While generative AI is sweeping the globe with its technological revolution, another more subtle yet equally pivotal technological shift is emerging: edge AI, or as it is now widely known - edge-side AI. If edge AI focuses on the decentralization of computing resources, edge-side AI involves deploying and running AI models directly on devices to enable rapid data processing and intelligent decision-making. Both belong to different facets of "distributed intelligent computing" with significant conceptual overlap.

Recently, significant actions by tech giants, such as NXP's acquisition of Kinara for a substantial sum and Qualcomm's announcement to acquire Edge Impulse, have sent a clear message: edge AI/edge-side AI is transitioning from a "proof-of-concept" to a "strategic pillar". Behind this technological shift lies not only the inevitable trend of computational restructuring but also a fresh wave of global competition centered around data sovereignty, real-time decision-making, and energy efficiency balance. Hence, this article delves into the strategic rationale behind these lavish investments by tech titans in acquiring leading edge AI/edge-side AI enterprises.

Tech Titans Rush to Deploy Edge AI/Edge-Side AI

On February 11, 2025, Dutch chipmaker NXP announced the acquisition of US-based edge AI chip startup Kinara Inc. for $307 million. Kinara specializes in developing neural processing units (NPUs) for AI workloads at the network edge.

Kinara's discrete NPUs (including Ara-1 and Ara-2) are industry leaders in performance and energy efficiency, making them ideal for emerging AI applications like vision, speech, gestures, and various multimodal implementations driven by generative AI. Both devices feature an innovative architecture that maps inference graphs for efficient execution on Kinara's programmable NPUs, maximizing edge AI performance. Additionally, Kinara provides a comprehensive software development kit to facilitate customers in optimizing AI model performance and streamlining deployment processes.

NXP stated that by integrating Kinara's discrete NPUs with its own processor, connectivity, and security software portfolio, this acquisition will enable it to offer a "complete and scalable AI platform from TinyML to generative AI".

NXP is not alone in valuing TinyML. On March 11, 2025, Qualcomm announced the acquisition of edge AI technology company Edge Impulse, aiming to integrate Edge Impulse's edge AI development platform and enhance Qualcomm's capabilities in artificial intelligence (AI) and the Internet of Things (IoT).

Edge Impulse, a star in the TinyML (Tiny Machine Learning) field, was founded by Zach Shelby and Jan Jongboom in 2019. They recognized that microcontroller computing power had evolved to the point where domain-specific AI models could run directly on devices. Despite mature hardware, there was a lack of a simple method to build, optimize, and deploy edge and domain-specific AI models to these devices. Thus, Edge Impulse emerged, and the AIoT platform it built significantly reduces the time required to create machine learning models for small devices like sensors, microcontrollers, and cameras.

Through this acquisition, Qualcomm will integrate Edge Impulse's end-to-end edge AI development platform into its IoT ecosystem.

Looking further back, other notable acquisitions include:

In August 2024, Amazon acquired chip manufacturer and AI model compression company Perceive for $80 million in cash. Perceive, known for its groundbreaking neural network inference solutions, focuses on providing technology for large AI models on edge devices. Its flagship product, the Ergo AI processor, can run data center-level neural networks in various environments, even under power constraints.

In July 2023, NVIDIA acquired AI startup OmniML, which launched a platform named Omnimizer. This platform compresses machine learning models to enable large models to run on smaller devices without relying on cloud computing power.

As future trends become clearer, tech giants are accelerating their pace. By acquiring and integrating edge AI/edge-side AI technology, they continue to solidify their leading position in the AI field, meet growing market demands, and propel technological innovation and application deployment.

Strategic Considerations at Play

Analyzing these acquisitions provides insights into the strategic rationale:

① Semiconductor Giants Lead the Charge

Among tech giants, semiconductor companies have reacted most strongly and swiftly to edge AI/edge-side AI. Through mergers and acquisitions, they quickly acquire core technologies to capture high-growth markets.

From a strategic perspective, cloud AI faces challenges like high computing costs, latency, and privacy risks. Edge AI/edge-side AI mitigates these issues by processing data locally, reducing bandwidth requirements and response time, enhancing data privacy, and lowering costs. Semiconductor companies help customers transcend cloud limitations by providing edge chips and solutions. Additionally, against the backdrop of saturated traditional PC and mobile markets, edge AI/edge-side AI presents a new incremental market for chip manufacturers, including high-potential scenarios like industrial automation, smart cities, intelligent driving, and intelligent security. For semiconductor manufacturers like Qualcomm, NXP, Intel, and AMD, deploying edge AI/edge-side AI is not just about seizing the next-generation AI computing entry point but also a practical path to extend the benefits of "Moore's Law".

For instance, NXP's significant investment in Kinara yields products delivering high-efficiency AI performance across various neural networks, including traditional and generative AI, meeting the rapidly growing AI demand in industrial and automotive markets. Similarly, Qualcomm's acquisition of Edge Impulse enhances its ability to provide comprehensive technologies for key areas like retail, security, energy and utilities, supply chain management, and asset management.

Clearly, edge AI's development is driving changes in the hardware ecosystem, becoming a "growth catalyst" for chip companies.

