Hello, and thank you for standing by for Baidu's third quarter 2017 earnings conference call. At this time, all participants are in a listen-only mode. After management's 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, Sharon Ng, Baidu's Director of Investor Relations.
Hello, everyone, and welcome to Baidu's third quarter 2017 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. Today you will hear from Robin Li, Baidu's Chief Executive Officer, Qi Lu, Baidu's Chief Operating Officer, and Herman Yu, Baidu's Chief Financial Officer. After their prepared remarks, Robin, Qi, and Herman will answer your questions. Before we continue, please note that the discussion today contains forward-looking statements made under the Safe Harbor provisions 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. Potential risks and uncertainties include, but are not limited to, those outlined in our public filings with the SEC, including our annual report on Form 20-F.
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 unaudited non-GAAP financial measures. Our press release contains a reconciliation of the unaudited non-GAAP measures to the unaudited most directly comparable GAAP measure and is 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 also be made available on Baidu's IR website. I will now turn the call over to Baidu CEO, Robin Li.
Hello, everybody, and thank you for joining today's call. We delivered solid financial results and made meaningful progress in this quarter on Baidu's strategic pillars, strengthening our mobile foundation and leading in AI. On strengthening our mobile foundation, Baidu's core business continue to build momentum from our twin engine mobile search and feed. In mobile search, Baidu is making good progress in product innovation that enables native access to rich content such as video and content within top mobile apps. Our strategy to leverage Baidu's leading AI technology to improve the algorithmic recommendation of our feed is showing measurable results in improving user experience. In September, the DAU over MAU of our flagship app, Baidu App, increased by two percentage points year-over-year.
Daily user time spent on Mobile Baidu grew 15% sequentially in the third quarter, the fastest among apps in China with DAUs over 100 minutes, according to QuestMobile. We will continue to work on increasing the user scale and improving user stickiness of Mobile Baidu by leveraging Baidu's leading AI technology coupled with user experience data based on the large array of Baidu apps that they use. This strategy will allow us to continuously improve user experience on Mobile Baidu, which ultimately is the key to growing market share. The natural result of our feeds increasing user scale and user time spent is better monetization. For the third quarter, feed revenue on an annualized basis exceeded $1 billion. This is an impressive milestone as our Feed product was officially launched about a year ago. Turning to Baidu's second strategic pillar, leading in AI.
We are making measurable progress across our initiatives on Apollo, DuerOS, ABC Cloud, and FSG. In September, we launched Apollo 1.5, an upgrade of Baidu's open-source autonomous driving platform, and created a multi-billion dollar fund to support the growth of Apollo's ecosystem. Last week, we announced a major milestone, forming strategic cooperation with two leading automotive companies in China, BAIC Group and King Long, which will help make Apollo an industry standard. On DuerOS, Baidu's voice-assisted operating system for smart devices, we are growing industry partners and increasing adoption through the expansion of DuerOS capabilities. For example, in the third quarter, we launched DuerOS intelligent device platform and DuerOS Bot Platform for third-party developers and launched our DuerOS global partner plan. ABC Cloud released new products and is expanding its industry solution to enterprise customers across more sectors.
Cloud is a key component of our voice-assisted and autonomous driving OS strategy. We are leveraging this foundation to extend cloud services to targeted industries as well as internally to companies such as iQIYI. FSG is making impressive progress leveraging Baidu AI and big data to access credit risk and customize financial products based on our user profile. At Baidu, we serve hundreds of millions of Chinese users every day through an array of over 20 products and apps. We place a high priority on continuously improving user experience and the trustworthiness of information and services found on our platform. At the end of September, we introduced a screening system to limit the distribution of fake news through search results. The system utilizes probably China's largest centralized database to combat fake news.
Employing Baidu AI and the network support of over 600 organizations to identify and verify false information and rumors. In addition, as a follow-up to the additional customer verification requirements that we added last year to improve ad quality, we are taking proactive steps to better gather, verify, and rank information from the healthcare and medical industry based on source credibility, quality, and relevance. We are enlisting the participation of the healthcare community, including top health and medical organizations and industry experts, to work with Baidu toward a long-term solution for better health information. On November 16th, we will be holding our 12th annual Baidu World event in Beijing. We look forward to showcasing Baidu's AI products and hope to see many of you there.
Last but not least, I'd like to extend a warm welcome to Herman Yu, who joined us a little over a month ago as Baidu's new CFO. I trust many of you know Herman from his last 10 years as CFO, first at Sina, then at Weibo. Herman joins Baidu at an opportune time as we are transforming into an AI-first company by strengthening our mobile foundation while developing technology platforms for two potentially huge markets, voice-assisted smart devices and autonomous driving. Together with the hiring of Qi, our COO, at the beginning of the year, and the reorganization of management over the last year, we believe these changes are increasing the depth and breadth of Baidu's leadership and allowing us to be more focused on what we do best as a technology company. I'll now turn the call over to Qi to go through our business progress.
Thank you, Robin. On our core business front, we remain focused on improving the user experience through product innovation for Search, Feed, Mobile Baidu, and IGE. AI is the underlying technology foundation that enables us to systematically drive product innovation in a number of areas. A big priority for us is to improve user experience so that the overall search experience is much more mobile-native. For example, our Top One initiative enables a faster and more immediate one-and-done search experience. It intelligently optimizes the top search result, and it has gained strong traction. In Q3, Top One results has grown to cover over 20% of Search page views. By applying intelligent filtering and advanced presentation technologies, the percentage of Search landing pages that provide a good mobile native experience has now exceeded 90%, which was 40% at the beginning of 2017.
