百度 (BIDU) 2016 Q3 法說會逐字稿

Hello, and thank you for standing by for Baidu's Third Quarter 2016 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 hand the meeting over to your host for today's conference, Linda Sun, Baidu's Senior Manager of Investor Relations.

Hello, everyone, and welcome to Baidu's third quarter 2016 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 and Jennifer Li, Baidu's Chief Financial Officer. After their prepared remarks, Robin and Jennifer will answer your questions.

Before we continue, please note that the discussion today will contain 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 measures 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 available on Baidu's IR website. I will now turn the call over to Baidu's CEO, Robin Li.

Hello, everyone, and thanks for joining today's call. In the third quarter, we continued to implement measures to improve customer quality and foster a healthy environment to enhance user experience and provide long-term sustainable growth. Beginning in the third quarter, we took proactive measures and required all customers on our platform to submit ICP licenses and verify enterprise bank accounts. We expect to complete this self-imposed verification measure across our platform in Q4. The new advertising law, which imposed stricter regulations on online marketing also became effective on September 1. The implementation of this regulation and our self-imposed proactive measures will have a short-term impact on our business. We expect the most pronounced impact on our business in the fourth quarter, followed by a recovery early next year. We remain highly confident in our long-term outlook, underpinned by the vital and attractive fundamental value proposition of search and our ongoing investment in technology, in particular, artificial intelligence.

As we improve the quality of our platform, we continue to focus on driving product innovation and building our long-term core capabilities. We are particularly excited about how our growing leadership in artificial intelligence is improving user experience, and user stickiness across our product lines while helping us reshape traditional industries such as finance and local services and driving future growth areas such as our digital personal assistant initiative Duer, and autonomous driving.

During the quarter, news feed proved to be an excellent complement to our core search product and existing content ecosystem, and contributed strongly to user stickiness, distributed content on the newsfeed platform grew 20 times since its launch in May, helping to drive a five-fold increase in newsfeed daily active users to about 70 million, and a three-fold increase in time spent per user over the same period. Our ability to help customers precisely target the right newsfeed users based on their search queries and interests has opened the door to new categories of advertisers such as real estate, and we are very encouraged by the result to date.

Mobile search represents a growing majority of our total search traffic, and during the quarter, we released Mobile Baidu version 8.0, a major upgrade of our app with enhanced voice input, improved newsfeed display, and personalized mobile homepage. Our initiative increased the number of images in search results, has shown great progress with nearly 90% of mobile searches producing results with images.

In August, we launched our mobile instant pages open-source initiative. When mobile websites adopt this backend technology, the speed at which they can be opened from both Baidu search and news feed increases by up to 80%. This not only enhances user experience, but also increased website page visits by up to 40%.

We are confident that Baidu mobile search has never been more powerful.

Turning to our services ecosystem...

Nuomi continues to play an important role as a part of our services ecosystem, further integrating Baidu Search and Maps, with local queries as a percentage of total search volume in mobile growing steadily from the beginning of the year, Nuomi supports search with richer, more localized results, enhancing mobile search's role as an O2O gateway. Nuomi supported maps by contributing the large majority of closed-loop points of interest, and added advanced AI-driven features such as customized restaurant and retail store recommendations which helped drive stronger user engagement.

Ensuring that we attract the best merchants to Nuomi is the top priority, and offering a suite of services for them to make the most out of the platform is an important part of this. Efforts such as helping merchants build pages for Local Express products and providing tools including CRM systems and templates to enhance their online stores have been paying off and we have seen very good growth, with 2.2 million merchants on the platform as of the end of September, a nearly three-fold increase year-over-year.

Consumers continue to fall in love with the speed and convenience of Baidu Deliveries, formerly named Baidu Takeout Delivery, with the service now expanding beyond the restaurant takeout delivery service. The platform registered more than 150% year-over-year GMV growth during the third quarter. This success is largely attributable to our AI-driven logistics technology. With an average delivery time of 32 minutes and punctuality rate of over 98%, Baidu Deliveries continues to define new standards in the industry. We are proud that Baidu Deliveries continues to expand into new service categories, such as fresh food product delivery, and on-demand logistics, a strong validation of the potential of our AI-powered logistics technology.

