If you’re new here, I am a fan of the Coffee Can Portfolio, an “Active Passive” approach to investing. The idea of a Coffee Can is simple: Buy a basket of the best stocks you can and let them sit for years. You incur no costs with such a portfolio, and it is simple to manage.
Coffee Can Next Results To Date:
If you’ve been following along, we’re still at 4 investments:
That said, I’ve short listed a few ideas for the 5th and final investment in this Coffee Can. More to come soon.
Did China just become “investable” Again?
Chinese stocks have been hated for a while. The peak to trough KWEB 0.00%↑ price decline (from Feb ‘21 to Oct ‘22) was over 80%. Not only that, but since 2022, we have had multiple opportunities to purchase Chinese stocks at 2013 prices! Sentiment has been ugly.
I just listened to a Bloomberg Intelligence Podcast titled “Can China Rescue its Stock Market and Economy?” It was published in February 2024. If you listen to it, you’ll notice that the conversation is extremely pessimistic. At one point, the host even says “Is there anything positive about the China economy?” to which the interviewee responds by laughing.
As investors, we bet on possible future states of the world. So just because people are bearish now, doesn't mean they will always be so. Therefore, as investors we must convince ourselves that current pessimism is temporary.
Go figure… KWEB is up more than 30% since its February low (see chart below).
Of course, this is a timely coincidence. But as investors, we must constantly remind ourselves that bears always sound smart, and there is always something to worry about, but remaining optimistic and rational is very important.
On the bright side, Chinese stocks have been extremely strong lately. And now we may know why… You see, as 13Fs came out last week, Appaloosa Management (David Tepper’s Fund) and Scion Asset Management (Michael Burry’s Fund) both revealed over 20% allocations in Chinese names. Several other high-profile investors also revealed stakes in Chinese companies. This is big news. This may just be the catalyst money managers have been waiting for, to invest in Chinese companies once again.
I believe Chinese stocks are still under appreciated, so this bodes well for our KWEB investment.
As I wrote before, although I invested in KWEB, I’d prefer to transition my investments into individual companies, as I become more familiar with them. To that end, I wanted to jot down some notes on Baidu.
Notes on Baidu
I’ve always known Baidu to be China’s leading Search Engine. However, the business is so much more. It is clearly one of the leading AI companies in China.
Have you ever noticed that the acronym AI appears right the middle of the name b(AI)du? Hmm…what a coincidence…
2024 Q1 Earnings Takeaways:
The Macro Environment Continues to be Challenging
Core search is being rebuilt with GenAI
The company believes they have the best AI Cloud
Their Robo-Taxi business is growing quickly and approaching breakeven.
The Stock is Hella-Cheap
1. Macro-Environment Continues To Remain Challenging
The majority of their advertisers are SMEs, and therefore their advertising revenue is sensitive to the macro environment, particularly to the offline economy.
The weak macro environment contributed to the softness in their ad business. The downturn in the real estate industry had direct and indirect impact on their ad business:
“Ad spending from developers and agencies was muted, but the impact also extended to both upstream and downstream sectors. For example, energy, chemicals, machinery, building materials for upstream and home renovation, furniture, which is kind of downstream, all saw constrained ad spending on our platform.”
2. Core Search is being Rebuilt with GenAI
Baidu believes GenAI complements traditional search, and expands the total addressable market.
Baidu believes Search will be one of the most likely killer apps in the GenAI era. That said, it’s still early days. Only ~11% of search results are currently filled with GenAI results, and these are not being monetized yet.
It will be interesting to see whether Gen-AI based search cannibalizes, expands or kills the Search business as it exists today.
Baidu has been reconstructing Search with ERNIE (Enhanced Representation through Knowledge Integration). ERNIE is a LLM-based AI chatbot service created by Baidu.
ERNIE helps Baidu generate realtime search results in a growing variety of formats like text, image, third-party links, points of interest, and citation.