At the same time, we continue to expand Search's coverage of richer content, especially video, and the percentage of video content within Search result pages continue to grow. Through deep link and share link, Baidu Search now can access high-quality content embedded within top apps. AI also enabled voice and image as search input modalities and are making sustainable progress in providing our users with a more natural and overall more compelling search experience. Feed, AI technology plays an even more potent role in product and user experience and our content ecosystem. First, we continue to strengthen the core foundation of Feed, which is a very high scale personalized information recommendation. Our core algorithms are iterating and improving at a rapid pace, leveraging Baidu's strengths in AI algorithmic capabilities and Baidu's data assets such as search data.
As a result, the core product metrics such as click-through rates and user satisfactions continued a healthy pace of improvement. At the same time, we significantly enriched the content in our Feed product. Video and short videos consumptions on the Baidu platform has grown very rapidly, as well as other key areas such as user-generated content, question and answers, new verticals such as sports, and highly engaging content areas such as animations and literature. Baijiahao is the publishing platform of a content ecosystem. We made strides in enabling more quality publishers and the better content productions on our platform over the past quarter. The ratio of clicked on content to displayed headlines has grown at a very healthy upward trend, doubling from one year ago. Overall, the user base and the usage of Feed product continue to grow at a strong pace.
Mobile Baidu is the flagship product for strengthening our mobile foundation and has the strong twin engine of search and the feed to propel its growth. We continue to make good progress across a number of areas, such as the synergies between search and the feed, the speed and the richness of the user interactions, such as voice output, content social sharing, mobile bandwidth discount offerings. We believe Mobile Baidu has the foundation and what it takes to build out the industry's leading mobile information content container and the corresponding ecosystem over time. We will share more about our product vision and the future plans about Mobile Baidu at the upcoming Baidu World Conference. On the customer front, with our twin engine of search and the feed, we're able to deliver a more comprehensive, richer, and a diversified offering to fulfill customer needs.
For example, we helped Chrysler Jeep with a unified campaign to expand their presence in China across search, Brand Zone, and the feed. The campaign was highly successful, exceeding the expectations for lead scales and ROI. With Baidu's robust ad platform and technology and data capabilities, we were able to provide Chrysler Jeep with targeted high-quality leads that resulted in better conversion rates compared with leads from other ad platforms. We continue to use AI and big data to systematically drive improvements on our advertising platform across a number of areas. For example, this past quarter, we launched dynamic ads in feed, which tightly integrates consumer shopping data with our feed ads. Dynamic ads in feed complement our dynamic ads in search. Dynamic ads show user products that are most relevant to them based on past actions, such as browsing or adding products in a checkout basket.
Customers share their data with us, which dramatically improves the relevancy and the level of targeting, resulting in a step function increase in ROI and a higher ad spend with Baidu. We're working closely with a select number of customers in e-commerce and the travel verticals, including JD and Ctrip. JD has been a longstanding customer using dynamic ads across search and the feed. Since the adoption of dynamic ads in feed in early July, JD saw ROI from dynamic ads in feed more than doubled, and JD meaningfully increased overall ad spend with Baidu. We plan to roll out dynamic ads to even more customers, such as those in auto, travel, and the real estate sectors. Turning now to iQIYI. iQIYI has continued to ride the secular growth of video and entertainment.
In Q3, iQIYI again showcased its industry leadership with The Rap of China's first rap reality show. The Rap of China was self-produced by iQIYI and garnered over 2.7 billion views in the first season. Over the years, iQIYI has invested to grow its library of self-produced content and has established itself as an IP platform with premium content. This strategy has differentiated iQIYI from its peers and underpins iQIYI's leadership and growth. In August, according to our research, iQIYI's PC and mobile app daily active users reached 78 million and 160 million respectively. Again, reaching new highs. Monthly time spent in August grew strongly, with users spending 461 billion minutes on the mobile app, 29.5% higher than August last year. In addition, we are leveraging the operating synergy between Baidu and iQIYI through content integration and data cooperation.
For example, iQIYI is an exclusive content partner for Baidu Feed. iQIYI's video feed uses Baidu's platform and the back-end technologies. Next, I'll talk about our progress on our AI-enabled new businesses. Our Apollo and our autonomous vehicle business. We made strong progress in Q3. Last month, we released Apollo 1.5, which opens up a number of important services such as HD map and a simulation engine, which will be key to drive developer productivity. A number of key capabilities, such as map engine, perception, planning and control, and lidar reference implementation. Apollo 1.5 has been very positively received by our partners and developers. Overall, in Q3, the number of Apollo ecosystem partners has grown from 54 to 74, with more OEMs, component and solution providers, as well as startups joining the Apollo ecosystem.
We reached strategic agreements with OEMs such as BAIC Group, King Long, one of China's largest bus manufacturers, and Chery, and Jianghuai, for commercial L3 productions in 2018 for buses and 2019 for cars, as well as L4 production by 2021. This month, we began a strategic collaboration with Shouqi, one of China's leading car-hailing platforms, and we will work with Shouqi to integrate Apollo technology and DuerOS across their fleet. We have also began putting Apollo funds capital to work, and have made investments in areas including hardware, software, and transportation services. On DuerOS, our conversational AI platform, we continue to improve our technology and the platform and strengthen our leadership position in this important emerging ecosystem because voice and natural language interface will become the future universal gateway to the digital world.
According to IDC, 27% of homes and 51% of cars will be voice-powered by 2020. In Q3, we launched our DuerOS intelligent device platform and DuerOS Bot Platform for third-party developers and our DuerOS global partner plan. We continue to expand the DuerOS market footprint to power more third-party devices, including select models of TCL and Lenovo television, Vivo and Xiaomi smartphones, Goertek and television set top boxes, and connected cars dashboard in select models of the First Automotive Works, Hyundai, Chery cars, and many more. Our ABC, our AI and big data powered cloud business, we held our 2017 ABC summit last month and launched a number of new products such as the ABC Stack, which integrates a broad suite of 60 AI technologies and nine open source big data technologies and powerful infrastructure services.
