Baidu, Buy 1 get 2 Free (Part II)

Previous: Part I: Baidu’s Lost Decade (History)

Uncertainty for large Chinese tech companies has created an abundance of fear in investors. Following the uncertainty, there has been an exodus of institutional investors in fear of failing unconventionally. Catchy one liners like “China is uninvestable” or “VIE structures own no assets” are simple narratives trying to explain a complex situation.

Great business like search engines rarely go on sale. There are only four countries in the world that has developed their own search engine: US, China, Korea, and Russia. In China, Baidu controls 76% of the search market and is the fifth most visited website in the world. 

From a peak market cap of $110 billion falling to a low $50 billion due to external forces, investors should examine Baidu closely. For $56 billion you are buying China’s number one search engine that generates $3.4 billion operating profit (TTM), a cloud businesses growing 71%, leader in AI, and $25 billion in cash and short-term investments. 

Baidu has been losing online advertisement market share from 35% in 2015 to about 15% in 2020, while new short video platforms like Douyin, Kuaishou, and Bilibili have been increasing market share.

COVID-19 hit Baidu the hardest in 1H2020 as online marketing revenue lowered 15% while competitors increased in revenue. Many of Baidu’s off-line customers stopped advertising because stores were closed to prevent the spread of the virus.

The core of Baidu search is to generate leads into sales from mostly higher priced off-line businesses like health care, travel, financial services, and auto. Competitors like Alibaba, Tencent, and ByteDance advertise smaller ticketed items, which grew from an increase in online shopping.

While Baidu was posting lower online ad revenue due to COVID lockdown in 2020, metrics like DAUs, search volume, and feed time increased. More users were using Baidu app to get information. The short-term revenue drop is not a permanent impairment to the search business.

A fundamental difference between search and social media is user experience. Social apps want users to spend as much time as possible to increase ads and earn higher revenue. Baidu wants the opposite. Users to spend the least amount of time as possible searching. User experience severely declines when spending five minutes to search for China’s national sport. (Ping Pong)

When Tencent unveiled WeChat search in 2017, headlines of Baidu going extinct were rampant. Baidu would be not match for WeChat’s 1.2 billion MAUs. Since then, Baidu has been able to defend mobile search with 95% market share in 2021.

China regulators have ordered large tech companies to open up their platform creating a more open internet. As more information proliferates the internet, search engines become more important to connect the user to the right information.

Search engines has been around for over 20 years. Is there anything left to improve? 

Yes!

Baidu has developed RocketQA, the next step where search evolves from keyword matching system into question-answering machine. The next leap of search technology would be if you could ask “what do i need to prepare to climb Mt. Everest in May?” The results would be if you were talking to a seasoned mounting climbing expert. AI would accumulate a list of all the gear you would need, physical training required, and important tips that a beginner would never think of. 

Even after 20 years, Robin spends 50% of his time improving search.

Improving search is a never ending quest of tiny improvements.

Revenue in Baidu’s online advertising reached $11.1 billion in 2020, decreasing 9% year-over-year. In Q2 2021, online advertising reached $3.2 billion revenue, growing 18% year-over-year.

Online advertising in China has grown 32% annually from 2010 to 2020. COVID-19 dampened growth in online advertising in China during 2020 to $75 billion. eMarketer estimates online advertising to grow at a 13% CAGR from 2020 to 2024 reaching $121 billion. China is still one of the fastest growing online advertising countries in the world. 

Baidu’s business is split into Baidu Core and iQiyi.

Baidu core are internally grown businesses which can be split into two groups— online marketing and non-advertising.

Baidu’s mobile ecosystem is a super app, with over 600 million MAUs. It is an open platform that allows search, feed, and content inside Baidu App. Users can experience third party apps with native-app-like experience without ever downloading anything through Baidu Smart Mini Programs.

Management is focused on switching users from the web browser search to Baidu App for a better user experience and higher monetization rate. 

Baidu earns revenue by: CPC (Cost-per-click), CPM (Cost per thousand impressions), CPI (Cost-per-install), and CPR (Cost-per-registration). Advertisers can choose to advertise on search-based and feed-based online marketing services.  

Two products that strengthens Baidu’s ecosystem are Baidu Smart Mini Programs and Managed Page.

Smart mini programs are mini applets, usually less than 10 MB, that work inside Baidu App. Users do not have to download apps they would rarely use to experience the native-app like feel. 

Higher traffic and retention are what draw app developers into Smart Mini Programs.  

Baidu Smart Mini Program was copied from WeChat in 2018 and surpassed over 416 million MAUs by Q1 2021. Baidu has taken a more open approach by allowing WeChat mini programs to easily transfer into Baidu mini programs; conversely, Baidu mini programs cannot transfer into WeChat mini programs. Additionally payment platforms like Alipay and WeChat Pay are both available. The open approach is better for the long term health of Baidu’s ecosystem. 

