Baidu, Inc. (NASDAQ:BIDU) Q3 2023 Earnings Call Transcript

Baidu, Inc. (NASDAQ:BIDU) Q3 2023 Earnings Call Transcript November 21, 2023

Baidu, Inc. beats earnings expectations. Reported EPS is $20.4, expectations were $2.45.

Operator: Hello and thank you for standing by for Baidu’s Third Quarter 2023 Earnings Conference Call. [Operator Instructions] 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, Juan Lin, Baidu’s Director of Investor Relations.

Juan Lin: Hello, everyone and welcome to Baidu’s third 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 newswire services. On the call today, we have Robin Li, our Co-Founder and CEO; Rong Luo, our CFO; and Dou Shen, our EVP in charge of Baidu AI Cloud Group, ACG. 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 provision of the U.S. Private Securities Litigation Reform Act of 1995. Forward-looking statements are subject to risks and uncertainties 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 SEC and Hong Kong Stock Exchange. Baidu does not undertake any obligation to update any forward-looking statements, except as required under auditable law. Our earnings press release and this call include discussions of certain unaudited non-GAAP financial measures. Our press release contains a reconciliation of unaudited non-GAAP measures to the unaudited most directly comparable GAAP measures available on our IR website at ir.baidu.com. As a reminder, this conference is being recorded. In addition, a webcast of this conference call will be available on Baidu’s IR website. I will now turn the call over to our CEO, Robin.

Robin Li: Hello, everyone. Baidu Core delivered solid revenues, profit and cash flow in Q3 despite navigating in a challenging macro climate for both our online marketing business and AI cloud business. I am proud that our team managed to strengthen operational efficiency and maintain stable margins, a full-scale reinvention of our product portfolio with ERNIE and ERNIE Bot. Today, I would like to share an update on the new opportunities that ERNIE and ERNIE Bot have opened up for us. After that, I will discuss some key highlights of each of our businesses. Presently, we are in the midst of a broad-based platform shift driven by generative AI and foundation models that is set to revolutionize every industry. On August 31, we received the approval to deploy ERNIE Bot on a large scale and open ERNIE API to enterprise customers.

Since then, we have witnessed a significant increase in queries handled by ERNIE Bot and through ERNIE API. Moreover, we have received valuable feedback from these users and customers enabling us to further refine our model’s performance. At Baidu World in October, we showcased our progress in ERNIE Bot and AI native products. During that event, we introduced ERNIE 4.0 or EB4, our most advanced foundation model. We believe that EB4 is the GPT-4 level model, displaying human level performance in understanding content generation, complex reasoning and memory retention. These capabilities are crucial for developing AI native applications and solutions. We are pleased to launch EB4 earlier than our expectations. It resulted from our unique end-to-end 4-layer AI infrastructure, which helped intense efficiencies in model training.

The input and feedback from our users and customers also played a big role. In products, as I said in the past, we continue to use ERNIE to reinvent our entire portfolio and introduce an AI-native experience. On the customer front, the ERNIE Bot enables Baidu search by generating direct answers to search queries, complementing traditional search. In the quarter, we initiated tests on our new features that recommend news feed-like information together with the generated search results and enable multi-round conversations to encourage further user expression. Our initial tests have received promising feedback. We believe these features will help deepen user engagement and prolonged time spent unleashing new monetization opportunities. In particular, they benefit as verticals like healthcare, education, travel, legal and auto, in which advertisers are willing to invest heavily in customer acquisition and reengagement.

ERNIE Bot, our new AI native product serves as a versatile multi-brand conversational AI assistant on both desktop and mobile. Given the exceptional performance of our large language model, we are confident in monetizing our services. Starting from November 1, EB4 was open to the public through ERNIE Bot at a subscription fee of about $8 per month. This marks us as the first company in China to implement user charges, distinguish us from other models in the market. Our primary focus is to encourage seamless collaboration between users and AI co-pilots, which we believe is a key trend in the new era. For example, with an AI co-pilot, Baidu Wenku has transformed into a one-stop shop for various document creation needs. We have already seen an increase in the paying user account, a trend that we expect to continue in the coming quarters.