Another key offering is ABC all-in-one hardware, which can be deployed in either hybrid or private cloud, enabling our enterprise customers to implement end-to-end applications with one-stop services. We continue to expand vertical solutions across manufacturing, media, marketing, financial services, energy, and logistics, and secure key customers such as Shanghai Pudong Development Bank. On FSG, our AI-enabled financial services business, we continue to have healthy growth in our core offering of wealth management and consumer credit. At the same time, we're making strides in building out a broad suite of fintech capabilities and a comprehensive fintech platform, with over 300 financial institutions sharing data and product into our platform. Since last quarter, we have put in motion a plan to provide the FSG business with an operating structure that allows FSG to operate more independently, enable stronger long-term growth.
At this point, we do not have a specific timetable and a structure to share. Overall, we're pleased with our operating focus and the business progress. We look forward to continuing our momentum and sharing our progress with you as we continue to move forward. With that, I will turn the call over to Herman to go through the financials.
Thanks, Qi. Hello, everyone. It's a great pleasure to join Baidu and participate on my first earning call. Baidu is an enviable tech company. Its search business delivers an operating margin of approximately 50%, then invests another 30% or so of its revenues back to strengthen its mobile and be in AI. In the third quarter, Baidu shed its delivery business as part of a year-long strategy to free up company resources and management attention to focus on AI and projects that have strong synergy. In addition to investing in mobile in areas such as Mobile Baidu and iQIYI, we are also putting significant efforts to develop AI-based platforms in the voice assistant autonomous driving markets. Like the rest of the world, smartphone sales in China over the last few years have been flattish. We are seeing heavy investments moving towards smart devices for homes and autonomous driving.
These trends present a unique opportunity for Baidu to offer operating systems for smart devices and autonomous driving vehicles. On a global scale, our peers in the voice assist and autonomous driving spaces are formidable. Baidu has been developing AI technology for the last five to six years, including voice assist in Chinese. Over the past year, we have been very focused on developing a strong network of partnerships with local companies in areas such as home electronics and auto manufacturers to adopt our technologies. We believe this is what it takes for an AI platform player to be successful in China. A big part of Baidu's investments today is in the new market opportunity presented by AI that will not have material revenues in the near future.
We believe our investments to become a platform player for smart devices and autonomous driving will bear significant return in the future, as this strategy plays to Baidu's core strength. We must exercise due care to avoid assigning a negative value to Baidu's enviable AI investments based on simple PE multiple or deviation thereof. Let me now turn to our financials. All monetary amounts are in renminbi unless stated otherwise. For the third quarter, Baidu's total revenues reached RMB 23.5 billion, representing a 29% increase from the same period last year. Online marketing services revenue were RMB 21.1 billion, representing an increase of 22% year-over-year. Online marketing customers reached approximately 480,000 this quarter, down 7% year-over-year and up 3% quarter-over-quarter, as our business steadily recovers.
Revenue per online marketing customer was RMB 40,900, showing a healthy increase of 31% year-over-year and 9% quarter-over-quarter. Two key drivers of our online marketing services revenue growth were feed and iQIYI, which we are quite excited about. From ad offering standpoint, Qi talked about the power of Baidu's dynamic ads leveraging customer data to increase ad efficiency and effectiveness. In addition, we experimented with AI ads. For example, Mercedes-Benz ran an ad using Baidu's facial recognition technologies to help potential customers select the most suitable Mercedes car model. Baidu's other AI ad formats include using artificial reality in the campaign where a mobile user can take a selfie, insert him or herself in the ad. The rich user experience that these AI brand ads create will generate new opportunities for Baidu to grow our display ads, especially as we venture into the feed business.
This is another example that Baidu's focus on AI is creating synergy across our platforms, from mobile to voice assist to autonomous driving. Three platforms weaved together through Baidu's leading AI. Turning to other service revenue. Other service revenue was RMB 3.4 billion, up 92% year-over-year, primarily driven by FSG and iQIYI membership. Let me quickly go over Q3 cost of revenue components. Traffic acquisition costs were RMB 2.5 billion, decreasing 5% year-over-year and representing 10.5% of our total revenues compared to 14.2% in the corresponding period last year. Bandwidth costs were RMB 1.4 billion, increasing 16% year-over-year, representing 6.1% of total revenues compared to 6.8% last year. Depreciation costs were RMB 852 million, increasing 6% year-over-year, representing 3.6% of total revenues compared to 4.4% last year.
Operational costs were RMB 1.4 billion, increasing 19% year-over-year, representing 5.9% of total revenues compared to 6.3% last year. Content costs were RMB 3.9 billion, increasing 7% year-over-year, representing 16.6% of total revenues compared to 12.1% last year. The year-over-year increase in content cost was mainly due to iQIYI's acquiring of more hot series content. Turning to Q3 operating expenses. SG&A expenses were RMB 3.7 billion, increasing 4% year-over-year. R&D expenses were RMB 3.2 billion, increasing 24% year-over-year. The increase in R&D expenses were primarily due to the growth of R&D personnel costs. Share-based compensation, which were allocated to related operating expenses and expense line items were RMB 841 million compared to RMB 418 million last year. The year-over-year increase was mainly due to additional share grants to retain talent and stay competitive in the high-tech industry.