Baidu Wallet also gained strength during the quarter, with activated Baidu Wallet accounts growing to 90 million at the end of the third quarter, up from 80 million at the end of the second quarter.

Baidu Maps further solidified its number one market leadership with nearly 350 million MAU. Baidu Maps serves as a powerful O2O gateway, leading the industry with over 22 million local services points of interest, and 1.9 million of which provide closed loop services to users within the app. Baidu Maps also provided indoor maps for over 3,000 large shopping malls which is unparalleled in the industry. We have expanded our footprint abroad to over 100 countries and regions to serve Chinese travelers overseas and Baidu Maps worked with over 1 million developers during the quarter, about 75% of the market, and provided open map services to 600,000 third party apps and websites, more than any industry peer in China.

As you know, Baidu has been investing in AI for years, and we are proud to be recognized as one of the leading AI innovators globally. Our AI infrastructure is already helping to make all of our products more efficient and useful. For example, AI enables better filtering of bad search results and keywords, helps generate more intelligent and targeted search results and news feed, automatically detects non-compliant images on Nuomi, enhances the accuracy of Baidu Deliveries' delivery time estimates, and predicts traffic conditions in Maps among other things. AI also enhanced app displays and broadened the range of ad formats, driving further value to our customers. And our amazingly accurate speech and image recognition have become the envy of the industry, helping to drive further rapid growth in voice and visual search queries.

Our AI-powered Baidu Translate product is the first online translation service to combine deep learning with a statistical model. With leading voice and image input and recognition, Baidu Translate services 28 languages and processes over 100 million requests a day on average, many of which come directly from search, and supports over 20,000 third party applications.

Beyond our core products, we are already applying AI and big data to many areas of people's daily lives. For example, Baidu AI medical assistant Melody, helps doctors to diagnose illnesses from online conversations with patients. Our intelligent voice-recognition based CRM system understands callers' questions and guides sales people with real time answers and data. We use augmented reality to display live advertisements and our digital assistant, Duer, provided live TV commentary for Olympic basketball games. We also recently began to test facial recognition-based entry control systems here on Baidu Campus.

In the beginning of last month, we open-sourced our AI platform, "PaddlePaddle" to the developer community, providing access to Baidu technology in the areas of voice and image recognition, natural language processing and machine learning.

Now, looking to some of our newer areas...

While Financial Services is still an early-stage effort, we are making strong progress in creating an innovative platform to transform the traditional finance industry that gives users unparalleled convenience and approval speed while minimizing risk through our AI-based risk control technologies including facial and fingerprint recognition, OCR recognition of identification documents and liveness detection. We continue to grow our leading position in the growing market for education loans in Q3. To date, we have worked with approximately 1,300 educational institutions, many of whom are our long-time search customers and have more than 50% market share. We are also making progress expanding into new areas, partnering with over 300 companies or merchants in new categories such as travel, where we partner with Qunar, cosmetic surgery, home decoration, and home rentals.

Autonomous driving is an important area for future growth, where AI technology has helped us take an early lead. Enabled by Baidu Brain's powerful image recognition and processing technologies, our autonomous driving car recently broke a number of records in the KITTI benchmark tests, achieving an accuracy rate of 90.13% in the vehicular recognition tests. Last year, we began road tests of autonomous driving under real world conditions in China and in August, we received our permit to begin testing Baidu's autonomous cars on the public roads in California. We have an ambitious development schedule but with our top team here in Beijing and in the US, leading technology and unrivalled data resources, I'm confident that we will lead the way in making autonomous driving a reality.

Our iQiyi platform performed very well in the third quarter, with strong growth in revenue and paid subscriber numbers. In August, according to iResearch, iQiyi's mobile app leads with 140 million daily active users, 520 million monthly active users and 590 million hours monthly user time, representing a significant lead over the closest industry peer. We look forward to further supporting iQiyi's growth.

In conclusion, while we are still experiencing some short-term challenges as customers adapt to the new regulations and our stricter standards, our core value proposition for users and customers remains solid. Just as importantly, we have a clear strategy for future growth as we enter the new era of AI. We are ahead of the game in many ways and thrilled to see the benefits of Baidu AI already becoming a reality in the lives of our users and customers.

