Science and Technology Innovation Board Daily, February 29th (Reporter Huang Xinyi) Baidu has released its financial reports for the fourth quarter and full year of 2023. According to the financial report, Baidu's total revenue in 2023 was 134.598 billion yuan, a year-on-year increase of 9%, and its net profit attributable to Baidu was 28.7 billion yuan, a year-on-year increase of 39%. In the fourth quarter, the revenue was 34.951 billion yuan, a year-on-year increase of 6%, and the net profit was 7.755 billion yuan, a year-on-year increase of 44%.
Among them, Baidu's core revenue in 2023 was 103.465 billion yuan, with a net profit of 27.4 billion yuan, a year-on-year increase of 38%. The monthly active users of Baidu App were 667 million in December. The service volume of Carrot Run exceeded 5 million orders in 2023, with a year-on-year increase of 49% to 839000 orders in the fourth quarter. The proportion of all autonomous driving orders in Wuhan reached 45%.
In the conference call after the release of the financial report, Robin Lee, chairman and CEO of Baidu AI Cloud, revealed that the total revenue of Baidu Smart Cloud in the fourth quarter was 8.4 billion yuan, of which the big model brought about 660 million yuan of incremental revenue for cloud business. He predicts that by 2024, this incremental revenue is expected to reach billions of yuan, mainly from the sales of advertising business and artificial intelligence cloud services.
Recently, Baidu has launched two MaaS products, Qianfan AppBuilder and Qianfan ModelBuilder. Robin Lee said that so far, Qianfan ModelBuilder has refined 10000 models. The daily adjustment usage of the Wenxin model has exceeded 50 million times, with a quarter on quarter increase of 190%. In December, about 26000 companies called the Wenxin Big Model, with a quarter on quarter growth of 150%. At present, companies such as Samsung, Honor, and Autohome have reached cooperation with Baidu.
The high usage cost of large models is one of the difficulties affecting their larger scale implementation. Robin Lee revealed that the reasoning cost of Wenxin model has been reduced to 1% of the version in March last year. Robin Lee also responded to the impact of AI chip supply on Baidu. "In the short term, this has little impact on our model development, product transformation or commercialization. Baidu has upgraded Wenxin model to 4.0. Model reasoning requires chips with lower performance. Our AI chip reserves and chips available in the market are enough to support the provision of AI native applications for customers, so that Baidu can continue to improve its revenue in the next year or two." Robin Lee said.
Robin Lee also said that in the long run, Baidu may not be able to obtain the most advanced GPU chips, "however, through the most efficient and self-developed software stack, the user experience can be unaffected. In the application layer, model layer and framework layer, there is room for innovation in end-to-end and self-developed AI architectures."
The explosive popularity of Sora, a large-scale model for cultural and educational videos, has attracted great attention in the industry. In response, Robin Lee said that the integration of text, audio and video is an important direction for the future development of the big model, and Baidu has invested in this field. Robin Lee believes that in the field of visual big model, a significant application and potential market is automatic driving, which can transform images and videos captured on the road into specific tasks, thus producing more intelligent, more adaptable and safer automatic driving technology.
Hu Qimu, Deputy Director of the Strategic Emerging Industries Research Center at Tsinghua University, believes that Baidu's revenue growth rate in 2023 was 39% year-on-year, and its net profit growth rate was 44% year-on-year. Among them, the performance growth brought by the AI business was the main driving force,
"Looking back at Baidu's previous development history, although it has laid out almost all tracks, it has not made any outstanding brands. What can be remembered by everyone is still the search engine. However, this core capability has become a weapon in the era of artificial intelligence. It should be said that Baidu has the best Chinese training set, thanks to its underlying capabilities of the search engine."
Previously, there were reports that Baidu planned to sell its 53% stake in iQiyi, but iQiyi denied it. Hu Qimu stated that based on the development of AIGC in the past two years, Baidu is only likely to increase its investment in the audiovisual business in the future, which is an important track for its technological commercialization. "Text, graphics, video, and sound are currently the best areas for monetization, and the demand scale is also large enough. The recently released Sora has actually sent a signal."
In terms of the integration of big models and industries, Hu Qimu pointed out that AIGC is still in its early stages. "Like the deep integration of AI and manufacturing, the existing Internet companies have not yet found a particularly good breakthrough. AI technology has been applied in the fields of administrative office and supply chain, and the scene continues to increase, but the core processing and manufacturing links are still not making enough progress."
Hu Qimu believes that Baidu needs to cooperate with manufacturing companies to achieve two-way cooperation. "Baidu has done a good job in the field of smart transportation, but it still needs to enter more complex processing and manufacturing fields. Other large factories have basically cooperated with traditional industries in landing cases."