Operating profit was RMB 4.7 billion, representing a 69% increase from the corresponding period in 2016. Non-GAAP operating profit reached RMB 5.5 billion, increasing 73% year-over-year. Other income net was RMB 4.2 billion, which mainly consists of investment gain recognized as a result of the disposal of Baidu Delivery. Income tax expense was RMB 1.1 billion for the third quarter. The effective tax rate for third quarter was 12% compared to 25% in the same period last year. The decrease in the effective tax rate was primarily due to a non-taxable investment gain arising from the disposal of Baidu Delivery. Net income attributed to Baidu for the third quarter was RMB 7.9 billion, increasing 156% year-over-year. Diluted earnings per ADS was RMB 24. Non-GAAP net income attributed by Baidu was RMB 9.1 billion, increasing 163% year-over-year. Non-GAAP diluted EPS was RMB 26. Adjusted EBITDA reached RMB 6.8 billion, increasing 62% year-over-year.
As of September 30th, the company had cash equivalents, and short-term investments of RMB 108.1 billion. Net operating cash flow and capital expenditure for the third quarter were RMB 9.7 billion and RMB 1.3 billion respectively. Total headcount at the end of the third quarter was approximately 39,800, representing a decrease of 6% from the end of last quarter. Turning to fourth quarter 2017 guidance. We currently expect total revenues for the fourth quarter to be between RMB 22.23 billion and RMB 23.41 billion, representing a year-over-year growth rate of 22.9%, 22%-29%. Based on our Q3 revenue run rate, Baidu Delivery would have generated approximately RMB 400 million in revenues. As mentioned above, we disposed of Baidu Delivery in August and will no longer have such revenue stream going forward.
Excluding disposed businesses such as mobile games and Baidu Delivery, we expect our revenue in the fourth quarter to grow between 28%-34%. Also built into our fourth quarter guidance is the assumption that iQIYI revenues will drop roughly RMB 700 million sequentially, primarily as a result of limited hot series being featured in China in the recent period. Over the past month, television stations in China have been broadcasting programs featuring the 19th Party Congress meetings and related content in China. To show our respect, iQIYI adopted a self-restriction to not show hot series during this time. iQIYI has been a robust driver of Baidu's revenue growth historically. Year-to-date, it has been consistently growing over 50% year-over-year each quarter. We expect this to slow down to less than 40% in the fourth quarter.
Excluding revenues from iQIYI and disposed businesses, we're expecting Baidu revenues in the fourth quarter to grow between 25%-34% year-over-year, and this compares to 26% year-over-year in the third quarter. In other words, we continue to see healthy growth for our core business. This reflects our current and preliminary view and is subject to change. I will now open the call to questions.
Thank you. Ladies and gentlemen, we will now begin the question and answer session. If you wish to ask a question, please press *1 on your telephone and wait for your name to be announced. If you wish to cancel your request, please press the pound or hash key. In order to be fair to all callers who wish to ask questions, we'll take one question at a time from each caller. If you have more than one question, please request to join the question queue again after your first question has been addressed. Our first question today comes from the line of Chi Tsang from HSBC. Please ask your question.
Good morning. Thanks very much for taking my question. I wanted to ask you about Baidu App. I guess my first question is maybe a housekeeping question. Did you guys disclose how many DAUs you have for News Feed currently? Secondly, I was wondering if you can sort of give us a little bit more commentary about engagement. I'm wondering sort of how many times people are opening up the Baidu App per day. Then sort of related to that, the number one news app in China has about 70-80 minutes of use per day. I'm wondering if that is a figure that you think Baidu App can reach at some point in the future. Thank you.
Let me take the first cut at it, Robin and Herman feel free to chime in. For Baidu App, as we talked in the past, we view this as the flagship product for the mobile era information content containers. Baidu is in a unique position to drive sustained improvement of growth on user base DAU and user engagement, such as time spent. The fundamental reason is we have the twin engine of feed and search. Feed, through the fundamental power of AI technology and more data, will be able to understand the user's interests, user's needs in increasing degree. The product will really know you and provide services that are timely and of high value to you. At the same time, we always know for our industry's last almost 20 years, search is a fundamental need for our users.
We always have information content service knowledge needs to access through search. By harnessing the synergy between these two fundamental trends and by those fundamental strengths, we believe we will be able to drive sustained improvements of product through better, more content in our ecosystem, overall better experience, richer modality, and most importantly, the twin engine of personalized recommendation and search. Baidu is in a unique position more than any other players in China market to sustainably drive a high-paced improvement on user base growth and engagement.
Yeah, this is Robin. Let me add that we basically view Baidu App as one product, including the search feature and the feed feature. Feed is highly personalized. We distribute feed content based on a lot of understandings about our users
Especially user queries that represent their true interest. We basically view feed and search as two major features of this one product, and we are able to provide a much more personalized information distribution channel for our users. We believe this still has a lot of potential to go. Our recommendation algorithms continue to evolve on almost a daily basis. Users who use our feed product will be able to see progress going forward constantly. I think there are still a lot of room for improvement on this front.
Our next question today comes from the line of Alicia Yap from Citigroup. Please ask your question.
Hi. Good morning Robin, Dr. Lu, Herman, and Sharon. Thanks for taking my call. Congratulations to Herman on your new role. My question is related to the traffic acquisition cost versus the SG&A spend. It seems like the TAC continued to decline on the year-over-year basis. Could we assume that this is mainly driven by overall slower growth in the search specific traffic from the union members, while the organic mobile Baidu App traffic improving as the user engagement in Baidu Feed is leading to more organic search traffic in return? It seems that the step up in SG&A this quarter is mainly related to more aggressive promotions in Baidu App installation. Any color as to how we should think about the SG&A line for the fourth quarter as well as the TAC for the fourth quarter and some color into 2018? Thank you.
