Under global attention, generative AI is ushering in a new era. The rich applications and scenarios it introduces not only promote the deep integration of multiple cutting-edge technologies, but also bring huge commercial value to various industries. The McKinsey report shows that globally, generative AI technology will achieve an overall economic growth of $2.6 trillion to $4.4 trillion annually.
Among them, the launch and popularization of large models have become the most eye-catching symbol of this wave of generative AI. With the popularization of 5G, large models are accelerating from the cloud to the edge and terminal sides. Billions of terminals, including smartphones, PCs, cars, XR, and the Internet of Things, are revitalizing with the support of AI and bringing abundant market potential.
As a industry leader in the global mobile communication field that has been deeply rooted for many years, Qualcomm has become the most active player in promoting terminal side AI in this round.
"Terminal side AI will become the key to promoting the scale expansion of generative AI." At the 2024 World Artificial Intelligence Conference held on July 4, Meng Pu, Chairman of Qualcomm China, pointed out that over 15 years of AI research and development experience have enabled Qualcomm to continuously lead the innovation of terminal side AI, provide the best hardware and software for terminal devices, and have excellent heterogeneous computing capabilities.
As of now, Qualcomm Snapdragon has become the core platform for nearly 3 billion smartphones, laptops, XR devices, automobiles, wearable devices, and other terminals worldwide. "This gives Qualcomm a unique advantage in deploying and promoting terminal side AI," Meng Pu said. Qualcomm has also established partnerships with multiple companies to provide support for developers and further promote AI application innovation in various industries.
Accelerate the scale expansion of terminal side AI
In the past year, the development of AI big models represented by ChatGPT has caught the attention, and with its derivative technological changes, the computing focus of AI is shifting from the cloud to the terminal. From smartphones, personal computers (PCs), cars to various network edge terminals, the sinking computing power enables these terminals to also perform AI accelerated computing. At the same time, AI can run locally and interact with the cloud based on user needs, making human-computer interaction more natural, real-time, personalized, and ensuring privacy.
In addition, Akash Palkhiwala, Chief Financial Officer and Chief Operating Officer of Qualcomm, pointed out that running generative AI use cases in the cloud requires very expensive cloud resources, while the computing power on the terminal side is idle and needs to be utilized. Therefore, the cost advantage makes it a very suitable choice to use the computing power on the terminal side to drive generative AI use cases.
In Qualcomm's view, smart terminals make AI an ubiquitous personal assistant, driving the integration of terminals and the cloud, bringing new ways of interaction to smartphones, powerful AI capabilities to the next generation of PCs, and turning cars into a new computing space. The combined efforts of various scenarios are driving the smart terminal market to usher in new growth momentum.
Faced with this trend, Qualcomm is promoting terminals represented by smartphones, PCs, and automobiles to become important carriers for the scale expansion of generative AI. At this year's World Artificial Intelligence Conference, a reporter from China Business Daily saw that Qualcomm brought AI terminal achievements covering smartphones, PCs, automobiles, and the Internet of Things at the conference, demonstrating its efforts and practices in empowering industry cooperation and innovation with numerous ecological partners through terminal side AI.
In terms of smartphones, Qualcomm has launched the third-generation Snapdragon 8 mobile platform, which can support generative AI models with up to 10 billion parameters. At present, more than 20 flagship mobile phone products equipped with this platform have been released, including Xiaomi, Honor, OnePlus, OPPO, vivo, Samsung, etc., supporting rich generative AI use cases.
In terms of AI PC, Qualcomm has successively released Snapdragon X Elite and Snapdragon X Plus platforms, bringing a new AI experience to PCs with NPUs with up to 45TOPS computing power. As of now, Qualcomm has partnered with Microsoft and numerous ecosystem partners to launch over 20 of the first batch of Windows 11 AI PCs exclusively supported by the Snapdragon X series.