On the enterprise-facing product front, we recently introduced a GBI, generated business intelligence with ERNIE Bot. GBI simplifies data analysis using natural language interaction facilitating faster decision-making for business operation. The introduction of GBI was prompted by the recognition that customers across various industries have the need for AI co-pilots to help them analyze data more efficiently. During our last earnings call, we also discussed how we use ERNIE Bot to create Baidu Comate, our AI coding assistant and Infoflow, our enterprise communication and collaboration platform. These products focus on boosting productivity and efficiency gains and each of them present upsell opportunities for our cloud customers. In fact, an increasing number of our cloud customers in China’s traditional industry and public sector have used the trial version and showing interest in these products and features.

Additionally, these products and features also allow us to acquire new customers across an even wider array of industries. In terms of ecosystem, we empower enterprises to leverage ERNIE through API to create their own AI native applications and solutions that will drive the development of generative AI and LLM. As more and more AI native applications built on top of ERNIE become successful, whether developed by us or by our customers, ERNIE will likewise be successful. Now, over 10,000 enterprises are actively using ERNIE through API on a monthly basis. This number has been growing quickly since we received the regulatory green light at the end of August. Currently, ERNIE is handling tens of millions of queries everyday. Right now, a large and rising number of these queries come from the Baidu family of products as we have been pioneers in building AI native products and have put a lot of effort into reinventing our offerings.

In the first half of November, the number of daily external queries has increased by over 50% compared to the same period in October. As we are actively assisting our cloud customers in creating AI native applications, we believe there will be a continuous and significant rise in external credits in the future. We are also actively attracting developers to connect their information and services to ERNIE Bot through plug-ins. With plug-ins, ERNIE can help people with more-and-more tasks unlocking a wide range of possible use cases. As of today, [indiscernible] plug-ins have already been accessible to ERNIE. The initial batch of third-party plug-ins include ctrip.com, CITIC Press Group, China Justice Big Data Institute, New Oriental, Autohome and Tree Mind.

In summary, during the quarter, we made significant progress in using Gen AI to revolutionize product usage and transform business operations for our users and customers. We believe that this is just the beginning. In the future, we will realign resources to invest in this growth opportunity and shift away from lower priority efforts and improve efficiency for existing businesses, thus balancing investment and margins. We are excited about the possibilities for Baidu for our users, customers, partners and the entire ecosystem. Now let me recap the key highlights of each of our businesses. Mobile ecosystem continued to exhibit steady growth for both user metrics and financial performance in the quarter. Baidu Apps MAUs increased by 5% year-over-year to RMB663 million in September.

Search queries and content distributed by Baidu app remained resilient. In particular, videos distributed by search and feed within Baidu App both experienced a double-digit growth in third quarter. Baidu Core’s online marketing revenue increased by 5% year-over-year in the third quarter, consistently generating strong profit and cash flow for the group. This growth was driven by the continuous recovery in verticals such as health care and travel, among others. In the quarter, we continued to use Gen AI to help advertisers increase ROI and conversion on our platform. Starting from September, advertisers can engage with our new marketing platform. This platform supports natural language input and multi-round conversations which help advertisers, articulate their requirements more comprehensively, enabling us to formulate more effective campaign strategies for investment.

Moreover, we made ongoing enhancements to our monetization system, focusing on improving targeting capabilities and the auction system. For example, Tarena Education and IT Professional Education Company achieved an increase of 23.3% in conversion rate and 22.7% boost in ROI after using this enhanced platform and capabilities. We are still in the early stages of using Gen AI to help advertisers achieve higher of conversions and ROIs on our platform. Our efforts will ultimately lead to significant improvement in marketization capabilities and will contribute to future revenue growth. In the quarter, we also used our AI capabilities to help merchants grow their sales on Baidu. One example is how we help SMEs with live stream shopping. With ERNIE Bot, we introduced a tool that allow them to easily create their own digital human, generate live scripts and more significantly lowering the barrier and cost for live streamers to sell merchandise on Baidu.

