This conference call transcript was computer generated and almost certianly contains errors. This transcript is provided for information purposes only.EarningsCall, LLC makes no representation about the accuracy of the aforementioned transcript, and you are cautioned not to place undue reliance on the information provided by the transcript.
Operator
Hello, and thank you for standing by for Baidu's first quarter 2023 earnings conference call. At this time, all participants are in listen-only mode. After the management prepared remarks, there will be a question and answer session. Today's conference is being recorded. If you have any objections, you may disconnect at this time. I would now like to turn the meeting over to your host for today's conference, Wan Lin. Baidu's Director of Investor Relations. Please go ahead.
Wan Lin
Hello, everyone, and welcome to Baidu's first quarter 2023 earnings conference call. Baidu's earnings release was distributed earlier today, and you can find a copy on our website as well as on mutual services. On the call today, we have Robin Li, our co-founder and CEO, Rong Luo, our CFO, Zhou Shen, our EVP in charge of Baidu AI Cloud Group, ECG, and Jen Yuli, our SVP in charge of Baidu Intelligent Driving. After our prepared remarks, we will hold a Q&A session. Please note that the discussion today will contain forward-looking statements made under the safe harbor provisions of the U.S. Private Security Litigation Reform Act of 1995. Forward-looking statements are subject to risk and uncertainty that may cause actual results to differ materially from our current expectations. For detailed discussions of these risks and uncertainties, please refer to our latest annual report and other documents filed with the SEC and Hong Kong Stock Exchange. Baidu does not undertake any obligation to update any forward-looking statement, except as required under applicable law. Our earnings press release and this call include discussions of certain annotated non-GAAP financial measures. Our press release contains a reconciliation of the anodities and gap measures to the anodities directly comparable gaps, and it's available on our IR website at ir.by2.com. As a reminder, in addition, a webcast of this conference call will be available on By2IR's website. I will now turn the call over to our CEO, Robin.
Robin Li
Hello, everyone. 2023 is off to a good start. with our bins performing well across the board and driving strong growth. In Q1, Baidu Core's revenue growth really accelerated year on year and supported an improvement in our operating margin. Our online marketing bins have seen a positive impact from China's overall economic improvement. After the Chinese New Year, there was a quick recovery. which was especially beneficial for many of our advertisers in offline sectors, such as travel, healthcare, bin services, and local services. Moreover, we've been working hard to enhance user experience for Baidu apps, and this has led to good user metrics and increased online marketing revenue for the first quarter compared to the previous year. Our AI cloud achieved profitability on a non-GAAP operating level this quarter. Starting in Q2 last year, we have focused on building sustainable growth for our cloud business by phasing out low-quality projects and businesses. At the same time, we have been building standardized AI solutions for key use cases and replicating them from one project to another. This approach has led to an increase in AI cloud margins over the past few quarters, and we have seen profitability on the non-GAAP operating level this quarter. Today, all eyes are keenly focused on the transformative impact of generative AI. For today's call, I will first provide an update on this topic and then walk you through our operational highlights. in AI cloud, intelligent driving, and mobile ecosystem. Generative AI represents a tremendous opportunity for us. It can be compared to the introduction of the internet and smartphones. To capture this opportunity, we leveraged our technology capabilities and our extensive experience in search, knowledge graph, and dialogue. We were able to quickly deploy Ernie Bot in May-March, which became the first GPT model owned and operated by a public company. This launch demonstrated our significant investment and achievement in AI over the past 10 years, and it also highlights our position as the operator of China's largest search engine. With our search engine, we have gathered a vast amount of data, facts, and knowledge for natural language understanding, which has been critical in training our model. Baidu has established a robust range of AI capabilities powered by its four-layered AI architecture, which encompasses cloud infrastructure, the PedoPedo deep learning framework, large language models, learning, and AI applications. We have this distinctive advantage that helps us optimize entire process and reduce cost for model training and inference. One of our significant advantages is owning Pedal-Pedal, which is the leading deep learning framework in China. We have developed the PedoPedo in a way that aligns with ErnieBot, and this creates a strong synergy between these systems. Let me give you an example. Training and running large language models require a lot of data, which can be complex and overwhelming, and demand significant computing resources. To tackle this challenge, we use our proprietary full-stack AI architecture. we are continuously looking for new techniques to make our four layers work seamlessly, which helps us achieve better performance and quicker throughput. Specifically, we have improved compatibility between Pedal-to-Pedal, the large language models, and the cloud infrastructure. As a result, since launching EarlyBot in March, our normalized Queries per second have increased by almost 10 times, indicating significant efficiency gains. Baidu is rapidly integrating EarlyBot into all of our businesses for testing. The goal is to leverage its capabilities to enhance, rebuild, and create both our enterprise and consumer-facing products and services, thereby revolutionizing our business. We are also working on building an ecosystem around