Okay, great. Hi, Alicia. Two items. First on the TAC, I think the factor that you mentioned, I think that it's correct. Part of the reason is the fact that part of our TAC is also coming from Baidu App, and that we're seeing good growth with overall Baidu App. I think secondly, on the SG&A within the SG&A expense this quarter, we had bad debt allowances for Baidu Delivery and also for Nuomi that amount to over CNY 522 million. That was part of the reason for the increase. When you back that out, actually looking at marketing expenses is not as significant.
Our next question today comes to the line of Gregory Zhao from Barclays. Please ask your question.
Hello, Robin. Hello, Dr. Lu, Herman, and Sharon. Thanks for taking my question. Congratulations to Herman on the new role. I have two very quick questions. The first one is about our core search revenue. If we compare our segment revenue growth trend with our Q4 total revenue growth guidance. Our newsfeed is very strong. iQIYI is very strong. If we backfill to get the search revenue growth. The search revenue for Q4 looks a little bit soft. Do we have any special reason behind? That's my first question. The second question is about our newsfeed product. In terms of revenue, the newsfeed is growing very fast and based on Robin's disclosures, it's catching up with our competitors such as Weibo and Toutiao.
In terms of our content generation capability and the user engagement, Weibo's social features and Toutiao strategies such as video, I think they have some special selling point. Compared to the Weibo, compared to Toutiao, what's our position into the market and what's our strategy to compete with such kind of competitors? Thank you.
Herman will answer your first question, and Qi will answer. Hi, Gregory. On your question of search, whether it's soft in the fourth quarter, as I mentioned earlier that the guidance that we're giving when you're looking at Baidu core business that excludes our spin-off business and iQIYI, we're actually seeing the range that we give you, most of it is above Q3. I don't actually see that. I think your point is can we break out the difference between feed and search? As Robin mentioned earlier, when we look at, for example, the traffic in Baidu App, we actually don't see a user and try to break out the time they're spending on search versus on feed.
We see it as just one product, as long as the user is getting good experience in that product and we're generating revenue, whether it's in search or whether it's in the feed, we see that as kind of one container and how do you optimize the time spent and also the monetization for that one app. I wouldn't break that out separately. The way I would just see it is that a lot of properties, including Baidu App, where you have the different experiences in one, when we, for example, do marketing and so forth, we look at it as how much can we get from this particular app versus how much can we get it from different functionality of that particular app
Let me answer the second question, also add to what Herman said. We fundamentally view Baidu App, the combination of search and feed as one unified product, we have what it takes to build the mobile era leading containers and the starting point for our users. With regard to competition, specifically against, you mentioned Toutiao and Weibo. There are several fronts we believe Baidu has a strong strength in our capability to compete extremely well and grow up in this at a very rapid pace. First is the fundamental strengths of algorithm and data. If you look across our competitors, Baidu has the strongest AI algorithmic capability in data because our search data encompass so much user signals about the user interest and the user needs. That's number 1. Number 2, on growing the richness of content offering. We mentioned videos.
Our video product, not only videos displayed within search, within feeds, also we have a separate video apps. They're all growing at a very rapid pace. You will see us investing more and growing more innovations in enriching the content in our offerings. Third front is monetization. Our ad platform has fundamental strengths. We have vast scales of customer base, we have rich algorithmic capabilities and datasets leveraging from our search advertising synergies. You combine all these three key factors, we have very high confidence that our product will compete extremely well and will grow very strongly going forward.
Greg, let me add another point, just to hone in on the point of looking at Baidu App search and feed together. As you know, one of the things in trying to get app adoption is channel costs, right? Because we have two functionality, when we look at channel costs, we look at it very closely on the ROI, if we can get a bang for it. That means that you're looking at the lifetime value of that user once you acquire versus the cost. Because we have two functionalities, we look at the functionalities together and what we can get from the user as we're acquiring this. You can see our peers who basically it is reinforcing our strategy by now trying to add search.
The thing is, it takes many years to actually build their quality, but it just shows that our business model works, that you're probably going to be able to get more monetization if you have Feed and if you have a very good Search engine.
Yeah. I would add that in terms of content generation, we view the whole Baidu platform as one content ecosystem, which includes the Search content in what format, the iQIYI video, which is mostly in the long form video, but we can cut that into shorter clips and distribute it through our Feed product. We also have Baidu Tieba, Baidu Zhidao, Baidu Baike. We have lots of content generation tools and apps. We continue to integrate all this kind of content into the Mobile Baidu platform. Mobile Baidu, both in terms of Search and Feed, has unmatched distribution capabilities for content. Therefore, we will be able to build a stronger and stronger content ecosystem.
Our next question today comes from the line of Alex Yao from J.P. Morgan. Please ask your question.
Hi. Good morning, guys. Thank you for taking my question and congratulations to Herman's new role. Wish you all the best in the new role. I'd like to follow up with the full Q revenue guidance. I think, Herman, you mentioned that the core Search could be growing at 26% on year-over-year basis for Q4. If my math is correct, that still suggests that the core Search is going to decline somewhere between 2%-4% QOQ from Q3 to Q4. Qi also mentioned that AI has improved the click-through rate, which supposedly should improve the RPM. Does it mean we are seeing some weakness in the traffic for the core Search? Thank you. I stop here.
Yeah, Alex. Thanks. Actually, the math I'm looking at is converse to what you just said. When you look at our sequential growth from Q3 to Q4, taking out iQIYI, taking out the businesses that we have spun off, if you look at the midpoint, we're talking about maybe a sequential growth of 2%. It's up 2%. This is very consistent with, for example, past year. When you look at 2015, using the same math, taking out businesses that we have spun off sequentially, you're also looking at around 2%. I'm looking at this number pretty consistently because it's an up quarter. I think that when you put in the calculation, you might get a different result.
