For AI to be fully integrated into consumers’ daily lives, end-user devices must be able to process AI workloads rather than relying solely on cloud computing. On-device AI reduces costs and latency, improves personalization and privacy, and enables real-time processing, all of which are essential for bringing AI to everyday applications and use cases. As a result, edge AI, which is closely related to the Internet of Things (IoT), is expected to continue to gain momentum.
Edge AI is spreading across a variety of applications in IoT and consumer devices, including smartphones, wearables, headsets, drones, industrial technology, and automobiles. It is estimated that by 2027, 62% of data will be processed on edge devices. It is also estimated that by 2028, 26 billion short-range IoT connections will be powered by on-device AI, benefiting a range of associated technologies that enable the infrastructure of connected devices.1,2 As a result, we believe Edge AI will be an attractive investment for investors looking for AI opportunities outside of the semiconductor companies and cloud hyperscalers that have traditionally dominated the AI space.
Key Takeaways
The rise of generative AI is accelerating and expanding the drive for greater computing power directly on consumer hardware. On-device AI integration is expected to reverse the downward trend in smartphone sales and drive the next major upgrade cycle, especially for Apple’s iPhones. As consumer behavior becomes increasingly receptive to AI technology, various devices across multiple industries will require upgrades to integrate AI capabilities, ensuring that all connected devices are equipped with advanced capabilities.
AI offers the largest near-term monetization opportunity of any consumer device in smartphones, a $500 billion annual market, by accelerating replacement cycles and driving AI-related premiums.3 Nearly every major smartphone provider, including Apple and Samsung, already has a dedicated AI strategy in place.
Apple recently unveiled Apple Intelligence, a suite of AI-powered features for new iPhones, iPads, and Macs. 4 Unlike the chatbot-centric approach that many AI companies have adopted thus far, Apple’s AI focuses on personalization and app integration to improve the overall user experience. These tools include an improved Siri, powered by an in-house Large Language Model (LLM) and OpenAI’s GPT-4o. 5 In our view, Apple has established itself as the most sophisticated consumer digital agent with the new generative AI-enabled Siri, where enhanced natural language processing (NLP) and in-app command capabilities generate more accurate responses and task execution across iPhone apps. Other features include enhanced predictive text, advanced photo and video editing, and generative tools that have the potential to redefine how users interact with their devices. 6
Samsung’s AI approach for smartphones is focused on communication and productivity, with a focus on specific tasks such as language translation. In January 2024, Samsung unveiled Galaxy AI, a collection of AI tools designed to combine on-device AI with third-party cloud-based AI tools to improve the user experience. 7 For example, Galaxy AI enables the Samsung keyboard to offer translation capabilities in all apps and enables Galaxy phones to handle live call translations, acting as a personal translator. 8 Additionally, several photo editing features allow users to quickly edit, change, or rearrange objects in an image. Another tool is Circle to Search, which allows users to launch a Google search by circling something on the screen with their finger. 9
The integration of generative AI is expected to reverse the slump in smartphone sales and drive a significant upgrade cycle. For Apple, the impact could be comparable to the launch of the iPhone 12, the first 5G iPhone, in 2021, and the iPhone 6 series, which features larger screens. Importantly, Apple Intelligence will be limited to devices with A17 Pro or M-series chips, i.e. just 5% of the current iPhone install base. This means that 95% of iPhone users will need to upgrade their devices to experience Apple Intelligence.
Historically, most U.S. users have replaced their smartphones every two to three years.11 In recent years, however, Apple’s iPhone replacement cycles have been longer. This is likely due to a combination of macroeconomic headwinds that have impacted consumer spending and a slower pace of innovation. That said, the most recent iPhone supercycle, which saw the launch of the iPhone 12 and 5G about three years ago, is in line with the historical average.12
The iPhone 12 topped 100 million units sold in just seven months, two months faster than the iPhone 11, which was released a year earlier.13 The iPhone 12 also nearly matched the sales volume of the iPhone 6 series, driving Apple’s first major super cycle.14 AI-enabled iPhones are also expected to accelerate replacement cycles, given that users will have a strong incentive to upgrade to AI-powered personalized handheld devices.