Looking forward, we are optimistic that the growth of our online marketing revenue will continue to exceed China’s GDP growth. At the same time, we will continue to test AI native marketing products that could potentially open up more opportunities than traditional general search ads. This gives us confidence in Baidu’s long-term online marketing growth prospects. Turning to AI cloud, we continue to generate positive operating profit on a non-GAAP basis in the quarter as we remained focused on the healthiness of our business. Gen AI and LLM have brought us a lot of opportunities, which have strengthened our competitive advantages in cloud and increased our TAM. A growing number of enterprises are using ERNIE API to develop their own AI native applications and solutions.

We are also helping customers build their own models efficiently by leveraging our unique 4-layer AI infrastructure and our years of experience in building and using foundation models. LLM training is very complicated. It requires a large number of GPUs working simultaneously. ERNIE GPU failures can impact the entire process. We have developed ways to identify and address GPU failures quickly leading to a significant reduction in training costs. Now about 98% of the training time on our platform is valid, setting an industry benchmark. We also have a set of different resources, including toolkits, data sets for enterprise customers to easily fine-tune their customized models. Gen AI has helped us grow our cloud customer base. A large number of cloud customers using ERNIE API are new customers.

At the same time, some of our existing cloud customers have increased their spending with us because of generative AI. AI cloud revenues declined by 2% year-over-year in the third quarter, mainly due to the weak demand in smart transportation projects. We believe AI cloud revenue should rebound to positive growth in the fourth quarter, driven by the increasing momentum in generative AI-related businesses. Also, since smart transportation revenue started to slow down in Q4 of last year, we will have an easier year-over-year comp base in Q4 this year. Moving on to intelligent driving, our target remains unchanged, which is to achieve breakeven on the regional unit economics for robotaxi operation in a couple of years before turning operationally profitable.

To this end, we are strategically concentrating our resources on pivotal regions. Wuhan remains our largest operational area and we believe it is also the largest region globally, providing autonomous drive aiding services currently covering a population of about 2.7 million. In the third quarter, the portion of fully driverless orders within the overall order portfolio in Wuhan exceeded 40%, that’s up from 35% in Q2. We are also particularly pleased to highlight that Apollo Go’s operation in Wuhan continue to expand. In late August, Apollo Go remains the first company in China to provide autonomous ride-hailing services to the general public at Wuhan Kinko International Airport, one of the busiest airports in Central China. The extended reach into the airport transfer involves longer travel distances, presenting an excellent opportunity for the future improvement of unit economics.

All of these developments contributed to the UE improvement and we aim to reach regional UE breakeven in a couple of years. In Q3, Apollo Go provided 821,000 rides to the public, marking a 73% increase year-over-year and the cumulative order volume has surpassed 4.1 million by the end of Q3. As part of our executive reshuffle program, we have recently named Dr. Wang Yunpeng as Corporate VP of Baidu, who will lead the Intelligent Driving Group. Yunpeng has been with us since 2012 and has been responsible for autonomous driving business since 2018. I take great pride in seeing another business leader developed within Baidu. Zhenyu has taken a rotational position as CEO Assistant and the Chairman of the Technology Ethics Committee. Now let’s proceed with Rong’s financial performance review.

Rong Luo: Thank you, Robin. Now let me walk through the details of our third quarter financial results. Total revenue was RMB34.4 billion, increasing 6% year-over-year. Revenue from Baidu Core was RMB26.6 billion, increasing 5% year-over-year. Baidu Core’s online marketing revenue was RMB19.7 billion, increasing 5% year-over-year. Baidu Core’s online marketing revenue was RMB6.9 billion, up 6% year-over-year. Revenue from IT was RMB8 billion, increasing 7% year-over-year. Cost of revenue was RMB16.3 billion, which remained essentially unchanged compared to the same period last year. Operating expenses were RMB11.9 billion, increasing 8% year-over-year primarily due to an increase in China spending, promotional marketing expenses, server depreciation expenses and the server custody fee, which support ERNIE Bot repurchasing costs.

