At the Arm Tech Symposia 2023 event held in Taiwan, James McNiven, Arm's Vice President of Product Management, emphasized in an interview the company's deep partnerships with market players. He also emphasized working with the Taiwanese supply chain to drive the growth of the Arm application product ecosystem, and expressed confidence in the future development trends of AI PC and AI device applications.
James McNiven stated that the Arm ecosystem is already quite robust, including achieving 2500 billion Arm architecture licensed chip shipments ahead of schedule at the end of last year. If calculated based on the total global population, each person would have an average of 30 sets of chips, and if the total population accumulated throughout history is added up, each person would even have two sets, emphasizing that Arm architecture application products have spread to every corner of daily life.
Currently, Arm operates by providing computing platform technology resources and continues to grow alongside its market partners. Therefore, it will continue to strengthen its computing platform resources in response to market trends and demands. In particular, with the recent boom in generative AI and device-side AI computing, Arm has begun to provide corresponding solutions through its Total Compute platform and further apply computing performance to the needs of the IoT, terminal devices, data centers, and the Internet of Vehicles market.
Focusing on the design direction of integrating artificial intelligence into all types of devices
Regarding the development of AI applications, James McNiven emphasized that Arm has been preparing for this for a long time and anticipates more comprehensive application development opportunities in the future. However, compared to other industry players, Arm is primarily focused on integrating AI into all device categories, rather than focusing solely on specific market segments. Therefore, the solutions it provides will focus more on computing architecture, performance, and software integration, rather than designing solutions for specific product categories.
In Arm's view, rather than proposing reference designs for different product categories, it is more important to continuously improve the basic computing architecture and performance. For example, in addition to continuously strengthening the computing power of the CPU, GPU, and NPU, the communication connection between different computing components is also very important. Therefore, when Arm announced the launch of the Armv9 instruction set last year, it also announced the addition of the Scalable Matrix Extension (SME) architecture to improve processor computing efficiency. It also further enhanced the data transmission speed within the processor, allowing for a significant increase in computing performance.
At the same time, Arm also simplifies the design to enable partners to quickly implement solutions. This, combined with commonly available large-scale natural language models (LLMs), enables rapid implementation of various generative AI computing applications. James McNiven believes that rather than providing a single reference design that allows companies to mass-copy and apply it, it is more important to reduce the design complexity for companies, allowing them to focus on connecting different AI technology applications, thereby accelerating the entry of more differentiated products into the market.
Hopefully, Windows on Arm will be a success.
Regarding the current development of the Windows on Arm market, James McNiven stated that he would not comment on Qualcomm's current efforts to build products using its own processor architecture and the Arm instruction set, or on Apple's renewed push for Windows on Arm market development with its Apple Silicon processor. However, he emphasized that Arm's position is to provide solutions that deliver improved computing performance, and that he is pleased to see Windows on Arm designs succeed in the market.
In addition, James McNiven does not believe that Windows on Arm will replace the existing Windows market dominated by Intel's x86 architecture, but will provide the market with more diverse choices. He also stated that he will continue to provide better solutions to help partners expand market development opportunities.
Future computing models will tend towards hybrid architectures
Regarding the future direction of computing, James McNiven believes that hybrid computing models will become mainstream. However, the question of what computing content will reside on the device and what will rely on cloud-based collaborative computing depends largely on the actual application scenario, and there is clearly no definitive answer. However, it is foreseeable that the future increase in on-device computing will be significantly related to privacy concerns, such as concerns about personal data being stored in the cloud and used by other applications and services. This is only in specific cases to accelerate computing response times.