With that, I will now turn the call over to Jennifer for an update on financials.

Thank you, Robin. Hello, everyone.

We continue to be in a challenging transition period as we implement higher standards and comply with new regulations. As we mentioned, we took proactive measures and required all customers on our platform to submit ICP licenses and verify enterprise bank accounts. The process is on track with the completion of the verification work in healthcare and finance verticals in Q3, and we expect to complete the verification across all verticals in Q4. As Robin mentioned, we expect the most pronounced impact to our business to be in the fourth quarter, with recovery beginning next year. We believe these higher standards and regulations will result in a higher quality customer base and healthier industry environment over the longer term. Looking ahead, we will focus on further enhancing our value proposition to our users and customers. We will continue to execute our plan to build out our ecosystem and to spend where necessary with disciplined approach while closely monitoring return on investment. At the same time, investing in technology and R&D will continue to be a priority.

Now moving to the financials. All monetary amounts are in RMB unless stated otherwise. For the third quarter, total revenues were RMB18.3 billion, representing a 0.7% decrease from the corresponding period in 2015. The year-over-year increase is 6.7%, excluding Qunar, in the third quarter of 2015.

During the third quarter, Baidu had approximately 524,000 active online marketing customers, representing a 15.9% decrease from the corresponding period in 2015 and a 0.9% year-over-year decrease, excluding Qunar. Revenue per online marketing customer for the third quarter was RMB31,300, a 10.6% increase from the corresponding period in 2015. No year-over-year change, excluding Qunar, and a 10.2% increase compared to the second quarter of 2016.

Traffic acquisition cost, as a component of cost of revenues in Q3, was RMB2.6 billion, representing 14.2% of total revenues.

Bandwidth costs, as a component of cost of revenues, were RMB1.2 billion, representing 6.8% of total revenues, compared to 5.3% in the corresponding period in 2015.

Depreciation costs, as a component of cost of revenues, were RMB802 million, representing 4.4% of total revenue, compared to 3.6% in the corresponding period in 2015.

Operational costs, as a component of cost of revenues, were RMB1.2 billion, representing 6.3% of total revenues, compared to 6.8% in the corresponding period in 2015.

Content costs, as a component of cost of revenues, were RMB2.2 billion, representing 12.1% of total revenues. This increase was mainly due to iQiyi's increased content cost.

SG&A expenses in Q3 were RMB3.6 billion, representing a decrease of 37% from the corresponding period in 2015 and a year-over decrease of 23%, excluding Qunar, in the third second quarter of 2015. The year-over-year decrease was mainly due to a decrease in promotional spending for transaction services.

R&D expenses in Q3 were RMB2.6 billion, a 2.8% decrease from the corresponding period in 2015 and a year-over-year increase of 14%, excluding Qunar, in the third quarter of 2015. Share-based compensation expenses, which were allocated to related operating costs and expense line items, were RMB414 million in Q3, compared to RMB400 million in the corresponding period in 2015. Excluding Qunar, the SBC cost was RMB278 million in the third quarter. This year-over-year increase, with Qunar's impact excluded, was a result of increased share grants to employees.

Operating profit was RMB2.8 billion, representing an 11% increase from the corresponding period in 2015 and a year-over-year decrease of 12.5%, after excluding Qunar's impact. Non-GAAP operating profit was RMB3.2 billion, a 10% increase from the corresponding period in 2015 and a year-over-year decrease of 7.5%, excluding Qunar.

Income tax expense was RMB1 billion for the third quarter. The effective tax rate was 25.3%, compared to 19.4% in Q3 of 2015. The increase of effective tax rate of the third quarter this year reflects that some loss-generating entities in the Group cannot be consolidated for tax purpose under PRC tax law.

Net income attributable to Baidu for Q3 was RMB3.1 billion, a 9% increase from the corresponding period in 2015. Basic and diluted earnings per ADS for the third quarter amounted to RMB8.53 and RMB8.51 respectively.

Non-GAAP net income attributable to Baidu for Q3 was RMB3.4 billion, a 6% increase year-over-year. Non-GAAP diluted earnings per ADS for Q3 was RMB9.92.

As of Q3, the Company had cash, cash equivalents and short-term investments of RMB78 billion. Net operating cash inflow and capital expenditures for the third quarter were RMB3 billion and RMB1.2 billion respectively.