Ernie Bot to promote the adoption of generative AI in a wider range of use cases. This strategy should enable us to gain market share across all sectors and achieve sustainable long-term growth. As more people use Ernie Bot, more products integrate Ernie Bot in seamless ways, we will be able to gather large-scale human feedback and iterate the model rapidly and tailor to real-life market needs. On the enterprise-facing bin site, Generative AI is expanding the total addressable market of the cloud industry. This is because cloud service providers can use Generative AI to add or enhance capabilities to their current offerings, developed new solutions and applications to address unfulfilled market demand, and drive productivity and efficiency. As a result, since we introduced ErnieBot, we have observed more and more enterprises from various industries started their model training on our cloud. Our powerful foundation models and our unique full stack AI capabilities differentiate us from our competitors, which gives us an edge in the market. In March, sales lead increased by more than 400% year-over-year, which is a significant achievement for us. We believe that our AI cloud will continue to gain market share over time. I'll provide some examples of how EarniBot is helping our customers to gain efficiency. In the financial services industry, we have partnered with a prominent domestic insurance company to help them enhance their digital experiences for employees and clients. To achieve this, we will utilize our AI capabilities, including PedoPedo and Ernie on their private cloud. Additionally, we will create customized tools and technologies to assist them in training and refining models on EarningBot with their unique data. And we will continuously improve the system over time. This model will power various applications, including helping financial advisors analyze large amounts of research reports and public information. We believe that this project has the potential to establish a new industry standard for innovation and efficiency in China's financial services sector. In the software industry, we are collaborating with some productivity software developers to test the integration of Ernie Bot into their application. with the goal of automating certain workloads and expanding the capabilities of their end users. Additionally, we are partnering with some ERP companies who are exploring the use of Ernie Bot to assist their programmers in writing code. In the education industry, we opened Ernie Bot to a leading online education company via Model as a Service. helping them fine-tune their own model on Baidu AI Cloud. One of the features we're working on together is an enhanced, sophisticated AI tutor, enabling them to better serve their students. We've observed that industries experiencing a surge in demand after the pandemic are interested in using Ernie Bot to improve their business operations more efficiently. Online travel companies are a good example of this trend, as some of the biggest players are testing the use of Ernie Bot to assist their customer service team in managing customer queries, suggesting trips, and directing customers to relevant information. The feedback so far has been positive. and customers are also beginning to switch their IT expenditures to our cloud. On consumer-facing businesses, we are utilizing ErnieBot to enhance our product for the users. We're currently beta testing an upgraded version of Baidu Search, which is powered by ErnieBot. This version provides natural conversational interactions for a more seamless search experience. We are also testing the integration of ErnieBot with Baidu WenKu, our online document sharing platform, to enable ErnieBot to work alongside users and assist with searching, summarizing, and generating content using natural language. Additionally, we are incorporating ErnieBot into other consumer-facing products such as Baidu Map, Baidu Keyboard, Baidu Post, and XiaoDu. Furthermore, we are testing a standalone application for ErnieBot. So far, we have been encouraged by the positive user feedback at the current testing stage and look forward to bringing these experiences to more users in the coming months. In the new era, AI Chatbots will become a traffic gateway for people to seek information and content. Therefore, we should continue to gain user traffic and mindshare, which will drive monetization over time. We're also working to build an ecosystem around early bots with our partners to fuel innovation and drive growth. On the regulatory front, as a leader in AI and an advocate for AI ethics, Baidu continues to promote the responsible use of AI. To comply with regulatory requirements, any new technology, product, or service must undergo government review and approval before large-scale deployment. We have applied for review and approval of ErnieBot and will roll it out at a large scale to our customers and users once we get the green light. Baidu has been operating search in China for more than 20 years and has extensive experience with Chinese culture and regulatory environment, which we believe will help us navigate the regulatory landscape. In early April, the Cyberspace Administration of China released a new draft regulation for comments on generative AI, which explicitly endorses the development of self-designed AI algorithms and deep learning frameworks. We believe that regulators' active engagement in generative AI in the early stage will raise the bar to entry, and we are well positioned for that. Overall, we are thrilled about the prospects that lie ahead with ErnieBot, and we will continue to invest in it to unlock the tremendous potential that we envision. Now, let's have a look on the first quarter operational highlights. In Q1, revenues from AI Cloud increased 8% year-on-year, which is an improvement from the 4% year-over-year growth in the fourth quarter last year. We're pleased to report that AI Cloud achieved profitability on the non-GAAP operating profit level during the quarter. The margin improvement underlows two points of progress. One is our ability to optimize operations and drive sustainable and healthy revenue growth. Two