In the automotive field, currently based on the Snapdragon 8295 cockpit platform, multiple manufacturers including Ideal, Xiaopeng, Jiyue, etc. have released their car end large model functions, making "AI on the car" a reality. At the end of May this year, Qualcomm also held an ecological conference in China with the theme of automobiles, presenting more than 60 innovative technologies and over 185 product demonstrations with numerous industry partners, achieving "smarter cars" driving on the "smarter road".
More favorable, according to IDC's forecast, the shipment volume of China's new generation AI smartphones will reach 150 million units in 2027, with a market share of over 50%. Meanwhile, in terms of PCs, consulting firms predict that the penetration rate of AI PCs will increase from 2% in 2024 to 65% in 2028. In addition, with the development of multimodal technology, generative AI is expected to open up new application prospects for fields such as intelligent cabins and autonomous driving.
"If 20% of the generative AI workload is transferred to the terminal side, it is expected to save $16 billion in computing resource costs by 2028. Meng Pu said that the close integration of this terminal with the cloud will be the key to promoting the scale expansion of generative AI and accelerating digital transformation, which is exactly what Qualcomm is committed to achieving.".
Connection advantages help AI end cloud collaboration
From cloud to end, efficient collaboration of AI requires "connection" as the foundation, which makes wireless connection technology represented by 5G a key foundation for the popularization of terminal side AI. As a leading player in the global wireless communication technology field, Qualcomm's accumulation in the 5G field gives it an advantage before the new wave of AI.
Akash pointed out that both computation and AI heavily rely on connectivity. Especially in the hybrid AI of cloud, edge cloud, and terminal side collaboration, extremely high requirements are placed on the connection, which must have extremely low latency, great bandwidth, and connection reliability. "Therefore, we believe that connectivity and computing are inseparable. Therefore, when determining which companies can build generative AI terminals, connectivity technology is one of the cores supporting excellent generative AI." Akash stated that Qualcomm has a unique advantage in this regard.
This advantage is more evident in the Chinese market, which has been in commercial use for 5 years and is leading the world in scale. According to data from the Ministry of Industry and Information Technology, as of the end of April 2024, China has built a total of 3.748 million 5G base stations, with over 26 5G base stations per 10000 people. Over the past five years, it has directly driven a total economic output of about 5.6 trillion yuan and indirectly driven a total output of about 14 trillion yuan. 5G has become the core technology base for promoting economic growth, innovative applications, and industrial upgrading.
Qualcomm is a witness and important participant in China's current round of 5G rise. Even before the commercial licensing of 5G in China, in January 2018, Qualcomm and its Chinese partners jointly launched the "5G Navigation Plan" to help Chinese mobile phone manufacturers seize the opportunities of 5G and continuously explore domestic and international markets, achieving good performance. As of now, more than half of Qualcomm's flagship mobile platforms have been launched globally by Chinese mobile phone manufacturers.
At present, the first phase of 5G technology evolution has been completed, and it has now entered the stage of 5G-A (5G Advanced), which will support more enhanced features, such as RedCap for lightweight IoT terminals, enhanced industrial IoT, non ground networks, etc., promoting 5G to enter a wider range of industries.
"The evolution of 5G technology has reached its midpoint, and 5G-A is also accelerating its implementation and supporting more extended features, which will enable 5G to enter a wider range of industries and applications." During the recently held MWC Shanghai conference, Meng Pu stated that as an important driving force for 5G research and development, commercialization, and scaling up, Qualcomm is collaborating with Chinese industry partners to promote the evolution of 5G-A, enabling 5G to enter a wider range of industries and applications, and laying a technological foundation for 6G.
The reporter noticed that during the MWC Barcelona held earlier this year, Qualcomm launched a new 5G modem and RF system - Snapdragon X80. This platform integrates dedicated 5G AI processors and 5G Advanced ready architecture, achieving multiple global innovations including the first support for 6-way reception for smartphones and the first downlink six carrier aggregation, committed to achieving a new level of 5G performance and experience.