Baidu cost operating expenses were RMB10.5 billion, increasing 10% year-over-year. Baidu Core SG&A expenses were RMB4.8 billion, increasing 14% year-over-year. SG&A accounting for 18% of Baidu’s Core revenue in the quarter compared to 17%, in the same period last year. Baidu Core R&D expenses were RMB5.6 billion, increasing 7% year-over-year. R&D accounting for 21% of Baidu Core revenue in the quarter, which maintained unchanged from the same period last year. Operating income was RMB6.3 billion. Baidu Core operating income was RMB5.5 billion and Baidu Core operating margin was 21%. Non-GAAP operating income was RMB7.6 billion, non-GAAP Baidu core operating income was RMB6.7 billion, and non-GAAP Baidu Core operating margin was 25%. Total other income net was RMB1.9 billion compared to the total other loss net of RMB4.8 billion for the same period last year, mainly due to the first recognition of RMB338 million gain versus RMB3.1 billion loss for the same period last year from a fair value change in loan investments; and second, a decrease in impairment of long-term investments by RMB1.4 billion.

Income tax expenses was RMB1.3 billion, increasing 41% year-over-year, primarily due to an increase in profit before tax. Net income attributable to Baidu was RMB6.7 billion, and diluted earnings per ADS were RMB18.22. Net income attributable to Baidu Core was RMB6.4 billion, and net margin for Baidu Core was 24%. Non-GAAP net income attributable to Baidu was RMB7.3 billion and non-GAAP diluted earnings per ADS were RMB20.40. Non-GAAP net income attributable to Baidu Core was RMB7 billion and Non-GAAP net margin for Baidu Core was 26%. As of September 30, 2023, cash, cash equivalents, restricted cash and short-term investments were RMB202.7 billion and cash, cash equivalent, restricted cash and shorting investments excluding IP was RMB197.4 billion.

Free cash flow was RMB6 billion, and free cash flow, excluding IP, was RMB5.2 billion, Baidu Core had approximately 35,000 employees as of September 30, 2023. With that, operator, let’s now open the call to questions.

Operator: [Operator Instructions] Your first question comes from Alicia with Citi. Please go ahead.

Alicia Yap: Hello. Thank you. Good evening. Robin, Julius and management team. Thanks for taking my questions. My question is on advertising. So it seems like Baidu ad revenue growth is tracking slower than some of the Internet peers, so besides macro can management elaborate any other reasons that contributed to the softer ad revenue growth? And then looking into the fourth quarter, have you seen any demand picking up? What is the e-commerce sector contribution? And how will AI change the advertising outlook. Thank you.

See also 20 Richest Countries that Mainly Speak English and 20 Most Advanced Countries in Education.

Q&A Session

Robin Li: Hi, Alicia, this is Robin. In Q3, apart from the macro weakness, online marketing revenue from e-commerce platforms was also relatively weak. Revenue from e-commerce platforms is one of our top revenue contributors, accounted for about 10% of our total online marketing revenue. Like many other Internet platform companies, we are building our own native e-commerce business. Revenue growth from our native e-commerce business is tracking very strong as we continue to improve the shopping experience on Baidu. I would like to highlight the strides we have made in our ad business through Gen AI. We are basically restructuring the overall ad platform, including creative construction, ad targeting and the bidding investments.

These efforts have started to pay off. And the incremental revenue from this kind of initiatives are expected to reach the level of hundreds of millions RMB in the current quarter, which is the Q4 of this year. And looking forward, we are optimistic that – the growth of our online marketing revenue will continue to exceed China’s GDP growth. Thank you.

Operator: The next question comes from Alex Yao with JPMorgan. Please go ahead.

Alex Yao: Thank you, management for taking my question. I have a few questions on cloud revenue. I believe Robin mentioned despite of moderate revenue decline in 4Q, the cloud revenue will – the growth rate will return to positive territory in Q4. And then from there, should we expect the cloud revenue to further accelerate into first half of 2024. With regard to the Smart City projects, are there any more projects that are still at risk? And then more importantly, as you guys start to monetize the AI capability, when will AI start contribute to the cloud revenue meaningfully. Lastly, any preliminary view on cloud revenue growth outlook for 2024? Thank you.