is our efforts to standardize our AI solutions and applications and efficiently replicate them across different customers and sectors. Since Q2 of last year, we have phased out low-margin businesses and projects for AI Cloud, leading to a healthier business portfolio and cost structure. At the same time, we continue to standardize AI solutions for key use cases for replication. In the water industry, for example, last year we replicated AI solutions we developed for predicting water usage and controlling water pressure from Quanzhou Water Company to Guangzhou Water Company. And in Q1, we deployed the majority of our earlier standardized solutions to the Lintau County Water Company in Gansu Province. This example demonstrated the scalability and margin expansion of our standardized approach in creating AI solutions for traditional industries. As I mentioned earlier, Generative AI is creating a host of opportunities for the cloud industry and expanding the cloud market significantly. Therefore, we will continue to invest and actively explore ways to grow our cloud services. Moving into intelligent driving, we continue to make solid progress. In Q1, the number of rides provided to the public on open roads increased over three-fold compared to the same period last year and reaching nearly 660,000. Thanks to the remarkable growth in rides and our strong track record of safe operations, we have received more permits from governments to offer fully driverless ride-hailing services in China. On March 17, ApolloGo was allowed to provide fully driverless ride-hailing service in Yizhong region of Beijing. As of today, we are providing the services in three cities, Beijing, Wuhan, and Chongqing. We are prioritizing fleet expansion and passenger adoption in regions where we can provide driverless commercial operation, meaning it's fully driverless and we can charge. Our efforts are yielding good results. In Wuhan, where we first received the permit to provide such services on public roads last August, we now have more than 135 driverless vehicles, and more than one-third of the paid rides ApolloGo provided were in fully driverless vehicles. We are confident that the number of rides provided by fully driverless vehicles will continue to increase, leading to cost optimization and creating a profitable business model. We firmly believe the demand for autonomous driving services will continue to grow at a rapid pace, and we will further dedicate our technology and operational capabilities to better provide safe and reliable services to the public. Moving to mobile ecosystem. I'm pleased to report that online marketing revenues returned to positive year-over-year growth in Q1, resulting in an increase in operating margins. One contributing factor to this growth was the overall improvement in China's economic recovery and commercial involvement as a result of China's reopening. As I mentioned earlier, many offline-related verticals, travel, healthcare, business services, and local services have substantially increased. they are ascending on our platform. Some of these verticals have already rebounded to above pre-pandemic levels, indicating solid signs of recovery. Our own initiatives aimed at improving the product experience, such as incorporating short videos and product listing information within the Baidu app. also played a significant role in the revenue increase. Additionally, revenues from retail and e-commerce continue to grow, mainly driven by merchants' direct spending on our platform. The quarterly GMV facilitated by Baidu Search grew by 55% year-over-year in Q1, demonstrating our continuous traction in the e-commerce sector. Our users and traffic continue to grow in the quarter, with Baidu App's MAU increased by 4% year-over-year in March, and feed distributed through Baidu App and mobile search queries maintaining steady growth. This reflects the high value that users place on Baidu Search as their primary source of information. During the quarter, we continue to focus on enhancing user experience and customer satisfaction. On the user side, Baidu app has a wonderful flow feature on the Explorer page, which is particularly attractive to young females, enabling an infinite browsing experience and catering to their diverse needs. we expanded our immersive video feature to more and more use cases on Baidu apps, including incorporating it into search results. Additionally, we continuously improve our one-shot search results through AI to provide better search experience. These efforts have helped to increase user engagement and expand our user base. On the advertiser side, since launching our AIGC capabilities late last year, we've been working with advertisers to create more effective ads. By allowing advertisers to specify their requirements, our platform generates customized options for them to choose from. For instance, a local travel agency turned to our AIGC capabilities to create ads, Our AI-generated ads outperformed those created by the advertisers, delivering an average of about 20% higher sales leads than them. In Q1, advertisers adopting AIGC increased their ad spending year-over-year and quarter-over-quarter on our platform, demonstrating the effectiveness of our technology in driving wins results. Earlier this year, we began using our AIGC capabilities to help advertisers respond to customer inquiries on our platform. By harnessing the power of AI, our advertisers can provide professional and timely answers to their customers' inquiries, which in turn drives an increase in sales leads. For instance, a B2B advertiser witnessed a monthly increase of over 100% in sales leads in March after using our AIGC technology. We are expanding the service to increasing number of advertisers, helping them better serve their customers and drive improved business results. These efforts have resulted in improved ROI for advertisers on our platform. Looking ahead, we remain committed to assisting our advertisers in reaching and converting their target audience by leveraging our cutting-edge generated AI technology. With that, let me turn the call over to Rong to go through the financial results. Thank you, Robin.