Total headcount on a consolidated basis, including invested entities, was about 44,800 at the end of Q3. This represents an increase of 2.5%, as compared to the end of last quarter.

Now let me provide you with our topline guidance for the fourth quarter of 2016. We currently expect total revenues for the fourth quarter to be between RMB17.8 billion to RMB18.4 billion, representing a 4.6% to 1.7% year-over-year decrease. Please note, this forecast reflects Baidu's current and preliminary view and is subject to change.

I will now open the call to questions. Operator, please go ahead with questions.

(Operator Instructions). Our first question today comes from the line of Eddie Leung, from Merrill Lynch. Please go ahead.

Good morning. Thank you for taking my questions. I have two questions. The first one is about your mobile traffic outlook. We have seen a slowdown in the mobile traffic growth, so just wondering how much of that is due to perhaps a cut in traffic acquisition cost and how much of that is perhaps due to competition or industry slowdown. Any comment on the upcoming quarter's mobile traffic would be great.

Then secondly, just a question on your online finance business, so we understand that there has been some education loan business, so could you comment on the credit risk and how we are going to account for that on the balance sheet? Thanks.

Okay, I will address the first question, and Jennifer can take on the online finance one.

For our mobile search traffic, as you know, the Internet penetration rate in China is already around 50% and so going forward, we cannot count on a rapid growth of new internet users coming online and learning to do search. So from that point of view, the growth driver for search traffic is less than before.

But overall, I think search is still very fundamental and critical to every Internet user, and users are still getting used to mobile search. We have seen the trend that, on average, people are searching more through their mobile phones, so this kind of frequency related driver will continue.

Of course, we are still innovating in mobile search, so we have seen very rapid growth in terms of voice-enabled queries, and we will continue to promote this kind of new input method including voice and images, and we think there's plenty of room for growth in our mobile traffic.

Eddie, for the financial services business that we are getting into, we think we have a natural advantage in the educational sector. Many of the current advertisers are our search customers and we have been doing business with them for a long time. And so our approach to educational loan is working with these institutions and also assess the individual credit risk, given our technology infrastructure as well as the data capabilities. So our approach to this financial services is quite, I would say, measured and prudent.

We have originated some education related loans and they are built up under the balance sheet under "other assets". We also, based on estimates, start to accrue loss provisions for these related loans and they're also embedded in the "other assets". Today is only the early stage. We don't really have the testing results to see the actual losses, but we're measuring them and monitoring them very closely.

We believe that, based on Baidu's capabilities in AI using the data and the technology and also our knowledge about that vertical, this is the entry point that we are experimenting with our financial services business.

The next question comes from the line of Alicia Yap from Citigroup. Please go ahead.

Hi, thank you. Good morning, Robin, Jennifer and Linda. Thanks for taking my questions.

My question is regarding the current plan and longer-term view on your O2O strategy. So with sales and marketing spend come down quite a bit and the transaction GMV also decelerated to 49% this quarter from triple-digit growth last few quarters, can you share with us your current thinking about the O2O business? And for longer term, let's say we have the opportunity to do some business combinations on O2O business, how should we think about Baidu core search growth and overall mobile positioning if we don't have operation control of the transaction service in the long term? Thank you.

Yes, Alicia, I think the O2O initiative is pretty much based on our vision of connecting people with information and connecting people with services, especially for connecting people with services. Our main focus is to make sure that Mobile Baidu users can start the search on our search app and finish the transaction wherever relevant. So we've been working very hard to make that kind of experience better and better and our O2O initiative is surrounded on that.

As I mentioned during the prepared remarks, we now have 22 million merchants on our Nuomi platform, that's like triple over a year ago. From that point of view, we're very determined to make Nuomi a success. But in the meantime, we realized that the business model for O2O may not be purely a commission-based or transaction-based model, so we started to shift a little bit the transaction-based nature to advertisement-based nature, and we also lowered the subsidies we provide for all kinds of transactions on the Nuomi platform. So as you correctly pointed out, the growth for GMV slowed down a little bit, but the loss also contracted too. That reflects our mission of making sure that the users have the best experience regardless of a transaction or non-transaction or regardless whether we control the O2O 100% or we rely on some kind of third party.