Dou Shen: Hi, Alex, this is Doug. Thank you for your question. So actually, as we mentioned before, we have been focusing on improving the health of our business for sustainable development. And as a result, we have achieved non-GAAP operating profits in the past few quarters. As Robin already mentioned right due to the weak demand for intelligent transportation and the cloud revenue experienced a slight decline in Q3. So while excluding small transportation, the rest of our AI cloud business showed a pretty solid growth. And we believe AI cloud revenue will return to put growth in the fourth quarter and the trend will continue down the road. What is even more exciting side, we keep seeing new opportunities brought by Generative AI and large language models.

Actually, last quarter, we already said that more and more customers across various sectors came to us for model training, application development and solution enhancement. Although the current revenue from Generative AI and LLM related business is still very small, but it’s growing very fast. We have seen more and more enterprises proactively adopting these new technologies for productivity and efficiency gain. Some of these customers, especially those from the internet education and the tech sectors, they have started to see efficiency gains through working with theirs. As a result, some of them have gradually increased their spending on our cloud services. Looking to Q4, we aim to leverage our leadership in generative AI and large language models to continuously attract new customers and encourage the existing customers to increase their spending on our Baidu AI cloud.

And we believe this should not only lead to long-term revenue growth, but also continuous margin improvement. Thank you, Alex.

Operator: The next question comes from Miranda Zhuang with Bank of America. Please go ahead.

Miranda Zhuang: Thank you, good evening. Thanks, management for taking my question. And congratulations on the results. My question is about ERNIE’s. So can management share with us the recent feedback for ERNIE 4 since the rollout last month? And any color on the contract find on adopting ERNIE 4? And also for consumer and – and how is the feedback after ERNIE Bot started to charge end user subscription fee? And lastly, among the various opportunities you mentioned, which one do you think can become the biggest revenue driver. Thank you.

Robin Li: Hi, Miranda, let me answer your questions. since the release of EB4 in mid-October, we’re receiving positive feedback from both users and customers. Many enterprises have reached out to test EB4 and have been impressed with it’s capabilities. EB4 has gained a reputation for advanced understanding and complex reasoning abilities. Comparing to EB3 5 and other LMs in the market, we have noted that EB4 generates more structured and clearer responses and excels in coding. From November 1, we started charging enterprises and users for using EB4. And we’ve seen a growing number of customers and users willing to pay for its us. We’re proud to be the first company to introduce GPT-4 level model in China. EB4 further widen our lead over other LMs in the market?

And we are the first LM to charge end user fees that sets us apart from other peers. And regarding your question about monetization opportunity, we see significant opportunities in AI native applications, either devised by Baidu or by our customers, who leverage our AI capabilities. If you look at our own products, we see significant opportunities in the new search and the revamped ad platform. The new search complements traditional search. It can address complex questions that were previously unanswerable. It also enables users to conduct a more personalized and in-depth research on various topics and projects. We will soon enable users to have multi-round conversation with us. As our search will be able to talk with users in natural language and in multi-round, it will create more potential on the commercial side, too.

We are experimenting with a chatbot type ad product for SMEs and brand. We believe this will not only help drive effective conversion, but also allow us to eventually transform from a CPC model to a CPS model. And at the same time, our capabilities – our capabilities will help our advertisers to better operate their business on Baidu. Our ongoing effort to revamp the ad platform have already shown positive results, and we will continue leveraging generative AI and to assist advertiser team achieving durable ROI growth on Baidu. In terms of empowering our customers with Gen AI, as Doug just mentioned, customer needs are different now. Some customers still prefer to train their own model, but the GPU export restrictions will put a break on that.

It will eventually become clear that training LLM from scratch is very difficult, especially when trying to achieve emergent abilities. So we will turn to advanced LLMs available on the market like ERNIE for devising applications. As customers become more advanced in using LLM to create applications and more AI native applications powered by ERNIE become more widely used. We should be able to see continuous revenue growth through model inferencing. Over the long-term, inferencing should become a major source of revenue forward. Meanwhile, we will also help customers to fine-tune our existing model offerings to suit their customized needs in each scenario because our models are better, faster and more cost effective. So in summary, Gen AI and LLM will bring us massive business opportunities.