Generative AI
Now let me walk you through the details of our first quarter financial results. Total revenue was RMB 31.1 billion, increasing 10% year-over-year. Revenue from Baidu Call was RMB 23 billion, increasing 8% year-over-year. Baidu Call's online marketing revenue was RMB 16.6 billion, increasing 6% year-over-year. Baidu Call's non-online marketing revenue was RMB 6.4 billion, up 11% year-over-year. In Q1, AI Cloud increased by 8% year-over-year to RMB 4.2 billion. Revenue for IT was RMB 8.3 billion, increasing 15%, 1.5% year-over-year. Cost of revenue was RMB 15.2 billion, decreasing 3% year-over-year. Vital cost of revenue was RMB 9.4 billion, decreasing 4% year-over-year. Operating expenses were RMB 11 billion, increasing 7% year-over-year, primarily due to an increase in channel spending and promotional market expenses. Baidu Core's operating expenses were RMB 9.5 billion, increasing 5% year-over-year. Baidu Core SG&A expenses were RMB 4.5 billion, increasing 16% 1.6 year-over-year. SG&A accounted for 20% of Baidu Core revenue in the quarter, compared to 18% 1.8 in the same period last year. Baidu Core R&D expenses were RMB 5 billion, decreasing 3% year-over-year. IMD accounting for 22% of Baidu Core revenues in the quarter and decreased from 24% in the same period last year. Operating income was RMB 5 billion. Baidu Core's operating income was RMB 4.1 billion. And Baidu Core's operating margin was 18% 1A. And Nongat operating income was RMB 6.4 billion. Nongat and Baidu Core's operating income was RMB 5.4 billion. And Nongat Baidu Core operating margin was 23%. Total other income net was RMB 2.6 billion, which is mainly including a fair value gain of RMB 1.5 billion from long-term investments. In the first quarter last year, we recognized a fair value loss of RMB 3 billion. A significant portion of the long-term investments include, but not limited to, investments in equity securities of public and private companies and private equity funds. This is subject to quarterly fair value adjustment which may contribute to net income volatility in future periods. Income tax expenses was RMB 1.2 billion, increasing 205% year-over-year, primarily due to an increase in profit before tax. Net income attributable to Baidu was RMB 5.8 billion, and diluting earnings per ADS was RMB 15.92. Net income attributable to Baidu core was RMB 5.5 billion, Nongat net income attributable to Baidu was RMB 5.7 billion. Nongat diluted earnings per ADS was RMB 16.1, and Nongat net income attributable to Baidu Core was RMB 5.3 billion, and Nongat net margin for Baidu Core was 23%. As of March 31, 2023, cash equivalents, restricted cash, and shorting investments were RMB 194 billion, and cash, cash equivalents, restricted cash, and shorting investments, excluding IT, were RMB 188.8 billion. Free cash flow was RMB 4.5 billion, and free cash flow excluding IT was RMB 3.5 billion. Baidu Call had approximately 36,000 employees as of March 31, 2023. On a separate note, in the quarter, ITC generates about RMB 1 billion in non-GAAP operating income and about RMB 1 billion in free cash flow. With that operator, let's now open the call to questions.
Operator
Thank you. Ladies and gentlemen, we will now begin the question and answer session. If you wish to ask a question, please press star and 1 on your telephone keypad and wait for your name to be announced. If you wish to cancel your request, please press star and two. Participants are requested to restrict one question at each time. Your first question comes from Alex Yao with JP Morgan. Please go ahead.
