The rise of generative AI has already had a significant impact, but what would it mean for the electronics and wider industries if this technology were to be deployed to edge devices?
The next major innovation in Apple’s “One More Thing” could very well be edge AI. While it once seemed that the tech giant’s next big leap would be in automotive cars, Bloomberg tech journalist Mark Gurman reported in early February that Apple had halted its decade-long car project. Instead, the company is now refocusing its resources entirely on generative AI initiatives.
At GTC 2023, Nvidia CEO Jensen Huang compared ChatGPT to the AI equivalent of the revolutionary “iPhone moment.” Then, in an intriguing parallel, GTC 2024 witnessed Nvidia’s announcement that Apple would be using Nvidia’s Omniverse platform to enhance the 3D digital experiences of Apple’s Vision Pro (headset). Launched by Nvidia in 2019, Omniverse serves as a platform for developing and deploying physically accurate industrial digital applications. Over the past five years, Omniverse has evolved to incorporate new features that cater to a wide range of industries, including automotive, manufacturing, media, film, architecture, energy, scientific computing, and simulation.
Omniverse is not the only instance of Apple venturing beyond its proprietary technologies for AI services. According to Bloomberg’s official X account, Apple has also recently reached an agreement with OpenAI to integrate the company’s AI technology in the iPhone, further enhancing Apple’s AI capabilities and offerings.
Apple’s recent partnership with OpenAI marks a significant departure from its historical emphasis on internally developed technologies. This strategic shift underscores the company’s commitment to acquiring the most cutting-edge AI capabilities, regardless of their source. The integration of OpenAI’s AI technology into the iPhone highlights the profound impact of generative AI on the consumer market and positions Apple at the forefront of AI-driven innovations.
Generative AI’s Potential to Revolutionize the Edge
Under the guidance of leading tech giants, AI technologies are transitioning from the cloud to the edge, expanding to PCs and smartphones, as well as smart Internet of Things (IoT) devices like smart speakers and in-car assistants.
Media reports suggest that Apple is prioritizing an edge-focused AI strategy, with AI models operating locally on devices rather than in the cloud infrastructure. Part of this focus includes Apple’s aim to incorporate Large Language Models (LLMs) into its devices. However, according to The Guardian, while Apple initially developed its own framework called Ajax for LLMs, it has subsequently adopted OpenAI’s technology.
Although the precise motivations behind Apple’s shift to edge AI remain unclear, it is likely driven by the increasing consumer demand for AI-powered solutions. Apple’s aggressive integration of the latest AI technologies into its products, regardless of their origin, signals the emergence of the “AI plus hardware” era.
In addition to Apple, other major players like Google, Qualcomm, and Nvidia, have all been actively exploring the integration of AI with hardware. For instance, Google released its PaLM 2 model, featuring Gecko, which is capable of running on mobile devices. Meanwhile, Qualcomm demonstrated the world’s first terminal-side Stable Diffusion model for AI image generation running on Android devices, generating a 512×512-pixel image in under 15 seconds on a smartphone. As these tech giants pave the way, generative AI applications are beginning to extend beyond the cloud, infiltrating devices and helping to revolutionize the value of hardware.
Redefining Growth in the Electronics Industry
In 2024, Statista reported that 95.9% of internet users worldwide preferred smartphones as their primary device for accessing the web, compared to just 62.2% who relied on laptops or desktops. This emphasizes the importance of integrating AI tools into smartphones to enhance user engagement, thereby fueling market growth.
Traditional hardware is showing signs of stagnation, with new smartphones and PCs entering an increasingly saturated market. Even newer innovations, such as smart speakers and headphones, are experiencing a slowdown in growth.
Global PC sales have steadily decreased since their peak of 365 million shipments in 2011, apart from a brief rebound in 2020 and 2021. In 2023 alone, PC shipments saw a 15% year-over-year decrease, totaling 242 million – roughly a 33% drop from the 2011 high.
Smartphone sales follow a similar trend, peaking at 1.47 billion units in 2016 but declining by 20.61% to 1.167 billion units by 2023. Likewise, despite being a key product for smart homes, smart speakers have seen only lukewarm market growth, largely due to subpar user experiences.
Edge AI has the potential to alter this downward trend. As generative AI use cases expand across edge, cloud, and hybrid environments, AI is set to become a standard feature for PCs, driving innovation and potentially market growth. According to Statista’s forecast, the market is expected to grow at a Compound Annual Growth Rate (CAGR) of 3.04% from 2024 to 2028, reaching a market volume of $247.10 billion by 2028. This growth is largely driven by the introduction of AI chips, with Statista predicting that AI-capable PC shipments will surge from 48 million in 2024 to 205 million by 2028.
The integration of new AI features may also help drive up the Average Selling Price (ASP) for advanced semiconductor components like Central Processing Units (CPUs), Graphics Processing Units (GPUs), and memory as the devices begin to offer smarter functionality at the hardware level.
Generative AI’s Impact on Edge AI
The rapid growth of generative AI models, such as ChatGPT, is fueling a constant need for AI-generated content (AIGC) and driving advancements in edge computing hardware. Edge AI is capitalizing on this opportunity to enhance smart hardware, driving significant changes in product design and expediting the transition towards intelligent, companion-like devices, including smartphones, home assistants, service robots, and automotive electronics.
The Future Belongs to AI at the Edge
The AI industry relies heavily on computing power for its development. As companies continue to release AI products and achieve breakthroughs with large models, the computing power required for training and inference is projected to grow exponentially.
The core concept of computing power lies in semiconductor chips, positioning the AI chip sector as a crucial battleground. AI chips are primarily categorized into two types, Cloud AI chips and Edge/Terminal AI chips, which can be further divided into training and inference chips.
Cloud AI chips are optimized to manage resource-intensive tasks such as model training, inference, data analysis, and tasks requiring high bandwidth. Edge and terminal AI chips are designed to address inference tasks, including data collection, environmental sensing, human-machine interaction, and localized decision-making.
Generative AI models like ChatGPT, depend heavily on cloud-based computing resources, necessitating significant resources for both model training and inference. However, this approach is neither sustainable nor cost-effective in terms of computing capacity and energy consumption.
Industry experts increasingly highlight the need to enable generative AI functionalities on edge devices. In addition to enhancing efficiency, edge-based AI enhances data privacy and security through local data processing, safeguarding data sovereignty and promoting environmental sustainability. This underscores the future trajectory of a multi-layer computer network in which sophisticated AI training tasks are centralized in the cloud. In contrast, real-time inference and localized tasks are decentralized to the edge.
Conclusion
The rise of AI, whether in the cloud or at the edge, is undeniably creating demand. Whether at the product level or within the realm of underlying technologies, AI has the ability to transform the functionality of a solution completely. This shift can potentially push industries and products out of stagnation and into a new era of innovation.
While the exact form of this progress is uncertain, one thing that is certain is the downsizing of generative AI into edge devices. This advancement will help to introduce a multitude of unprecedented applications, revolutionizing our lifestyle, work, and entertainment experiences.
Source: Mouser Electronics
