Reporter Chen Jialan reports from Shenzhen.
After AI has become a hot word in the technology circle, recently, the concept of the term "AI mobile phone" has also become popular. In the past, Samsung repeatedly mentioned the concept of AI mobile phones when it released a new generation of flagship devices, the Galaxy S24 series mobile phones, and then OPPO executives repeatedly shouted that the mobile phone industry will enter the era of "AI mobile phones".
On the first working day of the Year of the Dragon, Chen Mingyong, founder of OPPO, issued an open letter saying that AI mobile phones will become the third stage of the mobile phone industry after feature phones and smartphones. Two days later, on February 20, OPPO held an AI strategy conference and announced the inauguration of the OPPO AI Center.
We officially established the AI Center to integrate the AI resources of various departments and turn it into an entity organization. Liu Zuohu, chief product officer of OPPO, told ** reporters, including China Business Daily, "In terms of investment, we regard AI as the most important strategy in the next era of mobile phones, and there is no upper limit." ”
Starting from the second half of 2023, many mainstream mobile phone manufacturers such as Huawei, Xiaomi, Vivo, Samsung, OPPO, and Honor are vying for AI models. And Qualcomm President and CEO Cristiano Amon said in an interview that 2024 will be the first year of global AI mobile phones, and generative AI is entering mobile phones "very quickly".
What is an "AI phone"?
The concept of AI mobile phones is hot, but at present, the definition of AI mobile phones still seems to need to be clarified. In addition, in the context of the bottleneck of smartphone innovation and growth, can AI mobile phones really become a disruptor of smartphones? When will it stimulate users to form a large-scale replacement demand? These are still worth keeping an eye on.
In fact, smartphones capable of running device-side AI have been around for nearly a decade. AI algorithms have long been seen running on mobile phone ISPs (Image Signal Processors) and NPUs (Neural Processing Units). However, the recent emergence of large language models (LLMs) has prompted the industry to rethink and define what an "AI phone" is.
According to a recent report released by IDC Consulting, it divides the definition of AI mobile phones into two categories, including hardware-enabled AI mobile phones and a new generation of AI mobile phones. Hardware-enabled AI phones refer to smartphones that use accelerators or dedicated processors in addition to the main application processor to run device-side AI at lower power consumption. More recently, such phones have also included a shift to using NPU cores, using the INT-8 data type, with performance up to 30 TOPS, and examples of device-side AI include NLP (Natural Language Processing) and computational photography, which have been on the market for almost a decade.
The newer AI phones fall into the latter category, as defined by IDC"A new generation of AI phones"Refers to a system-on-chip (SoC) that uses a faster and more efficient device-side generative artificial intelligence (GENAI) model that uses a neural processing unit with 30 tera operations per second (TOPS) or higher performance and uses an int-8 data type. These new devices are generating strong interest from consumers and OEMs, making AI a key marketing segment for recent flagship product launches.
In OPPO's view, the new generation of AI mobile phones should have four major capabilities: first, they should be able to efficiently use computing resources to meet the computing needs of generative AI in the AI era; At the same time, it can keenly perceive the real world and understand the complex information of users and the environment; AI phones also need to have strong self-learning capabilities; In addition, AI mobile phones will also have more creative capabilities, providing users with continuous inspiration and knowledge support.
According to IDC, the United Nations mobile phone manufacturers released "AI mobile phones*** Current users expect more advanced technology to liberate users' energy and creativity, and AI large model technology will once again trigger an interactive revolution and drive the evolution of smart phones to AI mobile phones."
Is it expected to drive the wave of replacement?
IDC believes that from 2024 onwards, the new generation of AI mobile phones will grow significantly, driving a new round of replacement, among which flagship models will become an important growth driver in the early stage of the development of the new generation of AI mobile phones.
Zhong Xiaolei, a research analyst at Canalys, told reporters that the end-side large model has requirements for hardware computing power, and the penetration rate of AI mobile phones that can meet the requirements is not high, but they are all newly released new products (which will only be launched in the second half of 2023).
However, there is also an opinion that it will take a long time for this new generation of AI phones, which is seen as a potential disruptive smartphone revolution, to promote large-scale replacement. At present, the integration of mobile phones and AI still faces many challenges such as memory space, cloud computing power cost, data privacy protection, and consumer acceptance.
