Qualcomm’s Big Bet On Edge AI Could Reshape The IoT Market (2025)

AI is expanding beyond the cloud, turbocharging industrial equipment and computing systems at the edges of enterprise networks — and Qualcomm is hitting the gas. On March 10, Qualcomm agreed to buy Edge Impulse, a leading provider of AI development tools for IoT and embedded devices. With a community of over 170,000 developers and a history of over 474,000 machine learning projects, this acquisition jumpstarts Qualcomm’s edge AI software ecosystem.

But that’s only part of the story. The company is transforming edge AI systems from custom embedded mashups to scalable, AI-enhanced computing platforms, as evidenced by these recent milestones:

• March 2024: Foundries.io acquisition — Embedded Linux OS tools and services
• October 2024: IQ Series — Industrial-grade SoCs and IoT Solutions Framework
• February 2025: Dragonwing — Industrial AI umbrella brand
• March 2025: Edge Impulse acquisition — Edge AI development framework
• March 2025: Palantir collaboration agreement — Enterprise AI on Qualcomm edge AI platforms

These milestones address edge scalability problems that have plagued IoT products for over a decade, paving the way for AI at the edge. The result is an AI-focused industrial edge platform that lets software developers focus on applications rather than OS customizations, system configurations, security subsystems, DevOps toolchains and AI workflows. Here’s my analysis of this strategy, beginning with the business motivations for edge AI.

Data Is Driving Edge AI

While generative AI and large frontier models grab the spotlight, a quiet revolution is intensifying around the edges of our computing universe. AI is finding its way into the world all around us. Smaller, domain-optimized AI models, hosted by a wide variety of IoT devices, bring intelligence to all kinds of products — industrial, robotic, wearable, vehicular, agricultural, smart home, smart city, healthcare and human interface, to name a few. AI at the edge is growing rapidly, and here I’ll explain why.

According to an unsubstantiated legend, when a reporter asked notorious bank robber Willie Sutton why he robbed banks, he replied, “Because that’s where the money is.” Likewise, if you ask CIOs of enterprises with physical assets why they are betting big on edge AI, they’ll probably say, “Because that’s where the data is.” Edge intelligence integrates real-time operations data with enterprise analytics, transforms resource planning from reactive to proactive, expands the scope of industrial automation from individual machines to process orchestration, and upgrades human roles from operators to innovators. These and other benefits promise significant ROI from investments in edge AI.

AI-powered, cloud-based ERP, BI, CRM and SCM applications already grab data from edge IoT infrastructure, so why add AI to the complex, rugged, fragmented world of operational technologies? Why not just bring the data to the AI, rather than extend AI to the data? Here are five good reasons:

• Latency — Edge AI enables real-time analytics and closed-loop process control.
• Reliability — Cloud service interruptions and network outages can be costly and dangerous.
• Privacy — Regulatory and confidentiality requirements often require data to remain on-premises.
• Longevity — Embedded OT systems must run for the lifetime of the host equipment with minimal external dependencies.
• Cost savings — Edge AI reduces costs for cloud inference and network bandwidth.

In addition, conversational AI chatbots, agents and virtual assistants enable employees to interact with automated systems using natural languages, gestures and physical activities rather than keyboards, control panels and GUIs. AI is a new UI for integrated industrial operations, making employees more effective, reducing training time and providing in-context assistance when and where needed.

The Scalability Strategy Behind Edge Impulse, Dragonwing And Foundries.io

The business cases are compelling, so what’s holding back a wave of edge AI applications? OT diversity is the biggest reason. Unlike the uniform, button-down IT world, OT devices must meet stringent requirements for disparate industries, environments, form factors and operational situations. Hence, OT products are typically customized, one-off designs with unique hardware and software configurations.

The extreme diversity of embedded devices prevents large-scale OT platform evolution. Smartphones have app stores with plug-and-play applications, and PCs have thriving independent software vendor supply chains, but OT devices typically require customized, product-specific applications. There is no common platform for building, testing, securing, delivering and updating OT applications — hence, no embedded “app store” and no economy of scale for OT device software.

Diversity is a fact of life for IoT and OT products — that will never change. Still, we can improve how we manage diversity by providing off-the-shelf, scalable platforms that span families of architecturally similar SoCs. Qualcomm is on the leading edge of OT platform evolution with essential components of an edge AI platform that lets developers focus on AI-enabled applications rather than embedded system engineering.

The following sections explain how the Foundries.io OS, Dragonwing IQ series chips and Edge Impulse AI development tools comprise a scalable platform for edge AI. I’ll also explain how the recently announced Palantir collaboration suggests the next step for this platform — a multivendor application ecosystem.

Foundries.io — OS And System Software

Traditional software development on Linux-based edge (embedded) systems requires customizing system-level platform code such as OS components, hardware drivers, network stacks, security systems and lifecycle management services. System customization delays application development, wastes engineering resources on undifferentiated features and creates significant technical debt by saddling product companies with long-term system support, perpetual software updates and security liabilities.