As the AI smartphone story progresses, we believe IoT companies in the smartphone supply chain, including systems-on-chip (SoC) and memory players, are well-positioned. Smartphones are evolving from processing AI models with hundreds of millions of parameters to processing models with more than a billion parameters.15 Future models are expected to process, or perform inference, on models with more than 10 billion parameters.16 High inference costs have hindered the adoption of early consumer generative AI products. However, the ability to process more parameters allows model providers to move low-level tasks from data centers to smartphones to reduce inference costs and enable developers to leverage edge computing for diverse front-end consumer experiences.
Beyond smartphones, generative and edge AI could unlock demand for a variety of more powerful machines over time across industries and use cases. In Q2 2024, Microsoft unveiled Copilot+ PC, the first AI PC with an on-device AI assistant and capable of performing over 40 trillion operations per second (TOPS).17 In Q1 2024, AMD saw a strong 85% increase in chip sales year-over-year, driven by robust demand for Ryzen 8000 series CPUs tailored for desktop and laptop applications.18 As manufacturers ramp up production, an estimated 54.5 million AI-embedded PCs are expected to ship in 2024, accounting for nearly a quarter of total PC spending.19 By 2027, 60% of PC shipments could have AI capabilities.20
Wearables that measure biometric and healthcare data could also benefit from on-device AI. From 2016 to 2022, the number of connected wearable devices worldwide grew 240%, from 325 million to 1.1 billion.21 However, much of the data processing mediated by these devices takes place in the cloud or on connected smartphones. Computing power costs are high and latency can be an issue, negatively impacting feedback loops and user experience. As AI models become more sophisticated and personalized, on-device data processing should improve, leading to a better relationship between device and user. One of the most promising potential use cases is remote patient monitoring.
Industrial IoT has great growth potential, with one forecast indicating that the global market could grow from $394 billion in 2023 to $1.7 trillion by 2030, at a compound annual growth rate (CAGR) of 23%.22 Humans still input a lot of information into machines, which leads to inefficiencies, so advances in AI are expected to help, especially through automating manual tasks. Even in a technology-driven world, only 31% of manufacturing companies had a single fully automated process in 2020.23 For AI to drive more efficient processes and improve productivity, IoT nodes are essential, providing the hardware layer required to capture and process data. IoT sensors generate data that AI can collect and analyze for predictive maintenance, quality control, and workplace safety. Sensor data can detect anomalies early, predict machine failures, and proactively alert employees of required maintenance. By facilitating microsecond decision making, manufacturers can catch issues in real time on the assembly line and ensure higher quality in large scale production.
Increased investment to support the reshoring of manufacturing is also driving demand for industrial IoT. AI chipsets used directly in industrial equipment reduce the amount of data processed in the cloud, thus reducing costs and network traffic, and improving security. Companies with a diverse range of end markets, such as NXP Semiconductor, reported strong growth in industrial and IoT revenue in the first quarter of 2024, up 14% year over year.24
Edge computing is also important for autonomous vehicles (AVs) to make decisions in real time without relying on a stable internet connection or cloud servers. AVs need vast amounts of pre-processed real-world data to train computers to operate the vehicle, so they use sensors like video cameras, LiDAR, and ultrasound to understand their surroundings. Companies developing the AI infrastructure and advanced driver assistance systems (ADAS) needed to facilitate this process include Tesla and Mobileye. High-speed connectivity is also required for constant updates and notifications of road conditions. Tesla has designed its own computer chips for these purposes, and others are expected to follow suit, benefiting the custom silicon market.
In 2023, the buildout of generative AI has largely benefited computing hardware and cloud infrastructure providers at the bottom and middle of the enterprise stack. In 2024, given its expected impact, AI is gaining importance and momentum at the edge. IoT companies provide the layer that connects the physical and digital worlds, enabling computing at the edge with processing and sensing capabilities. These layers are essential for AI to deliver on its promise of improving efficiency, controlling costs, and mitigating labor shortages. For investors, we see the potential for edge AI-driven applications and consumer use cases as a new AI opportunity set.
Related ETFs
SNSR – Global X IoT ETF
AIQ – Global X Artificial Intelligence & Technology ETF
Click on the fund names above to see current performance and holdings. Holdings are subject to change. Current and future holdings involve risks.