In June of this year, Qualcomm teamed up with China Unicom to complete the first on-site verification of 5G-A high and low frequency NR-CA new technology, and launched the first domestic 5G-A high-frequency 10G testing platform based on Xiaomi 14 Pro and applied in the XR field with industry partners such as China Mobile Terminal Company and Xiaomi, laying a solid foundation for accelerating the maturity and deployment of the 5G-A industry.
Meng Pu pointed out that the collaborative development of 5G-A and AI is helping to unleash each other's greater technological potential, giving rise to revolutionary technological progress and application innovation, accelerating the development of new productive forces and the digital transformation of various industries. Qualcomm will continue to work with more industry partners to promote the two-way empowerment of 5G-A and AI, and open up new value spaces for the development of digital intelligence.
Joining hands with ecology to promote ubiquitous intelligent computing
In addition to showcasing multiple terminal products, Qualcomm also shared numerous innovative achievements in the field of AI and new thoughts on the development of the artificial intelligence industry through multiple forums at the World Artificial Intelligence Conference. In multiple cutting-edge technology demonstrations, Qualcomm has demonstrated its technological, product, and ecological leadership in the field of terminal side AI from multiple perspectives, including basic AI technology research, leading product applications, and future industry empowerment vision.
At the forum on empowering new industrialization with artificial intelligence, Xu Gui, the R&D leader of Qualcomm China, introduced a case of AI combining edge side terminals to further accelerate enterprise innovation and digital transformation. Taking the use case of robots in industrial scenarios as an example, Xu Xi stated that current robots mainly use computer vision or deep learning to complete understanding and functional execution. But with the breakthrough development of generative AI, future robots can understand the requirements we propose, coupled with the excellent positioning, navigation, and robotic arm control achieved in the field of robotics, which can create very intelligent applications.
"By combining 5G and AI, robots can achieve network connectivity, empower production processes such as operation, maintenance, and quality inspection, and create a more intelligent and modern industrial environment." Xu Jian said that technologies such as heterogeneous computing and quantitative compression can help cloud based large models land on terminals, thereby empowering a wide range of industrial applications. The combination of 5G and AI will help create a more intelligent industrial environment.
Meanwhile, at the conference venue, based on the powerful AI capabilities of the third-generation Snapdragon 8 mobile platform, Qualcomm also brought the first demonstration of Large Language and Visual Assistant (LLaVa) technology running on Android smartphones. According to Qualcomm staff, the LLaVa multimodal large model can empower many use cases, such as identifying and discussing complex visual patterns, objects, and scenes, and enhance privacy, reliability, personalization, and cost advantages.
In order to promote the scale expansion of generative AI and provide contextualized, customized, and personalized experiences, Qualcomm has also implemented the LoRA (Low Rank Adaptation) model on Android smartphones. It is reported that reducing the number of trainable parameters of the model through the LoRA adapter can reduce training costs, improve the accuracy of the model on specific tasks, and support more efficient, scalable, and customized AI applications on the terminal side. Qualcomm stated that LoRA and multimodal AI will be key technological examples for the future development of terminal side generative AI.
Various innovations also indicate that Qualcomm is moving towards a new transformation. "Innovation is Qualcomm's DNA." Akash pointed out that Qualcomm is in an industry that never stops, and this industry can always undergo tremendous changes, and innovation allows Qualcomm to change together with the industry.
Meng Pu stated that Qualcomm first proposed "5G+AI empowering thousands of industries" in 2021. Now, with the launch of large models and the widespread application of AI in the cloud, edge, and terminal sides, coupled with the rapid development of 5G, the industry has seen more clearly the rich development opportunities behind these technologies.
"Qualcomm has always adhered to an innovation driven development philosophy, continuously promoting AI research and development, hardware development, and software optimization, and is committed to providing comprehensive support to developers. Our goal is to penetrate numerous industries around the world, promote technological progress and application innovation, and make intelligent computing ubiquitous." Meng Pu further emphasized.