Alex Yao
Good evening, management team. Congrats on a strong quarter and thank you for taking my question. I have a few on the advertising side of the business. What is the month-on-month trend for ads recovery since March, and how big is the vertical exposures to offline-related verticals? How is the advertiser demand changing on your platform? And then lastly, how should we think about who your ads growth rate, so what is the need to longer-term growth potential for ad business beyond the COVID normalization? Thank you.
Robin Li
Hi, Alex. This is Robin. Thank you for your question. We're happy to see significant progress in our online marketing. For Q1, revenue made a remarkable comeback and a return to high single-digit growth. Recall that last year, online marketing revenue declined year over year for all four quarters. I think it was a 4% decline in Q1, 10% in Q2, 4% in Q3, and 6% in Q4. Now, we are in a positive territory, and the momentum got further strengthened in Q2. The Q1 growth was primarily driven by a number of verticals in the offline sector that I just mentioned during my prepared remarks. Offline sector contributes to a sizable amount of our online marketing revenue. Particularly, our largest revenue contributor, healthcare, has been improving and is back to the pre-pandemic level. So travel is also improving, apparently, and we saw that the Labor Day holiday was really, really hot. The growth in our online marketing revenue has also had a positive impact on the margins of our mobile ecosystem. So the overall margin for Battlecore got better. The success is due to our focus on user and customer experience on our platform. And furthermore, we have observed a strong growth in user engagement and traffic. Despite the reopening of China and people spending more time offline, our MAU for Baidu app grew 4% year-on-year in March. And our feed content distributed through Baidu app also demonstrated robust growth. And looking ahead, we anticipate advertisers will continue to allocate more ad dollars to performance-based platforms. And we believe Baidu Search is well-positioned to benefit from this trend. In addition, we see generative AI as a significant opportunity that complements our search capabilities and provides new avenues for users to access and create content. As we integrate Ernie Bot into our apps, particularly the search, we expect user traffic and engagement will go up, thereby further supporting the growth of our online marketing revenue. EarlyBot can also enhance our commercial products, presenting an exciting opportunity for us to strengthen our offerings. With EarlyBot's empowered commercial products, we can capture a larger market share and deliver even greater value to our advertisers. This additional dimension gives us further confidence in our online marketing business and its ability to drive sustained growth in revenue. Based on this development and our strong fundamentals, we expect our online marketing revenue to outperform the growth of China's GDP.
Alex
Thank you.
Operator
Thank you. Our next question comes from Alicia Yap with Citigroup. Please go ahead.
Alicia Yap
Hi. Thank you. Good evening, Robin, to the management team. Thanks for taking my questions. My question is related to margins and spending. be the major margins driver this year after large scale of the cost cutting effort last year. How should we think about core operating margin trend for the rest of, especially given there will be a few moving factors that will impact margins, such as, for example, the potential investment related to the large language model, the autonomous and others. Thank you.
Generative AI
Hello, Alicia. Let me ask you a question. I think we have implemented a number of initiatives last year to improve our operational efficiency, which you can see that in the past four or five quarters. And all of them have led to a notable improvement in our overall margins. These measures not only made the profit numbers look better, but also transformed our organizations, making it leaner and more agile, which will definitely drive the sustainable long-term growth. Purely in number perspective, as a result, you can see that our non-GAAP operating margin expanded from 17% in Q1 last year to 23% in Q1 this year. And the free cash flow also grew by over 200% from a year ago. Notably, all of this growth was also accompanied by a salary revenue in Q1. More specifically, as John has just mentioned, and number one, our mobile ecosystem profit margins improve, which is also together with the revenue reform of the online marketing businesses. Number two is we can see that for Baidu AI Cloud, by focusing on cutting low margin projects and businesses, as well as we're developing more standard AI solutions for key user cases, our Baidu AI Cloud has reached the profitability on a non-debt operating level. This is a very important and significant milestone for us. And looking forward, we aim to continue to enhancing the margins across all of our major operations. While at the same time, we also will continue to invest in the promising new areas, in particularly the earning board and intelligent driving. These striking investments actually are quite aligned with our long-term growth objectives. which also reflect the two areas where we have identified the significant potential opportunities. By focusing on these areas, we aim to capture the larger market shares, and we try to drive sustainable growth while we deliver more values to our shareholders. And as for the near term, I think incremental investment in the new opportunities is also quite manageable in my opinion. Overall, our focus on discipline execution and operational efficiency, and at the same time, we try to seize the long-term growth opportunities. All of these efforts have resulted in the improvement of the overall biblical margins in Q1, and this will continue in the future quarters. We will prudently manage our P&L and leverage our core business improvements. and further drive sustainable growth. Thank you so much, Alicia.