We have already made good progress in commercialization so far, and this is just the beginning of a promising future. Thank you.

Operator: The next question comes from Gary Yu with Morgan Stanley. Please go ahead.

Gary Yu: Hi, thank you, management for the opportunity to ask questions. Can management share the latest advertiser feedback on the AI-powered ad system upgrade, and how do you think about the level of revenue boost to the core adds in 2024 as it gets rolled out to more even all advertisers? Thank you.

Robin Li: Yes, we are very happy with the rapid AI transformation of our ad system and thrilled by the positive feedback from our advertisers. Overall, I think advertisers appreciate our efforts to help them improve their ROIs on our platform. Also they are fond of our new features, which help them to be more productive. As I mentioned in the prepared remarks, we’ve put a lot of effort into using Gen AI and our aims to reinvent our app system over the past few quarters. Now we have an integrated advertiser-facing marketing platform. Advertisers can use it to generate creative advertising materials. These materials have proved to deliver higher conversion rates than materials created by humans. And our platform also allows advertisers to engage in natural language by interacting with advertisers in multiple rounds of conversation, our upgraded platform is able to better understand their intentions.

This allows us to create campaign strategies that deliver higher ROI. Moreover, it significantly reduces the time that ad managers need to spend creating campaigns, because even an experienced ad manager have to spend hours developing an advertising strategy. And now with an AI copilot, the process only takes like a few minutes. As of today, we have a few thousand of advertisers already migrated to our new platform. While this number is relatively small compared to our 0.5 million advertiser base, it is certainly growing very fast. In addition, we continue to use AI to improve our bidding system and ad targeting capabilities. Such initiatives happened at the back end of the system, so advertisers may not directly perceive them, but they have observed improvements in their ad conversion and ROI.

So, on average advertisers using these capabilities probably achieved a high-single digit increase in conversions in Q3. All of our efforts should eventually attract advertisers to allocate more of their advertising budgets on Baidu. As I previously mentioned, online marketing revenue related to our upgraded ad platform has been growing rapidly and already became meaningful. And this is just the beginning, as I have said, we are experimenting with AI chatbot and chatbot could act as a replacement for landing page in the future. This is particularly useful in verticals where users typically research and engage in a long decision process before purchasing. Imagine searching for a training program and being directed to a bot instead of a web page, with our AI chatbots, users can quickly learn about brand-specific information, product details and other stuff.

Although, I think that multi-round conversation is also very engaging. And for advertisers, it helps them to stay connected with potential customers and guide them at key decision-making points, leading to better conversion and even direct sales. We believe AI chatbot could work well with Gen AI powered new search and bring us more opportunities. I mentioned earlier that the incremental revenue from these kind of initiatives are expected to reach the level of hundreds of millions of RMB this current quarter, Q4, and this is certainly growing very fast. The trend should continue and further strengthen in the year of 2024. Thank you.

Operator: The next question comes from Wei Xiong with UBS. Please go ahead.

Wei Xiong: Good evening management and thank you for taking my question. I wanted to follow-up on your cloud business. So, could management share more color on how to think about the industry competitive landscape in China cloud market, especially among the Internet cloud vendors and as well as when competing with the telcos and also with the development in generative AI and the large language model, how do we assess our competitive advantage against peers? Do we expect the competition to intensify next year as other companies try to make efforts and catch up with us? Thank you.

Dou Shen: Great question Wei Xiong. As you actually already noticed, right, the traditional cloud business is slowing down, while generative AI and large language models are sliding in and reshaping the competitive landscape of the cloud business industry. In the past, right, the focus in the cloud market was actually on house, which is like a commodity and people are competing on pricing. But now with the rise of generative AI and the large language models, things are changing. There is a growing interest among cloud customers coming to Baidu to utilize these sophisticated technologies to increase productivity and efficiency. They came to us not only because we have the most advanced AI technology, but also because we have experience and track record in using AI to help enterprises to solve problems.

