Edge AI Heats Up: Major Players Race to Redefine Intelligent Devices

Jul 8 2025 2025-07 Sensors STMicroelectronics

In recent weeks, the edge AI industry has seen a wave of strategic acquisitions, signaling a new phase of rapid growth and innovation. Leading semiconductor companies—STMicroelectronics (ST), Qualcomm, and NXP—have all made bold moves to expand their edge AI capabilities, targeting the growing demand for lightweight, efficient AI models that run directly on devices rather than in the cloud.

STMicroelectronics Acquires Deeplite to Bolster Edge AI Stack

ST's acquisition of Canadian AI startup Deeplite has attracted wide industry attention. Deeplite, known for its unique model optimization, quantization, and compression technologies, offers tools that make large AI models smaller, faster, and more power-efficient—ideal for edge devices like drones, cameras, sensors, and vehicles. The company’s core engine, Neutrino™, automatically compresses deep neural networks (DNNs), enabling deployment on resource-constrained environments.

This move is strategically aligned with ST’s rollout of its high-performance STM32N6 series microcontrollers. Deeplite’s software tools will enhance ST’s ability to support AI on microcontrollers (MCUs), an area traditionally served by TinyML, but now evolving rapidly into more powerful, general-purpose edge AI solutions.

Qualcomm Acquires Edge Impulse to Strengthen IoT AI Capabilities

In March, Qualcomm announced its acquisition of Edge Impulse, a prominent edge AI development platform known for its user-friendly, low-code interface and robust toolchain. The platform supports AI model training, deployment, and monitoring, and is used by over 80,000 developers across industries ranging from wearables to industrial monitoring.

Edge Impulse will continue to operate under its brand, now “Edge Impulse, a Qualcomm company,” and extend support for Qualcomm’s Dragonwing processors. The company’s compiler, EON (Edge Optimized Neural), enables AI models that are up to 55% smaller in RAM usage and 35% more compact in storage compared to TFLite Micro, making it a key enabler for battery-powered and memory-constrained devices.

NXP Moves Toward Generative Edge AI with Kinara Acquisition

NXP’s February acquisition of Kinara (formerly Deep Vision) further signals the strategic importance of edge AI. Kinara specializes in discrete neural processing units (NPUs) with high performance and power efficiency. Its Ara-1 and Ara-2 chips support advanced applications such as vision, speech, gesture recognition, and even generative AI on edge devices. The second-generation Ara-2 achieves up to 40 TOPS, consumes under 2 watts, and is tailored for next-gen edge workloads like LLaMA and Stable Diffusion.

Kinara’s SDK provides flexible deployment of trained models, and its programmable architecture ensures scalability across evolving AI paradigms, from CNNs to transformer-based generative models.

The Road Ahead for Edge AI

These acquisitions reflect two major industry trends. First, AI/ML is now central to embedded systems, not just as software but as part of the hardware-software co-design stack. Second, semiconductor manufacturers are expanding beyond chip production, offering full-stack AI solutions—including compilers, development platforms, and software ecosystems—to empower edge intelligence.

As cloud-based AI models become increasingly resource-hungry, edge AI offers a compelling alternative. Analysts predict that by 2025, 75% of data will be processed at the edge, unlocking massive potential for edge AI-enabled MCUs and NPUs.

In an era dominated by large AI models, these moves underscore a growing consensus: the future of AI lies not just in scale, but in accessibility, efficiency, and deployment at the edge.

The Products You May Be Interested In

TPCM-2.4-5	CMC 2.41MH 5A 2LN TH	383 More on Order
AA53002-015	XFRMR TOROIDAL 300VA CHAS MOUNT	464 More on Order
62054-P2S02	XFRMR TOROIDAL 15VA CHAS MOUNT	107 More on Order
62045-P2S02	XFRMR TOROIDAL 10VA CHAS MOUNT	135 More on Order
62044-P2S02	XFRMR TOROIDAL 10VA CHAS MOUNT	336 More on Order
62043-P2S02	XFRMR TOROIDAL 10VA CHAS MOUNT	304 More on Order
62034-P2S02	XFRMR TOROIDAL 7VA CHAS MOUNT	441 More on Order
62025-P2S02	XFRMR TOROIDAL 5VA CHAS MOUNT	218 More on Order
62021-P2S02	XFRMR TOROIDAL 5VA CHAS MOUNT	281 More on Order
62012-P2S02	XFRMR TOROIDAL 3.2VA CHAS MOUNT	178 More on Order
62075-P2S02	XFRMR TOROIDAL 35VA CHAS MOUNT	406 More on Order
62073-P2S02	XFRMR TOROIDAL 35VA CHAS MOUNT	294 More on Order
70072K	XFRMR TOROIDAL 35VA THRU HOLE	313 More on Order
70041K	XFRMR TOROIDAL 10VA THRU HOLE	157 More on Order
70015K	XFRMR TOROIDAL 3.2VA THRU HOLE	421 More on Order
70013K	XFRMR TOROIDAL 3.2VA THRU HOLE	244 More on Order
70011K	XFRMR TOROIDAL 3.2VA THRU HOLE	420 More on Order
62015-P2S02	XFRMR TOROIDAL 3.2VA CHAS MOUNT	210 More on Order
62084-P2S02	XFRMR TOROIDAL 50VA CHAS MOUNT	607 More on Order
62060-P2S02	XFRMR TOROIDAL 25VA CHAS MOUNT	3963 More on Order
70005K	XFRMR TOROIDAL 1.6VA THRU HOLE	401 More on Order
AC1200	CURR SENSE XFMR 200A T/H	119 More on Order
AC1050	CURR SENSE XFMR 50A T/H	409 More on Order
AC1015	CURR SENSE XFMR 15A T/H	287 More on Order