Guo Tianxiang, senior analyst at IDC China, said that the last wave of consumers who purchased phones at the peak of shipments in the second half of 2020 will enter the replacement cycle from the first half of 2024, but satellite communication and AI models will not become a rigid demand for consumers to promote large-scale replacement in the short term.
Large-scale model technology will propel mobile phones into the AI era. Huo Jinjie, President of IDC China**, the global shipment of a new generation of AI mobile phones will reach 1700 million units, accounting for 15% of overall smartphone shipments. At the same time, it is expected that the shipment of a new generation of AI mobile phones in the Chinese market will be 37 million units in 2024, and by 2027, the shipment of AI mobile phones will reach 1500 million units, accounting for more than 50%. The new generation of AI mobile phones will bring hardware upgrades and cost increases for storage, screens, and imaging devices, which will further increase the average customer value (ASP) of smartphones.
From the cost perspective, the cost of AI on the mobile phone side mainly includes computing costs, inference costs, and large model tool invocation costs, and the last cost depends on the frequency of use, the number of uses, and the pricing of the large model side, and the first two are closely related to computing power and chips.
From the perspective of IC, we have at least 30%-50% of the computing power optimization at the same cost every year, but it is not realistic to reduce the cost in the short term. Chen Yiqiang, deputy general manager of MediaTek's wireless communication division, said that the cost of hardware will decrease with the scale, but in the short term, due to the insufficient scale of the application, there is no way to form an effective business model in AI.
IDC believes that OEMs need to address other hardware requirements beyond the SoC for the next generation of AI handsets before moving to the next stage. One of the biggest variables is not only the premium cost of the SoC, but also the cost of the additional DRAM (Dynamic Random Access Memory) required to support the super-powerful SoC of these devices.
It is the first to drive the demand for flagship phones
For most smartphones, 16GB of RAM is already a lot of capacity, while for the new generation of AI phones, 16GB is considered the minimum requirement. As the cost of DRAM increases, the memory of these phones will generally increase by 1 times, which will far exceed the cost of memory by 2 times in the near future. Most flagship phones tend to come with premium configurations and components, including better screens and cameras, to optimize the use of multimodal GenAI in imaging.
This means that the new generation of AI phones will also come with higher bill-of-materials (BOM) costs, which is both an opportunity and a challenge for OEMs. OEMs can take advantage of innovative technologies and their clear advantages to pass on the increased BOM costs to the end consumer at a higher list price, and take the opportunity to add value to the industry or absorb additional costs between the channel and the OEM.
At present, AI capabilities are expected to be the first breakthrough point for manufacturers to promote high-end.
Zhong Xiaolei analyzed to reporters that end-side AI applications have certain limitations on hardware performance, but we see that some terminal mobile phone manufacturers are also accessing through some cloud large models, so there is not much requirement for mobile phone performance, but large-scale applications will invisibly increase the cost of manufacturers, the larger the scale, the higher the bandwidth cost and inference cost of manufacturers, so manufacturers will give priority to promoting the penetration of AI in the high-end market.
As for when AI mobile phones will be able to promote large-scale replacement demand, Zhong Xiaolei said: "At least after the end-side capabilities have further sinking, it is optimistically estimated that the penetration rate of AI mobile phones will begin to climb from 2025." ”
OPPO is committed to becoming a contributor and popularizer of AI mobile phones, and during the Spring Festival, OPPO has more than 10 models and more than 10 million users to quickly popularize some of the capabilities of AI. Liu Zuohu said that AI mobile phones are not exclusive to high-end mobile phones, they should be inclusive, so that everyone can quickly enjoy the benefits brought by technology, and OPPO will promote the function to the full line of products to promote the popularization of AI mobile phones.
On February 21, the reporter visited several mobile phone offline stores in Shenzhen, and in the face of several consumers who came to buy mobile phones, when asked whether they would pay attention to the functions brought by the mobile phone model, some consumers said: "There is no obvious perception." Another clerk said: "There are also consumers who will ask about the functional selling points of large-scale mobile phones, but the overall attention is not high." ”
When talking about the return on investment of AI, Liu Zuohu pointed out that in the early stage of technology development, everyone will definitely not look at the return first, and whether it is the identification of this general trend is the most important.
Editor: Wu Qing Proofreader: Yan Jingning).