According to news on November 22, at yesterday’s Baidu 2023 third quarter performance briefing, Baidu founder, chairman and CEO Robin Li said that the U.S. chip export restrictions to China will have a limited impact on Baidu in the short term. Baidu has a large reserve of artificial intelligence chips and can continuously update Baidu’s AI large model "Wen Xinyiyan" in the next 1 to 2 years.
Robin Li emphasized: "Wenxin large models do not require powerful chips, and our chip reserves and other alternatives will be enough to support a large number of AI local applications for end users."
"In the long run, the difficulty in obtaining the most advanced chips will inevitably affect the pace of China's AI development. Baidu is actively looking for alternatives, although these options are not as advanced as American chips." Robin Li further said.
Shen Dou, president of Baidu’s Intelligent Cloud Business Group, also mentioned that some of Baidu’s customers prefer to train their own models, but chip export restrictions will reduce this activity.
According to the third quarter financial report released by Baidu, Baidu's revenue for the quarter reached 34.447 billion yuan, a year-on-year increase of 6%; the net profit (non-GAAP) attributable to Baidu reached 7.3 billion yuan, a year-on-year increase of 23%. Both revenue and net profit exceeded market expectations.
Specifically, Baidu Core (BaiduCore, a combination of search services and transaction services) revenue in the third quarter was 26.6 billion yuan, and net profit (non-GAAP) attributable to Baidu Core reached 7 billion yuan, a year-on-year increase of 21%; Baidu Core online marketing revenue increased by 5% year-on-year to 19.7 billion yuan, which is basically in line with JPMorgan Chase’s expectations. JPMorgan Chase previously lowered its revenue forecast for this performance to 19.7 billion yuan. Non-online marketing business revenue was 6.9 billion yuan, a year-on-year increase of 6%. This part of the business includes revenue from innovative businesses including Baidu Cloud, Xiaodu and autonomous driving businesses. Among them, the cumulative service volume of Luobo Kuaipao has increased to 4.1 million. It is reported that Xiaodu continued to rank first in China’s smart screens and smart speakers in terms of shipments in the second quarter.
Baidu executives said on the conference call that the traditional cloud business is slowing down, and generative AI and large-scale language models are stepping in and reshaping the competitive landscape of the cloud business industry in the past, and are expected to achieve considerable profit margins in the long term. Some products among customers are still in the experimental stage, and there have been successful experiences in this area overseas, so we can see that new technologies are increasing Baidu's competitive advantage.
Baidu Executive Vice President Shen Dou pointed out that Baidu Cloud has achieved operating profits that are not in accordance with U.S. GAAP in the past few quarters. Due to weak market demand for intelligent transportation services, cloud business supply also declined in the third quarter. If the intelligent transportation part is excluded, the growth of cloud business is good. As the Group continues to explore the growth opportunities brought by generative AI and large-scale language model technology, it will leverage its leading position in related fields to continue to attract new customers and drive existing users to increase spending on Baidu AI cloud services. It is believed that the cloud business will also resume positive growth in the fourth quarter.
In October this year, Baidu launched Wenxin 4.0. Users can experience Wenxin 4.0 through Wenxin Yiyan, and enterprises can call Wenxin 4.0's API through Baidu Smart Cloud Qianfan large model platform. According to the financial report, Wenxinyiyan currently has 70 million users and covers 4,300 scenarios.
Robin Li pointed out in the financial report conference call that since November 1, Baidu has officially launched Wenxinyiyan Professional Edition and started charging. Version 4.0 has widened the gap between Baidu and other models and became the first to charge. "As for opportunities in commercial realization, we see many opportunities related to the native application of artificial intelligence technology, some of which are Baidu's own There are also some developed by customers using Baidu technology. In our own part, there are many opportunities in new search and adjusted advertising businesses. As our search service will be able to realize natural language conversations and multi-round conversations, the potential for commercial monetization will be greater in the long term, and the inference service will become an important contribution to the company's revenue."
In addition, Robin Li also said that considering factors such as the shortage of high-performance chips and the high market demand for data and AI talents, he believes that the industry will enter a stage of consolidation, and it is expected that only a few basic large models will be left in the market.
According to reports, Baidu is currently using the Wenxin model to reconstruct its advertising system, including services such as generative creativity and generative targeting. These initiatives are expected to bring hundreds of millions of yuan in incremental revenue in the fourth quarter. "Advertisers who adopted new AI features achieved an average high-single-digit conversion rate increase in the third quarter. Taking IT professional education company Dane Education as an example, after using the new features, the conversion rate increased by 23.3% and the ROI (return on investment) increased by 22.7%."
Baidu Chief Financial Officer Luo Rong said that in the future, Baidu will continue to prioritize investment in artificial intelligence, especially generative AI and basic models, with an unswerving focus on efficiency and strategic resource allocation.