Alex
Thank you.
Operator
Thank you. Our next question comes from Gary Yu with Morgan Stanley. Please go ahead.
Gary Yu
Hi, thank you, and good evening. I have a question related to cloud. So on your AI cloud revenue growth, what's the expectation for growth in 2Q and rest of 2023? And could you help us quantify any incremental cloud revenue in coming quarters coming from early adoption? And a related question on cloud is on margin. With your improvement in revenue growth and margin, how should we look at the normalized revenue growth and margin for cloud going forward. Thank you.
spk12
Okay, this is Do. Thanks for your question on the cloud. So since the second quarter of last year, we have made significant efforts in three ways trying to consistently improve our margins for the AI cloud. Firstly, we are resolute in phasing out the low-quality products and businesses. And secondly, So we are committed to building standardized AI solutions for key use cases and then efficiently replicate them across different customers. Secondly, we are continuously investing in the technical capabilities of our standard cloud offerings to reinforce their competitiveness. So as a result of the efforts, Baidu AI Cloud achieved profitability on the non-GAAP operating level this quarter. In addition to that, after China's reopen, now we finally can meet our clients in person to execute products and present our proposals. And we believe AI Cloud's revenue growth seems accelerated from quarter of last year and it will continue in the coming quarters. Moreover, we believe Ernie Bot could provide a massive opportunity for us to gain shares in the cloud industry. Since its launch in the middle of March, we have received hundreds of thousands of bad testing requests for the Ernie Bot from enterprises, customers, across a range of sectors, such as finance software, education, online travel, and so on. So this overwhelming interest indicates a promising start for the Ernie Bot's debut. So our customers are so eager to explore and utilize Ernie Bot for their businesses. And we all are confident that it will improve their productivity and enhance their competitive edge. Actually, as William mentioned, several customers have opted to develop their own large models on our cloud platform in order to leverage the advantages of the four-layered AI architecture, as Robin just mentioned in the prepared remarks. So as we look ahead, our focus will still remain on achieving healthy and sustainable revenue growth for the existing cloud business. In a while, striving to improve the margins. And in the meanwhile, so we will capitalize on the long-term huge market opportunity of large language models. And for sure, so we are committed to the investment and the value creation of our early bot. So overall, we believe we will continue outgrowing our inner peers in cloud revenue in the long run.
Alex
That's it, Gary. Thank you.
Operator
Thank you. Our next question comes from Lincoln Kong with Goldman Sachs. Please go ahead.
spk07
Thank you, Benjamin, for taking my question. So my question is related to Ernie Bot. So basically, can management further elaborate a bit more on the growth and monetization opportunity for Ernie Bot and the large language model? So how should we think about Ernie Bot will reshape our existing business and especially help us gain share in each of the business line and the markets? Thank you.
Robin Li
Yeah, Ernie Bot has been making great progress since its launch in May, March. For instance, we have already upgraded ErnieBot four times, and we have managed to reduce the inference cost by nearly 90% since its release. While some of our competitors are still focusing on model training, we've moved ahead with a significant reduction of our inference cost. This improvement, a result of our self-developed AI architecture with four layers, which sets us apart and enables us to swiftly reduce costs and enhance efficiency for model training and inference. This strong foundation will facilitate the widespread adoption of ErnieBot. We're using ErnieBot to upgrade almost all of our products and services, enhancing and rebuilding our existing offerings. This enables us to strengthen our position and gain market share in our target market, including cloud services, online marketing, and smart devices. In AI cloud, early body is being tested by more and more enterprises, as Doug just mentioned, and it's proving to be beneficial to customers in all kinds of sectors. helping them solve problems and improve their productivity and efficiency. It truly demonstrates the power of generated AI and large language models. Since launch, we have witnessed increasing adoption of EarlyBot with hundreds of thousands of enterprises and entities queuing for testing. This is an encouraging recognition of by this AI capability, particularly our powerful foundation model. To assist customers to better use ErnieBot, we have developed tools and technologies that support training and refining models using our customers' unique data, which also leads to continuous system improvement over time. And customers are becoming more selective than before. When choosing a cloud provider, they take into consideration the capabilities of generative AI and large language models. We see a growing number of new customers waiting to migrate to our AI cloud. So we expect Ernie Bot to enhance our competitiveness and our pricing power of our AI cloud. In March, we saw more than 400% year-over-year growth in the sales fleet, and we anticipate more opportunities in the future. Additionally, while the cloud industry presents significant opportunities for early bot adoption, there are great opportunities in the online marketing market too. As I mentioned earlier, We're utilizing EarlyBot to upgrade popular consumer-facing products, including Baidu App, Baidu Search, and other applications within Baidu family. The initial user feedback we have received is very promising. Upon widespread implementation, we should quickly gain user traffic and engagement, therefore reshape the competitive landscape in China's internet and mobile sectors. So in summary, early boss progress has been very encouraging and we're excited about the future. We're pushing the boundaries of AI and exploring new avenues of growth.