Furthermore, edge AI/edge-side AI's demands for high-performance chips, low-latency processing, and security rely directly on advancements in semiconductor manufacturing processes and hardware innovation. As AI migrates from the cloud to the edge, semiconductor companies, through technology investment and ecological cooperation, not only consolidate their core position in the industrial chain but also promote the transformation and upgrading of the entire industry. In the future, as edge AI devices (like smartphones and embodied robots) become more prevalent, the semiconductor industry will further benefit from this technological wave.

② TinyML: A Crucial Path for Edge-Side AI

Several of the aforementioned acquisitions involve a crucial technology - TinyML. TinyML refers to deploying machine learning models on resource-constrained embedded devices (like microcontrollers, sensor modules, etc.) to achieve localized, low-power, and low-latency intelligent inference.

Judging by the actions of tech giants, as TinyML technology matures, it will become a significant implementation path for edge AI/edge-side AI, finding widespread use in smart homes, wearable devices, industrial IoT, and other cost-, power consumption-, and response speed-sensitive scenarios.

TinyML's core goal is to run machine learning models on microcontrollers (MCUs) with milliwatt (mW)-level power consumption. It is currently the only technical path capable of stably running AI models on hardware with milliwatt-level energy consumption + KB-level storage + extremely low cost, naturally aligning with the typical deployment requirements of edge AI/edge-side AI. As AI applications shift from "centralized cloud inference" to "edge intelligence, edge-side intelligence", the focus of intelligent deployment shifts, and TinyML-supported deployment platforms (like Cortex-M series, RISC-V MCU, DSP cores, etc.) become the end nodes in the distributed intelligent computing network. TinyML empowers these tiny nodes with basic AI perception and judgment capabilities, realizing true "ubiquitous intelligence".

Moreover, past AI was often confined to large cloud models and high-performance computing centers, with high technical thresholds and development costs. The TinyML ecosystem is driving: ① low-code/visual modeling tools, enabling non-AI professionals to train and deploy models; ② generalization of hardware platforms, allowing one-click model migration; ③ separation of training and inference, enabling cloud-based model training and independent device-end inference, reducing device costs. This makes it possible for AI to be "developed and deployed like embedded software", significantly accelerating the popularization of edge intelligence.

With its unique advantages in low power consumption, real-time performance, privacy protection, hardware adaptability, and wide application scenarios, TinyML is one of the core technical supports for edge/edge-side AI. Without TinyML, there would be no true "edge ubiquitous intelligence" or "edge-side ubiquitous AI".

③ Emphasis on Large-Small Model Complementarity + Edge-Cloud Collaboration

While edge/edge-side AI has its advantages, its limited understanding capabilities make it challenging to handle advanced tasks like complex reasoning and semantic understanding. Large models, while powerful, come with substantial computing costs, making them unsuitable for edge or terminal deployment and real-time responses. Thus, large models excel in cognitive reasoning and complex generation tasks, while edge lightweight models dominate real-time perception and local responses, complementing each other well.

In the future ubiquitous intelligence architecture, edge lightweight models provide "fast" responses, while large models offer "deep" understanding. Only through collaboration between the two can an efficient and reliable AI system be constructed. Tech giants are accelerating their comprehensive layout of edge deployment capabilities and model management platforms precisely because they recognize this evolutionary trend.

For example, shortly after Qualcomm acquired Edge Impulse, on April 2, it announced the acquisition of Vietnamese AI startup MovianAI through its official website. Formerly the generative AI department of VinAI Applications and Research Joint Stock Company (VinAI), MovianAI is a leading AI research company renowned for its expertise in generative AI, machine learning, computer vision, and natural language processing.

This transaction creates a synergistic effect with Qualcomm's previous acquisition of Edge Impulse: MovianAI specializes in large-scale AI models for smartphones and automobiles, while Edge Impulse optimizes AI for resource-constrained small devices, providing Qualcomm with brand-new "end-to-end" expertise across the entire AI value chain.

Final Thoughts

According to new research by Future Market Insights, the edge AI market will reach $39.6 billion by the end of 2032, expanding at a CAGR of 20.8% over the forecast period.

As the market grows, the development of edge/edge-side AI is progressing towards a higher level of intelligent collaboration. The optimization of chip architectures, the maturity of TinyML technology, and the continuous evolution of cloud-edge collaboration mechanisms will contribute to the formation of a ubiquitous AI landscape where "intelligence is local and even smarter in the cloud". For enterprises aiming to seize the initiative in the AIoT era, an early layout of edge intelligence and model collaboration architecture has become an indispensable strategic choice.

References:

Qualcomm acquires edge AI development platform Edge Impulse to strengthen AIoT, Electronic Engineering Album

NXP plans to acquire Kinara, a pioneer in edge AI, to redefine the intelligent edge, NXP

NXP acquires Kinara for $307 million to deploy edge network AI computing business, Top Networks

Qualcomm buys MovianAI to boost generative AI capabilities, RCRWirelessNews

Giants enter TinyML, and edge-side and edge AI usher in a new turning point, IoT Intelligence