Deeplite has become a subsidiary of STMicroelectronics. By combining Deeplite's advanced edge AI software solutions with ST's most advanced MCUs and NPUs, ST will offer one of the most advanced edge AI platforms in the world.
Qualcomm Announces Acquisition of Edge Impulse
In March 2025, Qualcomm announced that it would acquire Edge Impulse, an edge AI development platform, to expand its AI capabilities for IoT-enabled products.
Edge Impulse will retain the brand and maintain its current website. The collaboration with Qualcomm will accelerate support for Qualcomm Dragonwing processors, and Edge Impulse's platform will continue to be open to MCU, CPU, GPU, and NPU users from the company's hardware partners.
Source from Internet
Qualcomm Dragonwing processors feature on-device AI inference, computer vision, imaging, and processing capabilities. Edge Impulse currently supports the Dragonwing QCS6490 and QCS5430 processors, with plans to add support for other Dragonwing processors for industrial and embedded IoT applications.
Edge Impulse's development platform includes tools for data collection and preparation, model training, deployment, and monitoring, with low-code or no-code interfaces. Developers use Edge Impulse's platform to add AI capabilities such as computer vision, time series data, audio events, and speech recognition to embedded systems in asset tracking and monitoring, manufacturing, anomaly detection, and predictive maintenance systems.
Edge Impulse was founded in 2019 with TinyML as its service. Its founders, Zach Shelby and Jan Jongboom, who worked for Arm before, are committed to providing the latest machine learning tools to enable all companies to build smarter edge products.
Edge Impulse solutions are widely used by health wearable device manufacturers such as Oura, Know Labs, and NOWATCH, industrial organizations such as NASA, and top chip suppliers. They are adopted by more than 80,000 developers and have become a trusted platform for enterprises and developers.
Edge Impulse features an Edge Optimized Neural (EON) compiler. According to its official website, the neural network inference model compiled by this compiler can use 25-55% less RAM and 35% less storage space than TFLite Micro. In addition, using their digital signal processing block (DSP Block) to pre-process the sound before inference can complete the inference faster and more accurately. Taking bird call recognition as an example, the speed is 48% faster and the accuracy is increased by 7%.
NXP Acquires Kinara
NXP announced the acquisition of Kinara for $307 million in all cash. This is an American semiconductor company that was founded in 2013 and was originally called Core Viz. After its establishment, it was renamed Deep Vision and renamed Kinara again in 2022.
Kinara's discrete NPUs, including Ara-1 and Ara-2, lead the industry in performance and energy efficiency. This makes them the preferred solution for emerging AI applications such as vision, voice, gesture and various other generative AI-driven multi-modal implementations. Both devices use an innovative architecture that supports mapping inference graphs for efficient execution on Kinara's programmable proprietary neural processing units, thereby maximizing edge AI performance. This programmability ensures the adaptability of AI algorithms as they continue to evolve from CNN to new methods such as generative AI and agent AI in the future.
The Kinara Al Software Development Kit (SDK) optimizes the application of trained A! models to Ara-1 and Ara-2 silicon and modules. Kinara's fully programmable compute engine allows our model compiler to quickly adapt to a virtually unlimited range of neural network architectures.
Conclusion
Industry experts have maintained that the world requires fewer large-scale models, asserting that true AI implementation emerges through edge computing and end-device deployments. It is predicted that by 2025, 75% of data will be processed at the edge, and the market potential of edge AI MCU is huge.
AI/ML technology is now an important part of the hardware and software stack in embedded system design. Chip manufacturers' AI strategies provide complementary tools for their semiconductor devices to meet the full range of embedded AI/ML learning needs.
In addition to the above three manufacturers, other MCU manufacturers are also acquiring startups in this field. For example, Renesas Electronics acquired Reality AI, Infineon acquired Sweden's Imagimob, and NXP launched machine learning software eIQ and AI tool chain NANO.
The demand for edge AI computing is growing rapidly, and MCU, as a core component of edge devices, will play an important role in this trend. It is expected that more manufacturers will join the competition in the field of edge AI in the follow.