Qualcomm Technologies has signed a deal to acquire US-based Edge Impulse, which makes miniaturised machine learning (ML) tools for microcontrollers (MCUs) in resource-constrained IoT modules. It is the latest move by the chipmaker to reorganise its IoT strategy and portfolio to combine hardware, software, and services to “scale across multiple verticals”. Last week, ahead of MWC, it introduced a new brand, Dragonwing, to give its IoT products their own identity, and to stand apart from its regular Snapdragon consumer units. The deal with Edge Impulse, a regular on the low-power IoT circuit, is timed with its appearance at Embedded World in Germany this week.
Qualcomm has also picked up the LTE portfolio of France-based IoT specialist Sequans, at the end of last summer. It explained its IoT strategy, now united under the Dragonwing brand, to put squarer focus on an ‘end-to-end’ chipset roadmap, software architecture, services suite, developer resources, ecosystem partnerships, optimised solutions, and industrial blueprints. All in, the message is Qualcomm is (more) serious about IoT – as if the market, including with the rise of private 5G networks and first sightings of RedCap in Industry 4.0, is ready to blow open. At least, that is how it reads. As with the proposed Edge Impulse deal, it is very clearly also about AI at the edge.
Edge Impulse’s no-code platform includes tools and features for data collection and preparation, and model training, deployment, and monitoring in IoT devices. It has been available since 2019, when on-device edge AI was popularly called ‘tiny ML’, to describe a more-miniaturised version of edge AI to go even in battery-powered IoT devices. As it is, on-camera AI/ML for computer vision applications is the standard application for tiny ML and edge AI, further orchestrated with distributed compute nodes within an enterprise-based edge setup, and linking to the cloud as required / permitted.
In a blog post about the Qualcomm deal, Zach Shelby, co-founder at Edge Impulse, wrote: “Recognising that the compute capabilities of MCUs had grown to the point where they were able to run domain specific AI models directly onboard, we realised there were an endless number of use cases that would benefit from moving AI from the cloud to the edge. But there was a missing piece: a way to easily build, optimise, and deploy edge and domain-specific AI models onto these devices.” Which is the firm’s whole origin story, right there – to remove the manual developer processes to collect and format data, hand-code and train architectures, and convert and compile models.
And then to test for viability. “[We] set out to democratise ML on edge compute for millions of developers… – to give developers a tool to automate data collection, simplify model training, provide advanced optimisation tools, and offer one-click deployment to many types of hardware, from MCUs to CPUs, GPUs, and NPUs,” said Shelby. The deal will mean its developers, which number about 170,000, can “target Dragonwing platforms” for AI inferencing, applications, and other processing capabilities, said Qualcomm. Target IoT sectors for Qualcomm are industrial, logistics, security, healthcare, retail, energy, and sundry “enterprise” markets.
Nakul Duggal, group general manager for embedded IoT at Qualcomm, said: “Edge Impulse’s advanced AI-powered end-to-end platform… will complement our strategic approach to IoT transformation. This acquisition will strengthen our leadership in AI and developer enablement, enhancing our ability to provide comprehensive technology for critical sectors… IoT opens the door for a myriad of opportunities, and success is about building real-world solutions, enabling developers and enterprises with AI capabilities to extract intelligence from data, and providing them with the tools to build the applications and services that will power the digital transformation of industries.”
Edge Impulse currently supports Dragonwing QCS6490 and QCS5430 processors; it will add support for additional Dragonwing processors for industrial and embedded IoT applications. Edge Impulse will maintain its current website (branded as “Edge Impulse, a Qualcomm company”) and remains dedicated to supporting developers and ecosystem partners. The deal is subject to customary closing conditions.