There are some of them who are still in the product experimental stages, but we have firm belief in the new technology to rebuild their products and services, because we have seen successful stories in overseas. That is why we are seeing the new technology is increasing our time and expanding our competitive edge. So, as Robin mentioned, EB4 is China’s first GPT-4 style model. We also shared the positive initial feedbacks we had garnered for EB4. So, currently our team are engaged in dialogues with our clients, assisting them in understanding the technology and utilizing ERNIE to redevelop their existing products and create new ones. So, I can say that ERNIE has already helped us attract new customers and additional IT spending from existing customers.

So, here I would like to briefly bring up two points for our advantage compared to the players in the market. The first one is our unique four-layer AI infrastructure, which gives us the flexibility to make adjustments or innovations at every layer to be able to be compatible with other leaders to keep driving efficiency in both model training and inference. And the second one, more specifically, is our capability to develop GPU networks for clusters for large language model training. So, as Robin just said, 98% of the training time on our AI infrastructure is valid, as a result in our customers, including several leading Internet and tech companies for increasing their investments in our service. Furthermore, we will continue to leverage our unique advantage of AI architecture to drive efficiency gains.

It will help us to greatly reduce costs in our model training and inference on our cloud, giving us the flexibility to offer more compelling prices to our customers and further strengthen our competitive edge in the market. Regarding the competition from telecom operators, I would like to highlight our focus is on different market segments since we differentiate ourselves with our AI capabilities, particularly earnings as we have mentioned. At least worth noting that we can cooperate and we are actually cooperating on many projects and as for other Internet companies in China, so our strong AI capabilities and ERNIE are well recognized by the market, which will set us apart from our peers. To sum up here, we believe that our strong AI capabilities and particularly in generative AI and large language models will allow us to eventually become the end market leader and gain share in the cloud market.

Operator: The next question comes from Lincoln Kong with Goldman Sachs. Please go ahead.

Lincoln Kong: Thank you, management for taking my question. So, my question is also about ERNIE, so given the successful upgrade of ERNIE 4.0, what would be the future strategy for model iteration to certify our tech leadership? So, do we foresee any competition in the foundation model in the industry, either to stabilize or intensify in the future? Thank you.

Robin Li: Hi Lincoln, the AI chips we have in hand already allow us to launch EB4. We are ahead of the competition. To take our lead in LLMs to the next level, we will take an application-driven approach. We will let the AI native apps tell us what to improve in ERNIE Bot capabilities. Given that there are only a very limited number of AI native apps on the market right now, the majority of ERNIE API cost from internal apps, our internal apps like search apps, Wenku, etcetera. The rebuilding and restructuring of our existing products drive ERNIE innovation in the right direction. What is equally important is that we are helping enterprises use ERNIE to build their offerings. And we have seen that over 10,000 enterprises are using ERNIE through API costs on a monthly basis, which propels ERNIE’s improvement too.

We also continue to improve the efficiency of our models. For example, compared to ERNIE Bot version in March, interest cost of the current version has been reduced by 98%, basically resulting in a 50x increase in QPS for the same amount of hardware activated power. We are also able to do this using our unique four-layer architecture and leveraging our ability to do end-to-end optimization. Continued inference cost reduction has further strengthened our model’s competitive advantage, and it gives us the flexibility to offer more and more compelling presence. From a long-term perspective, taking into account factors such as the scarcity of high-performance chips, high demand for data, AI talent and the huge upfront investments, the industry will soon translate into a consolidation stage.

We believe there will only be a select few foundation models in the market, and Baidu will certainly be one of them. In this stage of industry development, more and more enterprises will begin to leverage advanced foundation models like ERNIE to create AI products rather than spending resources on building their own large language models. So, we expect that the number of native apps based on ERNIE will reach millions in the future.

Lincoln Kong: Thank you.

Operator: The next question comes from James Lee with Mizuho. Please go ahead.

James Lee: Great. Thanks for taking my questions. Can you guys maybe quantify the investments related to AI and how that affects various cost items in your P&L, and should we expect these investments to accelerate over the next few quarters. I was thinking especially at the launch of ERNIE 4.0 and potentially higher inference costs as multiple using it. And then if we extrapolate that over a longer term, how should we think about Baidu core OPM over the next few years, given all the moving parts, including revenue shift, investment AI and also your continued improvement in cloud profitability. Thanks.