Managed Page is a hosted mobile alternative that improves load times under MIP (Mobile Instant Pages) standards. Baidu provides an industry specific template with tools for users to easily migrate their HTML5 website into Managed Page. The customer benefits with higher traffic and conversions, no maintenance or bandwidth costs, and access to Baidu’s AI.  

The benefit for Baidu is more control in their ecosystem and improved products through user insights. Managed Page was introduced in late 2018 and quickly adopted to make up 40% of core marketing revenue by Q2 2021. Over 300,000 Managed Page customers are using Baidu marketing cloud service.

Non-advertising revenue is the growth driver for Baidu. It contains AI cloud, autonomous driving, DuerOS, and smart products.

Revenue in non-advertising was $0.8 billion in Q2 2021 growing 80% year-over-year and made up 16% of revenue. Management expects non-advertising business to grow three times faster than online marketing at 30%.

In Amazon’s 2006 shareholder letter, Jeff Bezos describes how he thinks about growing new businesses.

High returns on capital in very large markets that addresses an underserved market; where we have the capabilities needed to bring strong customer-facing differentiation to the marketplace. We must convince ourselves that the new business can grow to a scale where it can be significant in the context of our overall company.

Patience is irreplaceable when growing new businesses— especially in China. Competition is fierce. What would usually take 5-7 years to meaningfully grow a business in the US, it would take over a decade in China.

Baidu’s new businesses like AI and autonomous driving are still seedlings. Discipline, patience, and a nurturing culture will be required to grow them into well-rooted young trees.

Baidu is the AI leader in China for three quarters in a row according to IDC. Over a decade of investments and R&D has allowed Baidu to be on the forefront of AI.

Andrew Ng, AI scientist who co-founded Google Brain and built the foundation for Baidu AI, speaks highly about the future of AI.

AI can be categorized into two groups— weak and strong. Currently we are deploying narrow (weak) AI to solve industry specific problems. For example, Deep Blue’s algorithm that was used to beat the world’s chess champion, Garry Kasparov, could not be applied to the more complex board game, Go. New training and data had to be used to create an algorithm called AlphaGo.

Baidu is working more as an integrator role deploying narrow AI to specific industry problems. The initial costs to set-up the projects are low margins or at a loss but can be reapplied throughout the industry vertical. The AI solution of deploying V2X (Vehicle-to-Everything) at major intersections in China to reduce traffic and pollution is costly in the beginning, but the same solution can be expanded to new cities at higher margins.

It is difficult to imagine how much AI will impact the world, but Baidu is in good hands with Robin Li being one of the best AI CEOs in the world.

The cloud industry has taken what used to be a fixed expense, buying and maintaining servers onsite, to a variable expense of pay-as-you-go model. Startups and even mature companies have switched to the cloud to save costs and time.

Baidu’s AI open platform is built on top of Baidu Cloud to allow anyone access to AI capabilities. Small startups who do not have access to the best data scientists or compute infrastructure can build on top of what Baidu has already figured out.

Baidu AI Cloud revenue was RMB3.3 billion in Q2 2021 growing 71% year-over-year, faster than the overall cloud market. Overall cloud growth in China was around 54% expanding Baidu’s market share to 8%.

China’s public cloud market in 2020 was $19 billion or 13% the size of U.S at $146 billion. In the past five years China’s public cloud grew at 61% CAGR compared to U.S. at 24%. There is still enormous growth and runway left for the cloud industry to develop in China.

The levels of IaaS and SaaS spending are different in China and in the U.S. IaaS spending in China was 62% of the cloud market, while in the U.S. it was only 16%. SaaS spending in the U.S. made up 67% and only 25% for China.

An informative research article “A Comparison of the Cloud Computing Market between China and the United States” by Ma Peng, explains the different levels in cloud spending. He mentions shorter life spans in companies, lower labor costs, skipping IT directly into cloud infrastructure as reasons for China’s increased IaaS market.

Competitors like Alibaba, Tencent, and Huawei will achieve better scale in IaaS. Baidu differentiates itself by providing superior AI through PaaS and Saas. 

Customers who use Baidu AI cloud for the first time usually sign up for one service to test and understand how it can benefit their operation. It can take months to gather enough data to optimize an algorithm to see meaningful benefits. After experiencing efficiency gains on the initial project, customers make follow-on purchases for their whole organization.

Baidu partnered with Geely, a leading automotive manufacturer, to help build a multiyear digitalization plan. The partnership will use Baidu’s cloud experience to build Geely’s private cloud infrastructure. All of Geely’s operation will take place in the cloud and will use Baidu AI for high-tech manufacturing capabilities.

Baidu has opened the door for Chinese companies looking to build their own private cloud network and being less reliant on public cloud.

Public cloud revenue in China is projected to be $100 billion by 2026. If Baidu can differentiate from cloud competitors with AI and maintain 10% market share, it will be equivalent to its online ad business revenue in 2020. 

Baidu’s mission is “to make the complicated world simpler through technology”. One company cannot solve AI by itself. Baidu is democratizing AI to accelerate the world’s adoption of AI.

