Meet the analyst: Operating system in AI era will be king, says DIGITIMES Research

Judy Lin, DIGITIMES Asia, Taipei 0


In the past, it was the chip that created the user's needs, but when hardware has little room to expand for running Artificial Intelligence (AI) models while the difference between each generation of chips gets smaller and smaller, it's the Operating System (OS) and applications that decide the winner in the AI PC, AI smartphone and AIoT era, said DIGITIMES Research analyst and project manager Joyce Chen.

Three obvious winners are likely to emerge from the AI PC, AI phone, and AIoT OS platform in the eyes of Joyce Chen.

AI PCs and mobile operating systems and applications are highly anticipated, and Microsoft's Copilot may lead the way for AI PCs to gain insights into users' needs; therefore, hardware vendors are now collaborating with Microsoft on a large scale in the hope of identifying consumers' real needs and achieving product success after developing a habit of using and relying on AI functionality.

"Microsoft still has 70% of the global PC operating system market, while Apple has 30%, so Microsoft's plan to build Copilot into Win 11, 12, and even CES, where we've seen a Copilot button on laptops, is a good indication that they're placing their bets on Copilot to help consumers build up a habit of using AI," said Chen. "But it also depends on how much trust users have for Copilot."

Google has little room to play in the PC market, but has lots to offer in the mobile connectivity market. On the cell phone side, Google is getting the most attention. The market share distribution between Android and iOS is close to 7:3. Google also has its own LLM, which was recently renamed "Gemini".

The way AI models operate on smartphones will be different from AI PCs because smartphones are always connected to the Internet, whether on Wi-Fi or telco services. "If the phone is connected to the Internet and can also run LLM, you can speak directly to command the phone to search for data and organize it, which will be a selling point," said Chen.

Credit: DIGITIMES Research

Credit: DIGITIMES Research

Everyone is trying to figure out how to optimize the functionality of smartphones by minimizing hardware. Almost all cell phone companies say that they can run models with 7 billion parameters, and MediaTek officially claims that its chip can run models with up to 33 billion parameters.

"All companies want to put bigger AI models in the limited physical space of their phones or PCs so that AI can be more and more accurate and do smarter things. Samsung has recently been shown to translate different languages on its smartphone in real-time, offline. But if it's important to the consumer that the data doesn't go to the cloud, the task can also be done on a PC," said Chen.

The first popularization of edge AI computing will be PCs and cell phones, followed by AIoT, which has been around for years. Chen said the dominant player in this market should be Arm. In addition to the PC and smartphone chip market, Arm's popularity will increase due to AIoT.

The global artificial intelligence of things market size is projected to grow from US$35.65 billion in 2023 to US$253.86 billion by 2030, at a CAGR of 32.4%, according to Fortune Business Insights. The AIoT applications are implemented in transportation, smart homes, smart manufacturing, and medical devices.

In terms of use cases and scenarios, Chen said PC or smartphone vendors would emphasize the addition of an NPU in their products to boost performance. Performance improvement cannot only rely on the CPU and NPU, but also requires the GPU and memory, and even the communications chip, to find the optimal mode of operation in the process.

For example, when processing images, the GPU will have to do more work; if it is necessary to achieve the best quality of network transmission, it may be necessary to allocate more computing power to the CPU and WiFi chip. It is necessary to decide which units should be allocated more for different tasks. The AI can then calculate the optimal allocation. There is a need for AI to find an optimal combination of computing in edge devices and optimize power consumption as well.

How Apple has been doing with AI?

Apple Inc., on the other hand, is taking a relatively low-profile approach to promoting AI. Apple's DNA has always been not to publicize a certain technology, but to emphasize the concept of user-centeredness so that all designs are designed to satisfy users. AI will be a tool for Apple to enhance a function that the user wants to do, but it may not be specifically emphasized that it's because of something similar to Copilot that the AI function is being done. That's the opposite of the market mainstream right now.

However, AI can still be seen in Apple's recent products. For example, the VisionPro already uses a lot of image-generating AI. So it's not like Apple doesn't have any generative AI technology at all.

Apple will likely use generative AI to optimize their laptops or phones, such as figuring out the best energy configuration or maintaining the best network transmission quality.

Apple's A17 Pro chip already uses 3nm technology, and it's installed in the iPhone 15 Pro, which already has an offline chat AI app. Bloomberg reports that Apple is actively working on adding a series of new AI features to iOS 18, and integrating generative AI into Siri, making iOS 18 the most significant update in its history. However, Apple didn't emphasize that it's an AI phone.

"This is Apple's approach, they will not be absent from AI PC or AI smartphone market, and their hardware specification is not lagging, their way of marketing it is just different from the mainstream market," said Chen.

Will there be monopoly concerns?

Large Language Model (LLM) is already a very powerful system. If it is combined with an operating system, it can be extended to many devices. However, Joyce Chen emphasized that trust has always been an issue for AI. The demand for generalized AI has yet to be found, and the behavior of consumers using AI still depends on their trust and acceptance of AI. For example, some heavy users use generative AI to make slides for work, but some people don't trust AI and feel that it doesn't matter to them with or without it.

There have been anti-trust concerns for IE browsers set up in every PC in the past. Will it be different for Copilot, especially with Microsoft's close relationship with OpenAI, ChatGPT is also associated with Microsoft.

"It may become the combination of the most powerful operating system and the most powerful generative language model on the earth. The regulatory agencies in various countries are also paying close attention to this trend, but Microsoft with Satya Nadella in charge is not the same old Microsoft either," said Chen.

About the analyst:

Joyce Chen is a DIGITIMES Research analyst and project manager of the semiconductor team. She specializes in artificial intelligence (AI) cloud platform applications and IoT applications.