Baidu, Inc. (BIDU) Earnings

Management highlights

### AI Infrastructure & Cloud - Full-stack AI capability is a core differentiated advantage, with proprietary technology across all layers from chips to applications enabling stable supply, end-to-end optimization, improved performance and lower costs for customers. - Self-developed Kunlun Xin AI chips have achieved large-scale commercial deployment in clusters of over 30,000 accelerators, with industry-leading cluster performance and stability. It is optimized for mainstream foundation models including Ernie, DeepSeq v4, GLM 5.1 and Minimax M2.7, with growing adoption across diverse industries. - Qianfan MaaS platform has expanded its model library to include popular third-party models besides Ernie. In March 2026, daily average token consumption from external customers was nearly 7x year-over-year, with rapid revenue scaling. - Baidu AI Cloud added multiple prominent new clients including Unitree, Honor, Opal and Bevo, with existing top clients deepening collaboration. ### Foundation Models - Ernie 5.1 was launched, with improved reasoning, code generation, agentic and deep search capabilities. It ranks first among Chinese models on the LM Arena text and search leaderboards, and is the only Chinese model ranked fourth globally on the global search leaderboard. - Baidu uses an application-driven iteration approach, continuously updating Ernie based on real-world product needs. ### AI Applications - **DoMate**: A 24/7 cross-platform AI productivity agent that autonomously executes complex multi-step workflows across apps and files, with seamless integration of Baidu's proprietary AI skills. - **Digital Humans**: Achieved ~80% cost reduction over the past two quarters, lowering adoption barriers. An overseas digital human platform was launched, supporting 24 languages with deep localization for global e-commerce use cases, with partners including Jingdong, Suoyebang, TikTok and Shopee. - **MiaoDa (Vibe Coding)**: A no-code AI development platform. March monthly active users grew 70% quarter-over-quarter, domestic paying user rate reached 3x the end of 2025 level. MiaoDa 3.0 launched with an enterprise version, mobile app and support for standalone mobile app generation. - **FAMO Agent 2.0**: An enterprise operational AI agent that enables domain experts to interact via natural language without coding. It delivers incremental efficiency gains for complex industrial operations, such as port scheduling at Qingdao Port. - **AI Search & Ernie Assistant**: AI search transformation continued, with reduced low-quality content and improved user experience. March daily active users of Ernie Assistant nearly doubled year-over-year, daily average conversation rounds tripled, and next-day retention improved meaningfully. ### ApolloGo Autonomous Ride-Hailing - Maintained strong scaling, with an industry-leading safety record. International expansion is progressing: open road testing is upcoming in Switzerland, test vehicles have arrived in London for upcoming testing, and fully driverless operations are live across multiple Dubai zones with a standalone ApolloGo app launched. - New use cases are being explored, such as fully driverless vehicle rental at Haikou Airport in partnership with Car Inc., expanding monetization opportunities beyond traditional commuting.

Guidance

• Management confirms AI is now the primary growth driver of Baidu, and expects the contribution of AI-powered business to continue growing, with significant open opportunities across AI cloud, AI applications and global Robotaxi markets. • Long-term operating margin trajectory is expected to be structurally positive, driven by a revenue mix shift to higher-margin AI businesses: GPU cloud has structurally higher margins than traditional CPU cloud, AI applications are naturally high-margin subscription businesses, and ApolloGo unit economics are improving consistently as it scales. • Management expects AI cloud blended margins to improve structurally as GPU cloud grows as a share of total cloud infrastructure revenue, with a durable long-term profitability trajectory for the cloud business. • Baidu will continue maintaining strategic investment intensity in AI while preserving financial discipline, and will balance long-term AI investment with shareholder returns as a core priority.