Alex
Thank you. Thank you.
Operator
Our next question comes from Miranda Wong with Bank of America Securities. Please go ahead.
Miranda Wong
Thank you, Majmin, for taking my question. Congratulations on the good results. My question is about competition. With many other Internet peers launching or about to launch their last language model, how to think about Baidu's competitive advantage? And how does Majmin think about the competitive landscape in China in the near term and the long term? especially for the enterprise-facing business? Thank you.
Robin Li
Yeah, that's a good question. AI and LLM are still in the early stages of development, and we are very happy to see the entry of more players into this field, as it will help the industry to develop rapidly. Moreover, we are actively constructing an ecosystem around Ernie Bot to attract new participants and drive innovation in the emerging direction. In our view, generative AI and the large language model present a very high barrier to entry, and very few companies possess the necessary capabilities to establish a significant presence in this market. The requirements include substantial upfront investment, robust AI capabilities and expertise, and vast amounts of data for model training. Also, extensive experience in delivering appropriate content to Chinese internet and mobile users is crucial. Given such high requirements, we anticipate highly concentrated market with probably two to three major players winning in the end. As mentioned earlier, as the first mover in China, we leverage EarningBot to upgrade and revitalize our entire product and service portfolio. Because we have a large product portfolio and large user base, we are able to get lots of user feedback about BernieBot, which will help us further enhance our leading position in LLM. This approach allows us to convert cutting-edge technology into practical features, products, and solutions that address challenges and support sustainable growth. Baidu has a distinctive AI architecture consisting of four layers that we have developed ourselves. The architecture gives us a competitive advantage in terms of AI capabilities and played a crucial role in the successful early launch of EarlyBot. With this architecture, we are able to quickly decrease cost and improve efficiency in both training and inferencing. For example, we have optimized the performance of PedoPedo for inferencing, and this optimization enables us to more effectively work between the framework layer and the model layer, which leading to a substantial efficiency improvement and cost savings. Devancing and training LLMs require significant resources, including robust cloud architect, cloud infrastructure. To this end, we have our own AI cloud business and have secured a substantial amount of computing power, including the most advanced GPUs to sustain the ongoing training and iteration of Ernie Bot. Another critical point is that our experience in operating China's largest search engine for over two decades has allowed us to build an extensive database in Chinese language, including very valuable data, facts, and knowledge. and enhancing our leadership position in NLP in China. We have extensive experience in reviewing the vast amount of content on search and within our mobile ecosystem. And for 2Beepin, we are able to offer our customers any combination of all different layers of our AI stack. If the customer only needs infrastructure, we can provide that. If the customer needs both infrastructure and framework layer, we can provide that. If the customer needs framework layer and model layer, we're happy to offer those too. And if the customer wants end-to-end solutions, we have that ready. So we are uniquely positioned and very flexible in serving our cloud customers. Based on this competitive advantage, I'm confident that we will be a winner in this market. Thank you.
Miranda Wong
Thank you very much.
Operator
Thank you. Our next question comes from Kenneth Fong with Credit Suisse. Please go ahead.
Kenneth Fong
Hi. Good evening, management. Thank you for taking my question. It has been noted that the Cyberspace Administration of China recently released a draft regulation on the generative AI. So what sort of additional investment are needed to meet the requirement? Thank you.