Foundries.io is a Linux lifecycle management company founded in 2017 and acquired by Qualcomm in 2024. The company’s main product, Foundries Factory, is a “DevSecOps” (development, security, operations) toolchain that aims to be “the OS of every thing.” The tools make it easy for application developers to create a Yocto-configured Linux OS distribution for use on the target product. The generated OS is production-ready, secure, open-source, cloud-supported and continuously updated. Foundries Factory lets embedded developers write production-ready application code immediately after unboxing the development kit, deploy the resulting software stack at scale and continually update the OS, applications, AI models and other components.

Dragonwing IQ Series — Industrial-Grade SoC Brand

I don’t often write about branding initiatives, but Dragonwing isn’t a typical fluffy institutional campaign. It’s strategically significant because it defines a unique identity for Qualcomm’s industrial products outside the consumer-facing Snapdragon umbrella, signaling portfolio alignment, focused innovation, unified messaging and a strong customer commitment. Edge AI is the forcing function driving industrial portfolio alignment, so this move makes perfect sense.

Dragonwing IQ series and Q series industrial SoCs combine Kryo CPUs, Hexagon tensor processors (up to “100 TOPS”), Spectra image processors, Adreno GPUs, ECC memory (rare for this kind of product) and an “MCU-like” subsystem with an industrial-grade “safety island” (SAIL), plus all the usual function blocks to bring accelerated AI applications to physically demanding, high-reliability edge environments. Details are available here.

Edge Impulse — AI Framework

Enterprises with physical assets derive significant business value from edge AI use cases such as production optimization, supply chain efficiency, predictive maintenance, quality monitoring, safety and compliance. In addition to improving productivity and reducing costs, these AI projects deliver significant ROI by transforming process automation from reactive to proactive and expanding enterprise resource planning from hindsight to foresight. However, these AI-powered applications require training and tuning custom AI models for each use case and physical environment. Developing those models is a bottleneck inhibiting the edge of the AI scale.

Edge Impulse simplifies and accelerates edge AI model development, “democratizing” the process so that more developers with operational knowledge can participate. The company has cultivated a community of over 170,000 developers, which only scratches the surface. The strong ROI for edge AI means we need millions of people training edge AI models — domain experts, not just data scientists. The first spreadsheet applications democratized data analysis over 40 years ago, and now we need a similar approach to AI training.

Edge Impulse simplifies and streamlines edge AI development from data collection through model deployment. The company’s website has complete technical details about these workflows, but at a high level, the process flows through these steps:

• Collect, process and manage data from real-world sensors, including cameras
• Extract features from sampled data
• Use those features to train AI and ML models
• Tune and optimize the models
• Test and validate model performance
• Optimize models for specific edge AI devices
• Deploy models in the field

Broad industry acceptance of Edge Impulse requires supporting these workflows across many types of edge AI devices from multiple suppliers — not just Dragonwing. According to CEO Zach Shelby, the company will continue supporting a variety of hardware. He described a “best of both worlds” approach: “Adding support for Dragonwing gives us superpowers on that platform.” Meanwhile, Dragonwing benefits from broad Edge Impulse adoption, which requires multivendor platform support, so this strategy makes sense. Shelby said it best at Embedded World: “There are five exhibit halls and thousands of vendors. That’s the world of embedded systems.”

The Edge Impulse acquisition signals Qualcomm’s commitment to industrial AI growth. And Edge Impulse will now have the resources to rapidly expand its user base beyond AI experts and enable people with operational expertise to develop and tune edge AI models.

Palantir — Enterprise AI On Edge Devices

Based on the agreement just announced, Palantir is extending its Ontology operations data layer to the Dragonwing platform, enabling OT-driven real-time insights and bringing data-driven, proactive decisions to industrial environments, including remote locations.

Qualcomm’s Palantir collaboration proves the importance of the industrial edge platform strategy discussed in the “The Scalability Strategy . . .” section above. It’s a great example of bringing AI to the data rather than bringing the data to AI. Platforms enable ISVs like Palantir to support the daunting diversity of edge devices with little or no customization.

Qualcomm is Positioned To Reshape IIoT And Edge AI

Qualcomm is committed to enabling edge AI across the board with integrated tools, native OS support, built-in OTA updates, Dragonwing industrial AI SoCs and edge AI workflows. These investments add up to a platform for edge AI development and deployment that accelerates Dragonwing adoption, expands the developer community and has the potential to redefine the industrial IoT supply chain by enabling an ISV software ecosystem.

Qualcomm’s edge platform is a wake-up call for the embedded semiconductor industry. The days of selling bare metal chips are over. The future belongs to platforms that enable application developers to rapidly build secure, production-ready, edge AI applications with little or no system-level engineering — just like their cohorts in the IT and mobile sectors have done for decades. Product companies will use platforms, not build them. And Qualcomm is building those platforms.