Apollo, Baidu’s autonomous driving division, was founded in 2013. Since the beginning, the strategy to create Apollo as a multi-directional autonomous driving platform was to gain scale and data advantage. Baidu has partnered with OEMs to provide driving software, created a JV with Geely to manufacture smart cars by 2024, and unleashed robotaxis in five cities.

Management estimates by 2025, total cost of robotaxi miles will be lower than a taxi driven by a human. In June, Baidu introduced its fifth generation robotaxi, Apollo Moon, at an estimated cost of $75,000, one third the cost of competitor vehicles. Apollo Moon reduced the cost per mile by 60% from the previous generation and an average decrease of 62% in the first four prior generations.

By 2025, the intersection between the cost of manned vehicle and robotaxi will see an acceleration of robotaxis on the streets.

China is filled with self-driving companies like AutoX backed by Alibaba, Didi, Pony.ai, and WeRide competing to get ahead. Baidu is still number one in autonomous driving in China and that is not reflected in its share price today.

TAM for robotaxi ride hailing in China will reach $224 billion by 2025. Baidu plans to roll out Apollo Go to 30 cities within the next two to three years and is currently operating in five cities today.

Bezos describes the big winners; when given a ten percent chance of a one hundred times payoff, you should take that bet every time.

iQiyi is an online video platform like Netflix but with ads. Members can choose between paying to become a VIP member or watch for free with ads. Popular dramas have the first couple of episodes available to everyone and the rest are behind a VIP paywall.

In 2018, iQiyi IPOed at a valuation of $14 billion with over 50 million subscribers and 2017 revenue hitting $2.7 billion. Today, with over 106 million subscribers and $4.6 billion in 2020 revenue, iQiyi’s market cap has fallen to $7 billion.

There are many reasons for the poor performance. First, revenue and subscriber growth have stalled. iQiyi has remained unprofitable since its launch in 2010. Competitors like Tencent Video and Youko backed by Alibaba have bigger wallets to acquire and produce their own content. Baidu looked into selling its majority stake in iQiyi to focus more on AI and autonomous driving, however, had no bidders due to potential regulation scrutiny. Finally, WolfPack Research released a 37 page fraud report on iQiyi in 2020 claiming inflated revenue and subscriber count.

Baidu would lose $3.5 billion at the worst case scenario of iQiyi going to zero, irrelevant to Baidu’s long-term businesses.

Baidu has $6.6 billion in cash, $19.5 billion in short-term investments (held-to-maturity one year debt securities, mainly deposits in commercial banks), and $15.6 billion in long-term investments totaling $41.7 billion.

Total liabilities is $23 billion.

Baidu is a founder led company with Robin Li owning about 17% and holds 57% voting power. Robin’s life work is in Baidu and he is aligned with the shareholders.

Since 2020, management has repurchased $2.5 billion shares outstanding. In the last quarter alone, $300 million shares were repurchased. The board of directors have approved a share repurchase plan up to $4.5 billion by Dec. 2022.

Share buybacks only create value when purchased less than its intrinsic value. Management is sending a message more value was created repurchasing $300 million in shares than investing more into its growth businesses.

Baidu spends over 18% of revenue on R&D to grow future businesses like AI and autonomous driving. However, research is expensed on the income statement, not on the balance sheet. On a GAAP accounting basis, Baidu’s earnings is lowered compared to if management wanted to maximize near term profit.

Amazon shared a similar story of spending all of its profit into R&D and capital expenditure making it look less profitable in the early days. Analysts who quickly glossed over the financials missed that Amazon was spending all of its profit to grow the future of the business.

Investments in AI and autonomous driving will have taken more than decade to become meaningful to Baidu’s business. Adding back R&D into operating income can seem absurd. But the point is to treat R&D as an investment. Management can make Baidu more profitable if it wanted to, but is focusing on the long-term.

Bruce Greenwald in his book, Value Investing, describes earnings power value as sustainable current earnings of a business divided by the company’s cost of capital. I like to use EPV for Baidu’s core business because forecasting growth is not considered as a part estimating EPV. It would be difficult to guess Baidu’s future cash flow from its AI and autonomous driving business.

R&D and SG&A needed to grow the business are added back into the current operating income. 50% of R&D and 25% of SG&A are added back to EBIT as a conservative estimate of future growth. Then a tax rate of 25% is applied to arrive at EPV at a cost of capital of 12%.

Combine $25 billion in cash & short-term investments, $15.6 billion in long-term investments, less $12.7 billion in interesting-bearing debt, to arrive at a total EPV of $63 billion. Baidu’s market cap today (10/15/2021) is $56 billion.

For $56 billion you are buying Baidu’s online marketing business and getting for free the future developments (AI cloud and autonomous driving) whose precise cash flow would be impossible to measure but significant. In the spirit of Graham and Dodd, this would be considered an investment and not a speculation.

Baidu is a great company at a fair price.