Next question. Our next question today comes to the line of Thomas Chong from Credit Suisse. Please ask your question.
Hi. Thanks management for taking my questions, and congratulations, Herman, on the new role. I have a couple of quick questions regarding the cloud and the financial services. On cloud
How should we ambition the business scale over the next couple of years, and how we think we can outpace our peers in terms of customers and revenue? My second question is about the financial services. Can management provide some more guidance about the timing for profitability, the business breakdown, et cetera? That would be great. If not, can we actually rank different business in descending order under the our revenue? Thank you.
Let me take both questions. First, with regard to our ABC Cloud business. The fundamental position differentiation is ABC, which is AI technology, plus big data, which is the data assets Baidu's ecosystem is able to harness. Overlay on top of a very strong set of cloud infrastructures and services. With that in mind, our focus is to leverage ABC, deliver industry vertical specific end-to-end solutions with big customers first, horizontalize, build industry-specific ecosystems to grow our business. We've been showing strong tractions in that approach. As I mentioned in the prepared remarks, we are continuing to expand the industry verticals that we're covering and deepen the penetration in each of the verticals.
In terms of the pace of growth, we are growing at a very rapid pace and on the pace essentially to double on an annual basis or more than double on an annual basis. Even though it starts from small numbers, it will become materials over the coming 2-3 years. That's for our cloud business. With regard to FSG, you mentioned the profitability. This is indeed something the teams internally has been working on super hard. At this point, we are not prepared to share the specific timing, we are on a very healthy path for this business quickly to become profitable.
Our next question today comes on line of Juan Lin from 86Research. Please ask your question.
Hi, good morning, Robin, Dr. Lu, Herman, and Sharon. Thanks for taking my questions, congratulations, Herman, for the new role. My first question is on iQIYI. I've noticed that your competitor have recently announced aggressive content purchase in the production plan. I'm wondering how should we foresee the competitive landscape for video content going forward, what is the trend for video price for the next year? Secondly, on the core business, as you have been adopting AI and innovative products for core business, I'm wondering what is the long-term potential for monetization improvements, such as lifting click-through rate and CPC? In particular, click-through rate and the CPC for search has been quite stable. I'm wondering whether you have seen new opportunities to improve search monetization due to the AI adoption going forward. Thank you.
Let me take both questions. First, with regard to iQIYI business and the content production and content cost question that you mentioned. First of all, let me reemphasize iQIYI, their core competency. The team has done a tremendous job over the years building up a core competency, a combination of content procurement and content self-productions. This is very important. If you look at the past track record, there is open public statistics available for you to look at. iQIYI has been consistently procuring top-ranked video content, and also iQIYI was able to show a strong track record of self-producing content. Overall, this is the key focus areas for the iQIYI management team. Continue to strengthen the leadership positions in content procurement and the content self-production. That's a very important part.
The other part I would like to emphasize is iQIYI also has been driving a lot of product innovations, particularly the personalized video feed recommendation product. It has been growing very strongly, and that product also has a lot of monetization potentials. Overall, we continue to have strong confidence in iQIYI as a leading video entertainment platform for China. Continue to drive strong growth by user engagement and monetization. The second, with regard to using AI to improve the monetization for our core business, it's a great question. That's indeed one areas we have strong potentials. Again, through a combination of AI technology and data. In AI technology, I believe in the last call I mentioned that we've been using deep learning, using the better models to be able to harness more signals from our conversion data and to improve the conversion rate for ad product.
As we all know, search advertising economics is fundamentally conversion economics. As long as we keep improving conversions, we will be able to keep improving our ability to monetize. That's number one. Number two is the direction highlighted by our dynamic ads, which is our partners, our advertising partners, are increasing willing to share their structured data, their product catalogs, and their user data. By a combination of their structured data By combination of their user data, again, we will be able to use AI technology to harness in more conversion signals, more user interest. Those are core dimensions that will pay off in both search advertising and in feed monetization.
Third area is as Baidu App continues to grow through the twin engine of feed and search, we will be able to draw more budgets, more diversified ad budget, not just focused on performance ads, but increasingly brand advertising and rich media advertising. Through AI technology, we will also be able to offer new interactions, as Herman mentioned, through face recognition, through AI. These are all key avenues for us to systematically increase the yield of our ad platform over time. Again, through the combination of AI and data assets.
Jin, let me add a few points on top of that. When we're talking about dynamic ads that she just talked about and how data structure with partners, customers and so forth, let me just give you a couple of examples. For example, when you look at Baidu Search, you can now search, for example, going to a destination. By working with Ctrip, we'll get a display. For example, you want to go to Beijing to Shanghai. When you search that, we'll actually get a display of the data coming from Ctrip. Then on our website, even though behind it's powered by a Ctrip system, on our website, you'll actually get the display. You'll actually be able to go directly to Ctrip's platform and be able to consummate that transaction. Everything will be done from a user interface perspective on Baidu's platform.
The very fact that you're able to go in the path of just kind of doing a search, now you're able to actually consummate the transaction from a user experience directly on here, having that closed loop, that's pretty powerful. Another example is if you're talking about financial services, on the one hand, we're making micro loans, having asset management. On the other hand, it's being able to get that user behavior, that technology, and then be able, for example, if you want to search for a loan based on different user profiles, we'll actually push you different products and be able to set the credit limit based on your past behaviors. These are kind of the steps we're going. It used to be just kind of contextual search, searching for work.
