Baidu, Inc. 2023Q3 Earnings Call Transcript: Baidu Focuses on AI Revenue Growth

1. Baidu Core delivered strong Q3 results despite challenges in the macro climate, with a focus on generative AI and foundation models like ERNIE and ERNIE Bot to transform product usage and business operations for users and customers. Baidu's mobile ecosystem continues to exhibit steady growth, while online marketing revenue has increased by 5% YoY.

2. Although AI cloud revenues declined by 2%, Baidu remains optimistic about positive growth in Q4 due to momentum in generative AI-related businesses. Intelligent driving targets remain unchanged, with a concentration of resources on pivotal regions such as Wuhan, which saw an increase in fully driverless orders and operational expansion for Apollo Go.

3. Baidu's ERNIE 4 model has received positive feedback since its release in mid-October, and the company now charges enterprises and users for its use. Baidu anticipates that inferencing will become a major source of revenue in the long-term and sees significant opportunities in AI native applications, including its own products and customer applications powered by advanced LLMs like ERNIE.

4. Baidu's unique four-layer AI infrastructure and capability to develop GPU networks for clusters give them a competitive edge over their peers. The company plans to leverage this advantage to reduce costs in model training and inference on its cloud, giving them the flexibility to offer more compelling prices to customers. Baidu intends to continue taking an application-driven approach to improve ERNIE, letting AI native apps direct innovation in the right direction.

5. Baidu's investment in generative AI and large language models is paying off as they see growth in revenue generated from AI-powered businesses. The restrictions on chip export from the U.S. have limited impact on Baidu in the near-term, but they are proactively seeking alternatives and innovating to improve efficiency and mitigate challenges. Baidu's end-to-end optimization approach allows for more efficient and cost-effective training and faster and cheaper inference, positioning them as the best option for developing AI native apps based on their foundation model.