Rong Luo: Hi James. Let me take your questions. This is Julius. Currently, the primary investments for generative AI and large language models is centered around the computing power, which is recorded as part of CapEx. I think in the past few quarters, we have put a lot of chips resources in training our new ERNIE models. And in the future, as more AI native applications, which is powered by our ERNIE can more widely used, which also more will put resources in small [indiscernible]. However, please note that all of this impact for the AI-related our investments all came out is quite manageable because all the hardware depreciation are spread on over a few years. And for example, all expenses linked to the coating power, we used in the training ERNIE are recorded through IND depreciations.

And the model inferencing courses, which is highly related to the usage of the model either internally or externally, and it should be supporting the funding by the future developments. And moreover, we are happy to see that our investment in generative AI and large language models are beginning to bear fruit. As Robin has mentioned just earlier, since we are receiving the approval from regulatory and more additional revenue generated from earnings powered by 2C or 2B businesses has been growing quite fast. While we are using the generative AI and large language models to renovate our businesses, we are still keeping a close eye on making sure our Baidu cost earnings stay solid. In Q3, we can see that mobile ecosystem continues its high margin, ensuring a very strong generation of cash flow.

And AI cloud business continued its healthy growth and achieved profitability one more time. And look ahead, we expect the traditional cloud business to remain quite profitable and the new opportunity arising from generative AI and language models are also expect to have favorable margin in the long-term. For intelligent driving businesses, our long-term growth opportunity will continue to invest with a mature pace. And all-in-all, we will concentrate our resources by reallocating them from non-core businesses to AI-related businesses. All of these were quite beneficial for our long-term growth. Thank you, James.

Operator: The next question comes from Thomas Chong with Jefferies. Please go ahead.

Thomas Chong: Hi. Good evening. Thanks management for taking my question and congratulations on a solid set of results. My question is on the chips side. Can management comment about the impact on AI development after the further restriction of the chip export from the U.S.? How does that affect our AI product offerings and user experience, if any? Thank you.

Robin Li: Yes, the restrictions on the chip export to China actually have limited impact on Baidu in the near-term. We have successfully launched EB4 in mid-October, our most advanced foundation model in China. It is a milestone for us. And as I have just said earlier, we have a substantial reserve of AI chips, which can help us keep improving ERNIE Bot for the next year or 2 years. Also inference requires less powerful chips, and we believe our chip reserves as well as other alternatives will be sufficient to support a lot of AI native apps for the end users. And in the long run, having difficulties in acquiring the most advanced chips inevitably impacts the pace of AI divestment in China. So, we are proactively seeking alternatives.

While these options are not as advanced as the best chips in the U.S., our unique four-layer AI architecture and strength in AI algorithm will continue to help us improve efficiency and mitigate some of these challenges. For example, we have made some innovations in [indiscernible], our deep learning framework and ERNIE Bot foundation model to allow them to be better compatible with different parts of AI chips, both model training and inference tasks. But given that all the other Chinese companies face the same challenge, we believe we are actually best positioned to service market. As you probably know, in the past, some of our peers, they tried to write on the Gen AI wave by investing in those startups to train foundation models. And they – basically we sell the computing power to those start-ups.

We didn’t do that. We tried to optimize everything from the infrastructure layer to stream of [ph] layer and to model layer into apps. So, we have invested in this kind of end-to-end optimization approach. Therefore, we can – for the same amount of computing power, we can do training more efficiently, more cost effectively and we can do inferencing faster and cheaper. And as time pass by, I think more and more companies will realize that they don’t need to train their foundation models. They just need to develop AI native apps based on Baidu’s foundation model, which is the best on the market. So, I am really happy that we basically invested on this kind of end-to-end optimization front for many, many years. And it’s time for us to show that the investment is worth it.

Thank you.

Operator: Ladies and gentlemen, that does conclude our conference for today. Thank you for participating. You may all disconnect.