Now you're actually being able to customize the product based on different users using a Baidu product. I think that is just significant amount of improvement, and that's kind of the technology we're heading with AI. Just a few more points on iQIYI. I think we need to recognize the business model of online video. If you look at in the U.S., there's basically type 2 of models. Type 1 of model is basically you're just a distributor. You license content that's already well-known. You pay for heavy content costs. What you're doing is you're just trying to arbitrage and trying to deliver on your platform. That's type 1 model. We've seen that over the last few years, people who license already very top brands and so forth, and the margin is very low.
The other type of business we're seeing, and we're seeing this is the one that's more successful, is companies that are able to leverage and bridge those with deep entertainment knowledge and technology. Those are the type of companies that can actually self-produce content that become top hits. At the end of the day, that business model is basically trying to be an IP powerhouse. Once you can make top hits, you can get a lot of memberships, attract a lot of people. When you have that scale, that base, you also prove to the entertainment industry that you really understand entertainment. Other people will want to license their content to you. All of a sudden, you have that inertia of both user scale and also that partnership.
By having that kind of creative, innovative leverage, you can continue to make more and more content. This is kind of the path that we're seeing with iQIYI. I believe in online video, it's important to understand that business model, those that actually can create great content while being a great technology company and continue to get more and more content, whether it's through self-production or through signing with a strategic partner because they trust you, they know that you understand entertainment. I think that's where iQIYI is positioned, and I think that's where our strategic advantage is.
Our next question today comes on the line of Alvin Jiang from Deutsche Bank. Please ask your question.
Hi, management. Thank you for taking my questions, and congratulations on Herman's new role. My first question is on AI. For AI new initiatives, we can see there are some other big internet companies are also doing the same exploration. Do you think there will be much more competition in this field in two to three years? What is Baidu's key strength in AI comparing with peers? Also on financial impact, how big is the investment, specifically on AI in 2017, and what's the monetization plan? Thank you. I have a follow-up.
Okay, let me take both questions, Herman will chime in as you see. With regard to AI competitions over the next two to three years, it is indeed the case there's a lot of industry energies and investment across a peer set of companies in AI. However, Baidu has fundamental strengths along a few key dimensions. First, as you all know, Baidu is among the very first companies to identify the importance of AI and proactively invest in AI technology and talent over many years ago. It takes time to develop the core competency of talent base and technology. That's number 1. Number 2, we view the best strategy and best path forward for AI commercialization is to platformize it and to build an ecosystem. That will have scale advantage. That will have increasingly bigger and bigger first-mover advantage.
For example, in our self-driving car, Apollo ecosystem, we are actually the world's very first open ecosystem, our ecosystem is growing at a very rapid pace. That, over time, will have increasingly competitive barriers in advantage for Baidu. Similar story is happening on our DuerOS platform. At this point, we are clearly, by far, the best leading technology platform for conversational AI in powered devices for home, for in-car navigations. Again, you will see scale advantage, first-mover advantage in those front. Overall, Baidu's combination of early investment, the buildup of talent base and technology, the platform orientation, and the ecosystem approach will enable Baidu to accelerate the pace of commercialization and increasingly build up competitive barriers in AI competition. With regard to AI investment, we've been very disciplined in running our overall internal operations.
Yeah. Alvin, let me just add a couple points. We use a very systematic approach to ensure our OpEx, particularly R&D and sales marketing, are allocated proportional to the economic long-term returns that we can see. For example, we've been investing heavily in Baidu App, in our feed, in our mobile search, because that's the area we see tremendous growth opportunity. At the same time, the AI-enabled new businesses, the financial maturity will take time to come. We have to invest at this stage proactively. The key is being very disciplined, being very efficient. We are very pleased how everything plays out, and over a period of time, Herman, our CFO, will share more operating highlights with regard to cost and benefit for our AI initiative businesses over the coming quarters.
Yeah. Alvin, let me just add a couple points. With regards to our competition, I think AI at the end of the day is just super computing, right? The power of your computing, the way you can actually leverage that is really the broad sets of data that you have, right? When you think about Baidu, you look at the properties that we have from maps to years and years of our search data and so forth. We really have the broadest set of data across the users in China. As a result, because we know the user so well, we know across so deeply across the spectrum of data, then when you have that super computing power, then you can have pretty amazing results. As we mentioned over and over, our AI is just focused on three platforms.
It's focused on mobile, it's focused on voice assist, and it's focusing on autonomous driving. Really, when you're looking at peers in this space, you're looking at basically a lot of global players that are kind of focusing on this direction. When you look at Baidu, I think you have to look at the computing that we're doing, which is AI. You look at the inputs that we have, I think it's very different than other internet companies because we're in the business of search for so many years and the properties that we have across China. Then separately, when we're looking at our sector that we're going into, trying to be an operating system for autonomous driving, for smart devices and so forth, it's pretty unique position, I think.
With regards to investment in AI, as I mentioned earlier, over the last year, we've been becoming more and more singular focused, right? We divested businesses that were not as synergistic as we think. For example, like Baidu Games, like Baidu Delivery and so forth. As we're divesting from those things, we now have more capital to spend. The great thing about Baidu's new business model is that we're leveraging AI on our traditional business as well as our new business. As you're further developing the power of computing mixed with the broad amount of data that we have, we're really able to leverage our existing business and also grow the new businesses. The total investment is quite huge, but a lot of it is already built into our financial model.
Our next question today comes on the line of Piyush Mubayi from Goldman Sachs. Please ask your question.
Thank you for taking my question. I have a very simple question. This is addressed to Dr. Qi. In which innings are we in the utilization of AI in improving search monetization, according to you? Thank you.
You are using a baseball analogy?
Yes.
I would say probably the second inning, somewhere between the second and third inning. There's a long trajectory ahead of us. Good question.
Thank you.
Our next question today-
Next question, operator.