The transaction we did last October with Ctrip is a perfect example. Now, our users can book a hotel from Baidu Maps or Baidu Nuomi or Mobile Baidu, and the experience is very smooth, but we don't own the Ctrip business. So this is certainly a possibility going forward for other verticals. But again our focus is to make sure that the user experience for transactions on the Baidu platform is smooth.

The next question comes from the line of Alex Yao from JPMorgan. Please go ahead.

Hi. Good morning, Robin and Jennifer, and thank you very much for taking the call.

Jennifer, I have a financial question about the fourth quarter guidance. Can you talk about the implied core search revenue growth within the guidance? You guys mentioned the core search will see recovery in early next year. Are you guys talking about on year-over-year basis or Q-o-Q basis? And how should we think about the magnitude of the recovery and the timing of the recovery in the first half of next year? Thank you.

Thank you, Alex. Obviously, as I mentioned, we're going through the transition period. Much of the work is related to us validating our customer base with the required documentations. In Q4, across all board, we're targeting to finish all the verification process with all the customers. And so what gets impacted is some of the customers may be removed from doing business with us because they cannot provide the proper documentations, or there's also a change of display or different ways of doing advertising or keywords that we monitored with restriction, so, customer cannot be spending as they used to be with us. So, very much closely a tighter control over the whole platform.

As you see, Q4, we will feel a more pronounced impact of the transition work that we're carrying on. But we do target that by year-end, all the verification work should be completed, and that will be the customer base that we're working towards going forward getting into next year and we will also recruit new customers. The effort will be focusing on building new customer base that with the same kind of standard.

As we look out into next year, we do anticipate that much of the quality assurance work has been done this year and next year we should start to resume our normal work just to build the customer service and also provide good services to the users.

Basically this is a very special period. It's not the time to talk about year-on-year growth yet, but you can see that Q3 for online marketing business is flattish and, going into Q4, we will continue to feel this transition impact, but that kind of sequential changes should finish by the end of this year, and going into next quarter we are focusing on building the business again. So this is the transition going forward and basically Q4 is what we anticipate to be kind of the bottom.

The next question comes from the line of Chi Tsang from HSBC. Please go ahead.

Great. Thanks very much for taking my questions. I wanted to ask you about two topics. Firstly, about the news feed. The news feed now has about 70 million DAUs, that's ramping very rapidly. I was wondering if you can give us a little bit more commentary regarding time spent and also whether or not the news feed is helping to drive more paid clicks, and what your monetization plans might be.

And then second topic, I was wondering if you can give us an update on the progress you're making as it relates to autonomous driving. You have a plan for small-scale production by 2018. Thanks so much.

On the news feed, we pretty much show the feed on the Mobile Baidu app. We leverage on our existing user base and we only count user that's an active news feed user when they click or they scroll on that related content. So I think the 70 million daily active users is already very significant. In the meantime, I mentioned that the time spent has tripled, so it's growing very fast. We will let you know the average time spent per user in absolute numbers when we feel appropriate.

And the monetization capability for news feed is also very promising, because we have a very large customer base for our paid search. Right now it's about mechanism, so a lot of advertisers are trying our news feed ads and they're very happy about the effectiveness. So we think when we decide to become more aggressive in monetizing the news feed, the revenue potential is huge.

What was the second question?

Autonomous driving. In particular, you have a plan for small-scale mass production by 2018. How are you sort of progressing towards that objective?

I think we are still on track for small-scale production by the end of 2018. Right now we have more than 40 testing cars in China and U.S., and we are making significant progress almost every day. We believe we have one of the strongest teams in autonomous driving technology, and we're also talking to a number of potential partners to supply us with cars and related technology. So we are very happy and we are on track to make that happen.

The next question comes from the line of Evan Zhou from Credit Suisse. Please go ahead.

Hi, good morning, Robin, Jennifer, Linda. Thanks for taking my questions. I have two follow-up questions on cost items.

So I was wondering, R&D expense on this quarter seems to have a pretty decent increase, I was wondering if we could break it down a little bit on what kind of the incremental spending on our longer-term AI-related initiatives and shall we expect them to have a kind of permanent step-up of these spendings down the road for this line?

