Baidu released its first-quarter 2026 financial results, reporting total revenue of 32.1 billion yuan. Revenue from Baidu's core business reached 26 billion yuan, a 2% year-on-year increase, exceeding market expectations. AI business revenue contributed 13.6 billion yuan, accounting for 52% of the core business revenue and marking growth for several consecutive quarters.
Following the earnings release, Baidu's Chairman and CEO Robin Li, Executive Vice President and General Manager of the Mobile Ecosystem Group (MEG) Rong Luo, Executive Vice President and President of the Intelligent Cloud Group Shen Dou, and CFO He Haijian attended the subsequent earnings conference call to discuss key points and answer analyst questions.
The following are highlights from the analyst Q&A session:
JPMorgan analyst Alex Yao: First, congratulations on the accelerated revenue growth in AI cloud infrastructure this quarter. Could management elaborate on what you see as the main drivers behind this growth momentum? Additionally, does the company currently have sufficient computing capacity to support future growth needs? Finally, could management help us compare the profitability levels between AI cloud business and traditional cloud business (e.g., CPU cloud)? How should we understand the long-term profitability trend for the AI cloud business?
Shen Dou: We have observed a significant increase in enterprise demand for AI infrastructure, both on the training and inference sides. Growth on the inference side is particularly strong, which is a very healthy and positive signal. It indicates that users are no longer just in the model training phase but are beginning to deploy and use AI technology more rapidly across more business processes.
Closely related to this, our MaaS (Model-as-a-Service) platform has also made strong progress. "Baidu Qianfan" is one of the few MaaS platforms in China. As Robin mentioned earlier, in addition to "ERNIE," we have rapidly expanded the model library of "Baidu Qianfan" to include currently popular models such as DeepSeek, GLM, and MiniMax. We continue to see growth in token consumption from external users.
It's worth noting that our "support for new model integration" is not simply "plug-and-play." It requires higher-level inference capabilities and more efficient model service capabilities. Baidu has the ability to run these models stably at scale and support more token demand under the same computing power conditions.
Based on our observations, user demand is widely distributed across various industries, including internet, autonomous driving, gaming, and advanced manufacturing. Not only are existing customers increasing spending, but we are also continuously acquiring new customers, including from industries not traditionally heavily reliant on AI and cloud computing, such as retail and IP-driven consumer brands.
Overall, the addressable market continues to expand. In an environment of strong demand and relatively tight supply, we are actively increasing computing capacity and improving resource utilization efficiency to better support growing customer demand.
Our confidence in capturing this user demand stems from our differentiated, full-stack AI capabilities. This provides us with several structural advantages. First is efficiency. Because we possess full-stack self-research and optimization capabilities, we can offer customers highly competitive cost-performance. Second, our self-developed AI technology and capabilities are widely recognized in the industry and have been validated in real-world scenarios.
Regarding profitability. The core driver of profitability is the change in business structure. GPU cloud typically has higher profitability than traditional CPU cloud for several reasons. First, GPU cloud is technically more complex and has higher barriers to entry. Baidu is actually one of the earliest cloud service providers in China to build GPU cloud at scale and has maintained a leading industry position. Second, current demand remains very strong while high-quality supply is relatively tight. Customers prioritize product stability and availability, not just cost. Additionally, our self-developed chips and full-stack AI capabilities provide us with further room to optimize costs, and the continuous optimization of our customer structure also helps improve profitability.
As the proportion of GPU cloud in total cloud infrastructure revenue continues to increase, we believe the comprehensive profitability of Baidu's cloud business will continue to improve. This is a structural, long-term trend. In summary, we are confident in the long-term profitability of Baidu's cloud business.
Citigroup analyst Alicia Yap: My question concerns Baidu's foundational model. In an increasingly competitive environment, how does Baidu view the positioning of the ERNIE model? Looking ahead, what are the company's investment plans, and what are the key directions for future model iterations?
Robin Li: The landscape for large models is changing very rapidly. Both in China and globally, various players are continuously releasing new models. We believe model capabilities will continue to evolve quickly, and strong self-developed foundational model capabilities remain indispensable. Therefore, we will continue to firmly invest in the ERNIE model.
At the same time, we have always believed that the value of a model is ultimately realized through applications. Therefore, we have consistently adhered to an "application-oriented" path. Each iteration of the ERNIE model is driven by real product needs and specific scenarios. Recently, we released ERNIE 5.1, which has achieved leading results in text capability and search capability among industry large language models, reflecting the continuous progress of ERNIE in text understanding, reasoning, and search capabilities.
Looking ahead, we will continue to advance ERNIE model iterations around core application scenarios, including AI search, digital humans, and intelligent agents. We firmly believe these are the application directions with the greatest value potential.
Our goal is to build the strongest capabilities in the most critical areas of demand. For example, we will continuously improve the ERNIE model's ability to understand user intent and assess content quality, enabling AI search to provide more accurate, higher-quality, and more intelligent results. Simultaneously, we are enhancing the ERNIE model's text and multimodal capabilities to make digital humans more vivid and realistic, and to interact more effectively with users and drive conversions in scenarios like e-commerce live streaming.
We will also strengthen coding capabilities to better support programming work, allowing users to build applications using natural language. As programming ability becomes increasingly critical in the AI era, this will also become a key focus for our investment. Furthermore, we will continuously improve the ERNIE model's ability to identify better solutions in complex, real-world scenarios, helping enterprises across various industries enhance operational efficiency.
To better support the above directions, we have also made organizational adjustments to the model team and will continue to optimize the organizational structure as needed. We are confident that Baidu's ERNIE model will continue to strengthen in all the aforementioned aspects.