Yep. Our next question today comes to the line of Grace Chen from Morgan Stanley. Please ask your question.
Hello. Hi. Thank you for taking my question. I'd just like to know, can you give us guidance about the various cost items such as content cost and G&A, traffic acquisition cost, and R&D? I remember we talked about to double the content cost year-over-year earlier this year, but if you look at the run rate so far, it seems like the spending is actually much lower than what we planned earlier. Can you help us understand the gap there and update us the latest guidance? Thank you.
Yeah. I think rather than giving line-by-line guidance, I think it's probably easier just looking at our operating margin. When you look at our non-GAAP operating margin for the third quarter, I think it's around 24%. I think historically when you're moving from Q3 to Q4, you're going to see several points drop. Really our operating margin at the end of the day, a big factor of that is looking at our channel cost. As I mentioned earlier, it's purely on an ROI basis. If we think that there are opportunities where we can spend, we can increase our users, and we think that the lifetime value of the user is worthwhile, we'll spend a little bit more. That's probably one of the factors that's going to decide what our ultimate operating margin is. We should expect a few points drop from Q3.
With regards to content costs, I think several factors in there. The biggest factor, as you probably know, it's from iQIYI. As I mentioned earlier, in the fourth quarter, we expect iQIYI revenue to be lighter because of the environment that we're in. Naturally, when your revenue is lower, that means that you're not going to buy a lot of hot series content. I think those things are related. I think that's probably the major factor for our content going into our Q4.
Our next question today comes on the line of Ming Zhu from UBS. Please ask your question.
Thank you, management, for taking my question. I just have one very quick follow-up on the video side. In last month, one of your competitors actually announced a very impressive membership number. I remember, Robin, you mentioned your membership number was over 30 million during the last earning call. Could you maybe update us on the latest membership count you have, and can you maybe share with us your long-term target on this front? Thank you.
Yeah. We noticed that some of our competitors are aggressively promoting their membership. They either bundle the memberships with their existing product or cut it with a very big discount. We no longer think the number of memberships matters that much, but the subscription revenue as a % of total revenue, iQIYI continue to lead, and we have a significant lead over the competition.
Operator, we'll take the last two questions. Our next question today comes on the line of Natalie Wu from CICC. Please ask your question.
Good morning, management. Thanks for taking my question. My question is regarding the feed ad. As it is mentioned, we should look search and feed ad together as they are in one single app or one unified product. Does that mean feed ad is showing cannibalization effect for core search? As you mentioned in prepared remarks, feed ad generated $1 billion revenue on an annualized basis in the third quarter. Does that mean your feed ad revenue in the third quarter reached RMB 2 billion? Is that number net of sales rebate or on a gross basis? Does that number include iQIYI's video feeds? How do you view the growth prospect of this business? Will you consider launch individual app for that? Thank you.
Let me take the first part with regard to cannibalization, and then Herman can talk about the number breakdown. The simple answer is clearly no. Feed ads has been pure incremental to our overall advertising business. There's a few key areas to further elaborate. First, our search advertising budget and ad spend continue to remain strong and stable and growing. That's very important. The second is we are increasingly adding more capabilities to offer to advertisers when they pick our feed ad product. For example, ad creatives, ad authoring tools, so that make it much easier for advertisers to use our feed product. It's also important to point out many of our search advertisers, they see the combination of search feeds is very useful and helpful for their advertising campaign because oftentimes they want to cover their target customers across the different layers consideration funnels.
There's indeed campaigns across both search and the feeds, but from ad spend perspective, it's very much incremental. More importantly, over time, the feed products would harness more brand advertising budget, particularly richer advertising creatives, ad interactions. Overall, we see the combination of ad platform offering both search advertising and the feed advertising is net incremental and have very strong headroom and trajectory for long-term and sustained growth ahead of us.
Hi, Natalie. Let me just add a few points to that. I think your question just assumes that feed and search are two separate products. Let me put it in context of why we think that we should look at this as one. If we start looking at these two as one product, then actually strategically for Baidu, we're going to be able to differentiate our product with other people. For example, one way to look at this is, for example, the content ecosystem, right? The idea is that the feed is where you have a healthy ecosystem, where you have a lot of bloggers that come in and actually use, for example, Baidu's feed account. What Baidu is able to do is once these users come and actually generate content on these accounts, Baidu also makes it available for people to actually search these accounts.
In other words, we really don't have a preference whether the user actually looking at the content through feed or looking at the content through search. We see it as two ways to display the same amount of content. On the one hand, we're expanding our content base using Baijiahao in addition to what the properties that Robin mentioned earlier, like Baidu Zhidao other properties like KMI and so forth. That's why I think the more we allow the users to leverage different ways to look at the content, we see it more as one product, and that's a huge differentiator than the other properties. Secondly, your question on the $1 billion. When we say analyze by $1 billion, what you need to do is look at it for Q3, just divide it by 4 to get approximately where we are with Q4.
Okay, the question on AI feed. AI, we see that as separate category, as I mentioned in the guidance. Feed is really the feed product within Baidu's other products.
Let me just add that for search and feed, we can look at this from two fronts. From the advertiser front, I think as Qi mentioned, feed can attract a lot of brand advertisers. On the other hand, we do share a large amount of advertisers both on search and on feed. I view that as a strength instead of weakness. We have a very large advertiser base, and they can distribute their ads both on search result page and the feed page. This is a very advantage point for Baidu. From the traffic point of view, I've been saying this over the past few years many times. Our business is very much traffic-bound, not budget-bound. The more traffic you generate, the more revenue you generate. Advertisers always want to buy more from our platform.
Ladies and gentlemen, we are now approaching the end of the conference call. Thank you for your participation in today's conference. You may now disconnect.