Also, what's the roughly RMB1.2 billion of the other income line, what caused that line to increase in this quarter?

And finally, maybe whether you can share with us some thoughts regarding the use of cash, as we are basically decreasing our cash burn on, say, O2O, and still decent cash flow for the core search business is good news. So, any plans to maybe return more value to shareholders? That would be helpful. Thank you.

Okay. Thank you, Evan. For the R&D expenses, throughout our history we've been steadily investing in the R&D capabilities, and the R&D expenses line item is particularly related to the talents, the manpower capabilities. And what you are seeing is I would say a consistent approach to R&D investment, and going forward that will continue, I would say, to be a priority for us, and our focus on accruing and attracting and retaining the top talents will continue to be our focus. So the pattern, there is no surprise and there is no swift change and we'll continuously to invest in R&D.

The other income line for this quarter is related to spinoff assets that the transaction itself generated gains, and so we do have, you know, some of the business like Baidu Video that was spun off, and because of that transaction, we do have other income line that reflects the gain of the transaction.

For cash, obviously as we continue to look into the future, we're very excited about the AI age and the opportunity it can bring. Many applications can be born out of these technology competencies, and our cash is a strategic asset for us to invest in the future. So, currently we don't have other plans to deploy cash other than for strategic purpose and for organic growth.

The next question comes from the line of Natalie Wu from CICC. Please go ahead.

Hi. Good morning, management. Thanks for taking my question.

So, regarding your feed ad product, can management share with us more color on this product, say, the current revenue run rate and ad load target for 2017 or longer future? Also, is it combined within the existing advertising system? Is the rebate rate for agencies in alliance with your search product, and et cetera? Thank you.

Are you talking about news feed?

Yes, news feed in the Mobile Baidu app.

Right now we just began the monetization process. So the ad load is relatively low. I think it's much lower than industry peers. I mentioned previously that the current mechanism, it's an opt-out mechanism for Baidu advertisers, so they share the same mechanism of rebate or channel policy, things like that. So we just got started. I think that the growth rate is very high but the ad load rate is very low right now.

The next question comes from the line of Piyush Mubayi from Goldman Sachs. Apologies. Please go ahead.

Thank you. Thank you for taking my question.

Could you talk about iQiyi's performance across advertising and subscription businesses during the quarter, as well as their content spending plans for two years, and if possible, the path to profitability for that business?

And second, would it be possible at all to talk about the split between medical and non-medical through the last quarter? And when you talk about potential recovery into 2017, how much of the merchants overall can be expected to be coming back? Or put it another way, of the 70,000 decline that we saw in the last quarter, could you give us a sense of what the split was? Thank you.

Thank you, Piyush. For the iQiyi business, both the advertising business and subscription business are growing nicely, and I think perhaps with a little bit of that in the breakout of the revenue line. Content strategy continues to keep priority for iQiyi as it carries out solid growth in advertising and subscription. So as I mentioned at the beginning of the year, the content plan last year was kind of at a rate of doubling, and this year you should expect a similar rate, and that plan has not changed. Going forward, I think the investment in content will continue to be the key and over the years iQiyi has built a core competency in both content identification and also self-production content capabilities.

So I think we're very encouraged by iQiyi's progress, and because of also the subscription user business growth, I think its financial performance is improving significantly. So, very encouraged by the iQiyi development.

For the customer base, yes, as we're going through last quarter, for example, of all the customer validation process, medical and financial services were the customer sectors that were mostly affected. Obviously the effect is across the board to all the customer base. As I mentioned earlier, the medical and financial services sectors have pretty much completed their customer validation work and the Q4 actual work is related to across the board or other sectors.

So we expect the medical sector is steady and slowly recovering, and the other sectors may be a much-affected in Q4 but going into next year, we should start to build the customer base again. Because of this really strict control and regulatory review from our perspective, some of the customers will not be able to do business with us, and so there will be a portion of the customer base that used to be doing business with us will be removed.

And of course, we think the potential advertising customers, the pool is big and we will provide unique advertising platform services and products and combined with the news feed, a kind of new inventory that we have, we have more product portfolios that we can service for customers. So the short answer is that some customers will be cut out and that there will be plenty of new customers and we'll continue to build the customer base going forward.