Robin Li
Okay. During our testing of ErnieBot, we have engaged in close discussions with regulators. Through effective communication, we have found that The principles of content review for generative AI are quite similar to those applied to search. Baidu has been operating China's largest search engine for over 20 years, and we have extensive experience in providing appropriate online content. We also put a lot of effort into both technology and compliance development to ensure our products and services meet applicable regulatory requirements. For important and sensitive topics, we have to make sure AI will not hallucinate. Given that LLM is more or less a probabilistic model, this task is not trivial at all. The requirements are not final yet, so we have to continue to update our strategy as it evolves. Furthermore, we maintain an ongoing dialogue with the relevant authorities and keep them informed about the latest technology trends. We actively share our insights on supporting the responsible and beneficial growth of technology, particularly for generating AI. This proactive effort position us well to meet the requirements and offer industry-leading generative AI solutions in China. Conversely, companies which do not have extensive experience in providing appropriate online content or lacking a track record of working closely with regulators will face significant challenges in offering scalable generative AI solutions in China. So here I would like to emphasize that Baidu has been a leader in AI for social good for a long time. We are one of the early adopters of AI for responsive use in China. We believe our new initiative, Ernie Bot, is the best in its kind in China, and it has the potential to increase productivity and efficiency across various industries, which can contribute to China's economic growth and social development. We expect Ernie Bot to power agricultural enterprises to promote innovation and to optimize energy use, resource allocation, and waste reduction, and to improve disease diagnosis in healthcare industry, therefore enhancing people's well-being and the quality of life. Also, we always attach great importance to AI ethics. Back in 2018, I came up with the four principles of AI ethics. That's like one, AI should be safe and controllable. Two, AI should promote equal access to technologies for humanity. Three, AI should empower humanity to learn and grow instead of replacing them. And then four, AI should bring more freedom and possibilities to humanity. I want to make sure that within Baidu, everyone understands this principle, one, it comes to new AI technologies and products. So as to enable the co-existence between AI and mankind. In addition, we also prioritize data privacy and data security in Bin's operation and Bin's strategy. We have a dedicated committee to oversee data management, and we are currently establishing another committee for AI ethics. And all of this effort aimed at promoting AI for social good. Thank you.
Operator
Thank you. Our next question is from Wei Xiong with UBS. Please go ahead.
Wei Xiong
Thank you, management, for taking my questions, and congrats on the good quarter. My question is about ApolloGo. So could management provide some updates on your robo-taxi business in terms of your strategy and business plan for the year, especially for the fully driverless ride-hailing operation? Is there any timeline you can share regarding the scale-up or the break-even of the business? And what are some catalysts or milestone events that we could expect for ApolloGo this year? Thank you.
spk14
Thanks, Sui-Hsiung. This is Junyu. Our Uppercore remains the largest autonomous ride-hailing service provider in the world. Just as Robin mentioned, in the first quarter, we offered almost 660,000 rides to the public, and this number keeps growing. We aim to continuously enhance our autonomous driving technology by providing more rides to the public. In addition, we will always keep in mind the goal of achieving break-even, gaining skill, and building a sustainable business model over time when we make our operations. To reach this objective, it's important to reduce operational costs and improve operational efficiency. On cost, we are focusing on lowering labor and hardware costs. Those are the major costs for RoboTaxi operation. So, we are working hard to find ways to cut down on those expenses. To reduce the labor cost, we continue to drive development of fully driverless operation. As Roby mentioned in his prior remarks, we are doing really well. in getting closer to our goal, and adding more wheels while we can provide fully driverless rides. In the Yizhuang region of Beijing, just three months after getting fully driverless testing permits, we got the green light to offer fully driverless ride-hailing services to the public on the open roads in middle match. This progress is because of our solid record of providing money-safe and reliable rights to the public in the past, so that we gained trust from both users and the government thanks to this. And in Wuhan, we started offering fully driverless ride-hailing services last August. Now, we are still adding driverless wheels, and expanding operation areas. As Roby mentioned earlier, more than one-third of the rides provided by UpGo in Wuhan are from fully driverless wheels. We think this number will keep going up in the future. We will prioritize safety and reliability for the future fully driverless operation. For hardware cost, We believe that after the release of RT6 and other new models of Robotaxi wheels in the future, it will continue to decrease. On operational efficiency, we have seen a steady increase in the average daily ride nationwide. In the future, we will continue to automate our operational strategy to achieve long-term success in Robotaxi.
Operator
Thank you. Thank you. Ladies and gentlemen, that concludes our conference for today. Thank you for participating. You may all disconnect your lines.
Disclaimer