ERNIE has also helped refine Baidu’s ad auction system, automate creative generation, and ad strategy formation. Advertisers have been seeing better conversions and leads, leading to increased spend with Baidu. As a result, AI-related incremental advertising revenue grew on a quarter-over-quarter basis, and is expected to continue growing.
GenAI has generated several hundred million RMB per quarter in incremental revenue. That’s a positive sign but still very small… single digit percent of total revenue.
Open Question/Concern:
How much share is Search losing to video-based discovery platforms like TikTok?
3. AI Cloud
GenAI and foundational models like ERNIE are transforming the cloud industry from general purpose computing to AI computing.
AI Cloud revenue reached RMB 4.7 billion (~$650mm), up 12% year over year.
Baidu believes they offer China's most efficient AI infrastructure and the most advanced MaaS platform for model training and inference. More and more enterprises are choosing Baidu for model training, fine-tuning, and AI-native app development on their public cloud.
Since H2 2023, Baidu’s AI cloud revenue growth has started to accelerate from a year-over-year decline in the Q3 2023 to an 11% increase in Q4 2023, and then further acceleration to 12% in Q1 2024.
Long term, normalized margin for GenAI and foundational models related business is expected to be higher than traditional cloud business. In addition, Baidu is also ramping down low-margin businesses.
ERNIE:
Has increased its training efficiency to 5.1 times (unclear what metric this is?)
Has reduced Inference cost to only ~1% (aka 1/100th!) compared to the March 2023 version.
Has 3 Tools available for Developers:
AppBuilder
ModelBuilder
AgentBuilder
Has API distribution partnerships so that ERNIE-enabled applications can become an entry point into their Generative AI Ecosystem:
Smartphone brands (Oppo, Vivo, Xiaomi, Samsung China and Honor) to enhance UX
Lenovo to enhance default browser AI assistant.
Nio, China's leading smart EV manufacturer, is using the ERNIE API to enhance the encamping experiences for its vehicles.
AI Agents:
Baidu believes AI Agents will become one of the most important types of applications powered by GenAI and foundation models.
With the ability to use natural language as the programming language, developers will be able to build AI agents without the need to write any code.
This is very exciting
For Example:
Advertisers can create customized ERNIE agents. When advertisers express their intentions to these agents, they can more effectively achieve their goals, whether it's helping potential customers understand their products or improving customer service quality.
This should lead to better conversions and ROI. (Time will tell…)
AI Chips:
Baidu claims they have recently made a breakthrough by integrating GPUs from different vendors into one large scale, unified computing cluster, allowing them to use less-advanced chips for highly effective model training and inference.
If the above is true, that could be a huge competitive advantage. Recall this is precisely what Google did in their data centers (which led to one of their competitive advantages): They were the first to leverage commodity hardware instead of expensive servers, at scale.
Baidu believes that ultimately China will have its own ecosystem of less powerful chips, but a more efficient “homegrown” software stack.
4. The Robo-Taxi Business
I had no idea Baidu even had robo-taxis…But Baidu’s autonomous ride-hailing platform is called Apollo Go. And they’ve already provided an impressive 6 million autonomous rides to the public! (826k in Q1).
Notes about Apollo Go:
Its Commercial operations were kicked off in Wuhan in 2022.
Apollo Go is now operational 24/7!
Their fully driverless vehicle fleet in Wuhan is now 300 vehicles. Expected to reach 1000 by year-end.
The operational coverage area for fully driverless ride-hailing service increased by 8X from a year ago, now covering over 7 million people in Wuhan.
Both, daily rides per vehicle and distance per ride, have been growing.
The main cost drivers of this service are (1) labor and (2) hardware expenses. Both should trend down over time.
Note: Current cost of vehicles excluding battery is <$30k
Apollo Go Strategy:
Improve operational efficiency and reach breakeven in Wuhan first.
Then scale up operations elsewhere, quickly.
License their vision-only autonomous system to OEMs (available for 100+ cities).
5. The Stock is Hella-Cheap
The stock trades at whopping EV / FCF of ~5x!
I think it is highly unlikely that an AI company such as this would trade at such a valuation if it were not based in China…