I would just add that there is also a macro trend with the new normal, we see a lot of consolidation across many different industries, from medical to education to finance, you name it. But this may not be a bad thing for us and I think when the industry is more consolidated they will care more about their brand and have a more systematic way of evaluating the effectiveness of their advertisement.

So we expect to actually benefit from this. But the number of customers driver may be affected because of this macro trend.

The next question comes from the line of Juan Lin from 86Research. Please go ahead.

Hi, good morning, Robin and Jennifer and Linda. Thank you very much for taking my question.

I have a follow-up question on the news feed ads. You mentioned that you will leverage the existing advertiser base. I'm wondering whether the news feed product will start generating incremental revenue right away, and advertisers will increase their budget on Baidu because of the news feed ad, or whether the growth of the news feed ad at the early stage is mainly due to advertisers shifting their search budget to news feed ads.

And also, when do you expect these new products should become a meaningful revenue contributor? Thank you.

I think I talked about the nature of search business quite a few times. I think it's always traffic-bound instead of budget-bound, meaning that many of our customers or advertisers, they have the budget. They just cannot buy enough amount of traffic on our platform.

And news feed provides a perfect incremental traffic for our advertisers and so far, based on our existing testing, I think advertisers are pretty happy about the performance of news feed ad. So I think the revenue opportunity here is very much incremental and we do expect a meaningful contribution for next year.

The next question comes from the line of Wendy Huang from Macquarie Capital. Please go ahead.

Thank you. My first question is about traffic acquisition cost. So that declined in both dollar terms as well as the percentage ratio. So how should we expect the TAC ratio to trend going forward? And also, can you give us an update on the gap between the mobile and PC in terms of the CPM and where it is standing now?

Lastly, on the SG&A cost, assuming there is no change with the O2O competitive landscape, how should we expect the SG&A cost in dollar amount in 2017? Given that currently we are seeing the sequential decline, should we expect this kind of sequential decline to continue into 2017? Thank you.

Hi, Wendy. For the TAC cost, Q2 obviously was a very special quarter. I mentioned last time, in Q2, it was very much the medical-related revenues were affected and much of that was organic. And so you could see that in Q2 it was a very pronounced increase of the TAC rate.

In Q3, other than that we're implementing the measures across the board, and some of the other verticals are also of course affected by this and the TAC ratio was not only because of the medical or organic revenue and therefore it fluctuates a bit.

Directionally, over time, you should expect the TAC rate to continue to increase as a percent of revenue, simply because we're growing on one, the mobile contextual-related revenue and much of that is yet to be fully developed. And that generally pays a higher TAC rate and if they bring a bigger contribution to the overall revenue, we should expect TAC rate to go up.

For the mobile CPM, obviously it continues to improve, but I think at this moment of time that the comparison may not be very meaningful because much of the changes it is going through. PC, obviously we're changing the screen display methods and so there are a lot of other variables that's moving around to affect. But generally of course, we continue to believe that mobile has tremendous monetization power and would continue to improve its capability and an area -- and eventually we think it surpasses the mobile monetization capability.

For SG&A expenses related to O2O, we mentioned earlier that we have continued improvement or thinking or revelation in the O2O area. We want to connect people with services and as we are executing on this business, we're very much focused on the user value proposition and merchant value proposition. And of course, at the same time deploy our SG&A investments in a very disciplined manner.

Our three-year plan of Nuomi for O2O has not changed but I think as I mentioned earlier, we are approaching the business in a very disciplined approach and a focus on ROI added value generation to our users and merchants.

The next question comes from the line of Erica Werkun from UBS. Please go ahead.

Morning, Robin and Jennifer. Thank you for taking my questions. This is Ming Xu asking on behalf of Erica.

So I have two follow-up questions on news feed. The first question is, could you clarify, in terms of DAU, is it 17 million or 70 million? And also, could you comment on the time spent trend?

Secondly, in terms of technology, how do you differentiate from the competitors? So in particular I think one thing is the AI technology that Baidu has. The other thing is we know that you invest in Taboola, the recommendation engine for the US. So how would these unique technology advantages help you to differentiate from your competitors in terms of recommendation? Thanks.