Beyond the ERNIE model, we also have a series of smaller, faster, and more efficient models, as well as model combinations optimized for different scenarios. Different application scenarios have varying requirements for capability, cost, latency, and deployment efficiency, and our goal is always to provide the best solution for each application scenario.
In the long term, we believe the potential space for AI applications remains vast, with much untapped potential. As more AI use cases emerge, the value of our "application-driven" path will become more prominent, and the ERNIE model will also become stronger and more valuable in this process.
UBS analyst Wei Xiong: First, congratulations on the continued strong growth of the company's cloud business. I would like to ask management about profitability. With the sustained and rapid growth of AI cloud infrastructure revenue, AI-driven businesses now account for over 50% of total revenue. As investors, how should we view Baidu's long-term operating margin level? Furthermore, what key factors does management believe will drive future margin improvement?
He Haijian: In the first quarter, as you have seen, the AI-driven portion of Baidu's core business (primarily businesses outside traditional online marketing) accounted for over 50% of total revenue for the first time. This is an important milestone, reflecting both the continuous increase in AI technology's contribution to the company and the fact that our revenue structure is becoming more diversified.
These high-growth businesses are currently still in the scale expansion phase. As their proportion in the revenue structure further increases, we expect they will not only drive revenue growth but also lead to margin expansion, thereby providing the company with multiple, sustainable long-term drivers for profit improvement.
At this stage, we are investing with firm conviction in the most strategically significant opportunities, while also placing great importance on the return on investment (ROI) of these investments. We believe the capabilities we are building today will influence our profit structure for many years to come and help Baidu form a lasting competitive advantage.
Let me discuss several key business areas specifically.
First, the AI cloud infrastructure business. GPU cloud is structurally more profitable than traditional CPU cloud, driven by stronger demand, tighter supply chains, higher technical barriers, and more favorable pricing power. Therefore, as its share in the overall business structure continues to rise, we expect it to become a significant driver for margin improvement and expansion.
Second is the AI application business. This type of business naturally possesses higher profitability, primarily due to its sticky subscription model and the gradual formation of operating leverage as the business scales.
Third is the autonomous taxi (Robotaxi) business. Since achieving unit-level breakeven in Wuhan, the unit economics have continued to improve. Although our business is still in the investment phase, as scale expands, we believe the path to future profitability is becoming clearer.
Additionally, there are some extra levers worth emphasizing at the company level: First, we are continuously driving cost optimization and operational efficiency improvements across the entire company; second, we are applying AI technology on a large scale to enhance internal productivity; third, at the infrastructure level, we are continuously improving server utilization, which over time will also directly improve margins.
Overall, our revenue structure is shifting towards higher-margin, higher-growth businesses. Full-stack AI capabilities provide us with cost efficiency advantages, and company-wide productivity improvements are continuously accumulating. We believe the company's margin trajectory is very attractive in the medium to long term and are confident in its sustainability.
Morgan Stanley analyst Gary Yu: My question is about Baidu's autonomous taxi business. Could management update us on the latest progress of Baidu's Robotaxi business in overseas markets? How should we understand the operational scale of the business? What is the revenue structure (proportion) between domestic and overseas Robotaxi business? Furthermore, how do domestic and overseas businesses compare in terms of profitability? In the long term, how does Baidu view its role in the Robotaxi industry ecosystem? Does management see Baidu more as an operator, a technology provider, or an operating platform?
Robin Li: First, in terms of scale, "Apollo Go" remains a globally leading autonomous taxi operating platform. As of April, our cumulative rides have exceeded 22 million.
China is one of the most open Robotaxi markets globally, so our domestic operational scale currently significantly leads overseas markets. This is a natural result. However, we also see that more markets worldwide are gradually opening up to Robotaxi, and the regulatory environment is becoming more positive and relaxed.
We are very pleased that our long-term operational experience in China has laid a solid foundation for our international expansion. Over the past few quarters, we have made significant progress in overseas markets. In fact, we only began accelerating overseas expansion a few quarters ago, but our business footprint already covers several key markets in Europe, the Middle East, and Asia.
This pace of expansion reflects the replicability and scalability of our technology and operational systems in different market environments. Our confidence in overseas expansion primarily stems from our large-scale operational capabilities validated in the Chinese market.
Over the years, we have accumulated experience in fully driverless operations in real road environments, covering complex road conditions, operational challenges, and various "long-tail problems" that only appear at large scale. This experience is not quickly replicable, and it has continuously optimized our algorithms and operational standards, making Baidu's Robotaxi system more robust with every ride.
When we enter global markets, this accumulated experience is also exported, helping us localize faster. For example, our business progression from Hong Kong to London is a good example. Hong Kong has been an important right-hand drive market for us. Over the past year or more, we have accumulated substantial operational experience there, which has also helped us smoothly enter the London market recently, as London is also a right-hand drive market.
Regarding profitability, Apollo Go has achieved unit-level breakeven in its largest operating cities in China, even though our pricing remains relatively low. As we gradually expand globally, the overall pricing environment is more favorable. Therefore, we believe our overseas business will have stronger profit potential after scaling. Meanwhile, the overall size of the international market (excluding China and the US) may even be larger than the domestic Chinese market.
Finally, regarding our long-term role in the overall Robotaxi ecosystem. I believe it is still premature to draw conclusions, as the entire Robotaxi industry is still evolving rapidly, with industry chain structures and business models still forming. Our current business focus remains on continuously expanding scale, deepening technological and operational advantages, and maintaining global leadership. On this basis, as the overall ecosystem matures, we will have greater strategic flexibility to define our role and capture long-term value.
Comments