Segment performance

Baidu Group total Q1 2026 revenue: RMB 32.1 billion, down 2% quarter-over-quarter. - Baidu General Business: RMB 26.0 billion total revenue, up 2% year-over-year, flat quarter-over-quarter. Core AI-powered business revenue reached RMB 13.6 billion, up 49% year-over-year, accounting for 52% of Baidu General Business revenue, marking the first time it has exceeded half of total general business revenue. - AI Cloud Infra: Revenue grew 79% year-over-year, with GPU cloud revenue accelerating to 184% year-over-year growth. Combined AI Cloud (AI Cloud Infra + AI applications) total revenue was RMB 11.3 billion. - ApolloGo (Robotaxi): Delivered 3.2 million fully driverless rides in Q1, maintaining triple-digit year-over-year growth. - iQiyi: RMB 6.2 billion revenue, down 8% quarter-over-quarter. Profitability: - GAAP operating income: RMB 3.2 billion, operating margin 10%. - Non-GAAP operating income: RMB 3.8 billion, non-GAAP operating margin 12%. - Net income attributable to Baidu: RMB 3.4 billion, net margin 11%, diluted EPS RMB 8.76. - Non-GAAP net income attributable to Baidu: RMB 4.3 billion, non-GAAP net margin 14%, non-GAAP diluted EPS RMB 12.06. - Operating cash flow: RMB 2.7 billion, positive for 3 consecutive quarters.

Risks & headwinds

• Forward-looking statements are subject to general risks and uncertainties that could cause actual results to differ materially from current expectations, detailed in SEC and Hong Kong Stock Exchange filings. • Domestic AI chips face near-term challenges related to manufacturing capacity and supply chain maturity, as demand is growing faster than current supply capabilities. • Scaling fully driverless Robotaxi operations at larger volumes encounters increasingly complex real-world scenarios and system/operational complexities that require continuous refinement of operations and technology. • The Robotaxi industry and global value chain are still evolving, with business models not yet fully defined, requiring flexible strategic positioning as the market matures.

Analyst Q&A

• Q: What are the key drivers of AI cloud infra growth, is there enough capacity to support future growth, and how will AI cloud margins evolve?

  A: Strong broad-based enterprise demand for AI infrastructure, with inference demand growing particularly quickly as customers move beyond model training to production AI deployment. Qianfan MaaS is seeing rapid traction from expanded third-party model support, enabled by Baidu's high-performance inference capabilities. New customers are being won across non-traditional industries like retail and consumer brands, expanding the total addressable market. Baidu is actively expanding capacity and leveraging full-stack AI advantages to improve efficiency, and GPU cloud has structurally higher margins than traditional CPU cloud, so blended cloud margins will improve structurally as GPU cloud grows its share.

• Q: How does Baidu position Ernie in the competitive foundation model landscape, and what are the investment priorities for future iterations?

  A: Strong in-house foundation model capabilities remain essential, so Baidu will continue investing heavily in Ernie, following an application-driven iteration approach aligned with real product needs. Future iterations will focus on improving capabilities for Baidu's core high-value applications: enhanced intent understanding and content quality assessment for AI Search, better multi-modal capabilities for more engaging digital humans, stronger coding support for MiaoDa, and improved complex scenario problem-solving for FAMO Agent. Baidu also maintains a portfolio of smaller optimized models for specific use cases to deliver the best outcome for each application.

• Q: With AI-powered business now over 50% of revenue, what is the long-term operating margin outlook and what are the key drivers of expansion?

  A: Fast-growing AI businesses are still scaling, but as they grow as a share of the revenue mix, they will drive margin expansion. AI cloud infrastructure has higher structural margins than traditional CPU cloud, AI applications are naturally high-margin subscription businesses, and ApolloGo has already achieved unit breakeven in its largest Chinese market with a clear path to profitability as it scales. Additional margin levers include company-wide operational efficiency optimization, widespread AI adoption to improve internal productivity, and improving server utilization rates, leading to a compelling and sustainable long-term margin trajectory.

• Q: What is Baidu's AI agent strategy, and how do you expect AI agent monetization to evolve?

  A: AI applications are a core strategic priority, with four key high-value focus areas: AI search (advertising-based monetization), digital humans, MiaoDa no-code development, and FAMO enterprise agents (all three primarily use subscription, usage-based or token-based pricing). DoMate is a new general-purpose productivity agent that will bring AI to daily workflows. The global AI agent industry is still early, but over time monetization is expected to shift from token-based pricing to results-oriented pricing for completed tasks and productivity gains, creating a much larger total market than token-based models.