Ming, in terms of the DAU yield, that's 7-0, 70 million. In terms of time spent, I mentioned it has tripled since our launch back in May and is still growing steadily. So users stickiness is very good.

You are right that the news feed product is essentially an AI product. We tailor the content for everyone on a very personalized basis and we leverage our AI technology in the form of machine learning and many other different AI technologies to make sure that we will feed the users with the most relevant, most interesting, high-quality content.

And also, for the monetization, it goes without saying that we are leveraging on a very large advertiser base and we also have the industry-leading targeting technology based on the user profile, location, their credit history, things like that. So we are able to provide very diversified and creative sources for the users and for the advertisers. So we think we are best positioned to win this game.

The next question comes from the line Alan Hellawell from Deutsche Bank. Please go ahead.

A basic question. Just what is the breakdown in GMV between food delivery and group buying? And then secondly, over the past many years, whenever you've introduced a successful innovation -- whether you move into travel or mobile app stores or other refinements, we've been able to kind of quantify that breakthrough.

And I'm just wondering, now that we're several quarters into leadership in AI, whether you can give us any sense whether it's CPC or CPM basis. The difference of purchasing a keyword that kind of benefits from AI versus one that would have been produced from your previous algorithm. Could you try to capture that improvement?

Alan, for your question on the GMV breakdown, the larger GMV continues to be the Nuomi business. The takeout food delivery service at a frequency level is much higher and is really growing very nicely. It's narrowing the gap but the biggest is still Nuomi.

And on how AI has helped our CPC and CPM, we started to invest in AI technology about five years ago and starting from about three years ago we implemented AI-related technology to help monetization. It has contributed to CPC and CPM and click-through rate for quite some time, I would say at least for the past about three years we're seeing very significant contribution from our AI technology on monetization and I believe there is still lots of room for improvement on that direction.

The next question comes from the line of Thomas Chong from BOCI. Please go ahead.

Hi, thanks for taking my questions. I have a quick question about the top 5 advertising categories. If my understanding is correct, should I expect the contribution from healthcare to increase from third quarter to fourth quarter, while some other sectors should be on a declining trend in the fourth quarter? Thanks.

The top sectors, just for your information, are the medical services, the local services, education, retail and real estate and home renovations. I think as you can see, even going through these large changes we continue to have the relatively steady, the top sectors, that remains largely similar to prior quarters.

Yes, for what you just said, I think I mentioned much of the review work for these verticals like medical will have finished and they should gradually resume some of the business that these advertisements can do with us. Other sectors will be going through the same exercise that we have been doing.

The next question comes from the line of Eric Wen from Blue Lotus. Please go ahead.

Hi, good morning. Thanks for taking my questions.

I have two questions, first a short follow-up on news feed. Is it mostly branded ads or did we also see effectiveness with local ads in this media format?

And my second question is more open-ended. As the Company shifting to more fundamental technologies like autonomous driving and AI, given similar companies like iFLYTEK have much lower revenues than us, what's our thoughts on our business model going forward, taking into account our past experiences and lessons learned with search? Thanks.

Okay. On the news feed, right now I think super majority of the advertisements are performance-based ads. We share the same advertiser base with Phoenix Nest or search ads.

But we do see a number of advertisers who are more brand-oriented, especially for those advertisers who aim to target a specific geographic location, for example in real estate, that they would like to target a specific geographic location. Sometimes you can see that they are also performance-driven, but sometimes they are looking for branding effectiveness too. So I think overall news feed is still pretty much performance-based advertising, but we'll increasingly add some flavor of brand ads.

On AI technology, I think it has helped almost all of our existing products, helped monetization, helped to deliver better user experience in search, in many other news feed and many other content-based products. But also I think for our newer initiatives, be it Internet finance or autonomous driving, we've seen significant contribution from AI-related technology.

I think that will position us uniquely in those very large markets. But the newer initiatives generally take longer to see material impact on revenue and earnings, so we'll need to wait for maybe a couple more years.

We are now approaching the end of the conference call. I will now turn the call over to Baidu's Chief Executive Officer, Robin Li, for his closing remarks.

Well, again, thank you for joining us today. Please do not hesitate to contact us if you have any further questions.

Thank you. That does conclude the conference for today. Thank you for your attendance; you may all disconnect.