In the past, it was often the development of chips that led and created user demand. However, as hardware becomes progressively more limited in its ability to scale for running Artificial Intelligence (AI) models and the differences between generations of chips shrink, we are entering a new era - in the age of AI PCs, AI smartphones, and the AI Internet of Things (AIoT) - in which it will be the operating system (OS) and apps that will determine the market's Winner.
According to analysts, three companies are likely to be the clear winners in the race for AI PC, AI phone and AI Internet of Things (AIoT) OS platforms.
The market has high expectations for AI PCs and mobile operating systems and their apps. Microsoft's Copilot could be a pioneer in the AI PC space in terms of insights into user needs. As a result, numerous hardware vendors are currently engaged in large-scale collaborations with Microsoft aimed at accurately grasping the true needs of consumers and driving product success by fostering habits of continuous use and dependence on AI features.
Compared to the PC market, Google has more room to play in the mobile connectivity market. In the mobile operating system space, Google is getting a lot of attention. Currently, the market share of Android and iOS is roughly 7:3. Google also has its own large-scale language model (LLM), recently renamed "Gemini," which will add new leverage to Google's competition in the mobile space. Companies are trying to figure out how to improve the functionality of their smartphones with less hardware. Almost all handset makers claim that their products can support running AI models with 7 billion parameters, while MediaTek has publicly stated that its chips are capable of handling models with up to 33 billion parameters.
All companies are trying to integrate larger-scale AI models within the limited physical space of a phone or PC, with the aim of making AI calculations more precise and able to perform more intelligent tasks," the analysts noted. For example, Samsung recently demonstrated that its smartphones are capable of performing multilingual translations in real-time and offline. However, if consumers are more concerned about the privacy and security of their data and don't want it to be transferred to the cloud, then such tasks can be accomplished on a PC as well."
It is predicted that the popularity of edge AI computing technology will be seen first on PCs and cell phones, followed closely by the AI Internet of Things (AIoT), which has been in development for many years but is still set to expand its impact even further with advances in edge AI. She believes that Arm will be a major leader in this market. not only does Arm have a significant presence in the PC and smartphone chip markets, but its market influence will further increase as AIoT grows.
According to Fortune Business Insights, the global IoT AI market size is projected to grow from USD 35.65 billion in 2023 to USD 253.86 billion by 2030, at a CAGR of 32.4%.AIoT is widely used in transportation, smart homes, smart manufacturing, medical devices, and other areas.
Analysts mentioned that while discussing various application scenarios, manufacturers of personal computers or smartphones tend to highlight the addition of neural network processing units (NPUs) in their products to enhance the overall performance of the device. However, performance enhancement doesn't just rely on CPUs and NPUs, but also requires GPUs and memory to work in tandem, and even communication chips need to be involved to find the optimal mode of operation during operation.
For example, when performing image processing, the GPU needs to take on more computational tasks, while when pursuing the highest quality of network transmission, it may be necessary to allocate more computational resources to the CPU and WiFi chip. For different tasks, it is important to decide wisely which components should receive more resource allocation. Then, AI can perform intelligent computations to find the optimal resource allocation scheme. In addition, on edge devices, AI needs to find the best combination of computing resources and effectively optimize the energy consumption of the device.
