Baidu, Alibaba, Tencent: AI Showdown
Takeaways from BAT earnings and announcements
Disclosure: My writing is from choosing and piecing together brain cells, investment bank sell-side reports, international & HK media coverage, the interweb of information, earnings transcripts and reports. I’m not a financial analyst, and I’m definitely not here to help you pick stocks. I’m purely writing about BAT’s earnings highlights because I’m interested/ curious about how China’s big techs are investing/ adopting/ building AI.
Overall Comments - Commonalities Across BAT
So these things happened this week:
11/12 Baidu World 2024;
11/13 Tencent 2024 Q3 Earnings;
11/15 Alibaba 2024 Q3 Earnings
In comparison to the US big tech’s earnings two weeks ago, the BAT seemed to not have focused on AI as much during these public-facing addresses. Many of the analyst questions seemed to want to get more from the management on each of their comprehensive AI strategy.
The FT wrote in August that Chinese Big Tech are finding ways to go around US restrictions and are buying GPUs aggressively. So far, no one has mentioned anything about infrastructure hurdles (unlike Zuckerberg during Meta’s earnings) so seems like my point about China having infrastructure advantage still holds.
So a fact is, they definitely are all doing what they can to push ahead AI integration, AI infrastructure development and even AI hardware exploration. With Baidu more focused on application innovation, and Alibaba more focused on infrastructure, but so far, still nothing WOW. Scroll down to see detailed breakdown takeaways from each company.
High-level Takeaways
Overall hype on AI continues but nothing with a WOW factor
Application on large-model technologies still lag compared to US counterparts
BAT still leading on large-model in China
Baidu, Alibaba’s AI agents are expanding its reach across industries
Innovation in hardware is drawing attention
We’re Back Baby: Baidu World 2024
For a while, Baidu seemed to have lost its relevance in China’s mobile internet era. Many said ByteDance overtook the “B” in “BAT” but at Baidu's World 2024 event, held in Shanghai this week, Baidu showcased significant advancements in artificial intelligence (AI). So I guess we can say, Baidu is back!
Founder and CEO Robin Li is ALL about applications. His strategic vision is that intelligent agents will become the primary interface for delivering content and services in an AI-driven world. Baidu (BIDU) is not pursuing a "super app" model but instead aims to empower users to create a multitude of practical applications through its platform. And this is consistent with he talked about at WAIC as well (see op-ed I wrote about attending WAIC here, for Fortune)
Source: Baidu data
Here are the key takeaways from the conference:
the average daily API calls of Ernie LLM saw significant growth over the past 6 months: Wenxin large-model application daily usage had exceeded 1.5 billion, growing by 7.5x compared with data six months ago.
iRAG is important to resolve the industry issues on illusion in order to provide high quality and accurate image and video content;
Four types of AI agents (Baidu AgentBulder) were introduced – corporate, Digital human, Tools and Industry;
Baidu introduced MiuTa, which aims to facilitate AI agent development without programming expertise.
Let’s break it down.
A. iRAG Technology Launch: key for higher quality image and video generation
Image-Based Retrieval-Augmented Generation (iRAG): This new technology aims to tackle the issue of "image hallucinations" in AI-generated content. By enhancing the realism of images produced from text prompts, iRAG reduces production costs and improves workflows for content creators. It leverages Baidu's extensive image library and integrates with LLM, to create ultra-realistic visuals suitable for various applications, including comics and marketing materials.
Baidu highlighted the number of daily API calls increased by 7.5x over the past 6months to over 1.5bn. With continued upgrades on ERNIE, the user experience has also been improving drastically while the cost has been brought down.
B. Miaoda (秒哒): AI agent “no-code” development tool
Baidu’s Qianfan platform has over 33k fine-tuned LLM models and 770k applications used in various use-cases/ sectors.
Based on this, Baidu introduced this no-code development platform designed to democratize software creation. Users can build complex applications using natural language, without any programming skills. It features multi-agent collaboration and integrates various tools, allowing users to manage projects and automate tasks effectively.
C. Roll out of four types of AI Agents:
Baidu’s AI agent platform has attracted 150,000 enterprises and 800,000 developers. Looking at broader industry application, Qianfan LLM platform has fineturned 33,000 LLMs and developed 770,000 enterprise applications.
There are four types of AI agent: (1) Corporate; (2) Digital human; (3) Tools ;(4) Industry
The company offers developers low entry barrier to AI tools with strong functionalities.
(1) corporate agents offer users good recommendations and personalized responses. It helps corporates to enhance conversion and interaction.
(2) digital human has emotions and attitudes with more natural interaction as live streaming hosts.
(3) tools agents can be used across different sectors. The new product can enable users to create industry analysis reports based on private/public information and tools for editing, creativity and sharing. [A free Canva-like tool]
(4) industry agents cover legal, financial, medical, travel and sports sectors.
D. Xiaodu AI Glasses:
It combines large models with hardware (they kind of look like classic raybans). The glasses are 45g in weight and have 5h long battery life, ultra wide-angle 16-megapixel camera embedded. And it will be the first AI-native model to be embedded with Chinese LLMs in the market. They can perform tasks such as real-time translation, content summarization, and music curation based on the user's environment. The firm expects the product to be officially launched in the first half of next year.
Tencent Earnings: Gaming is King, AI meh
Chinese social media and gaming company Tencent (0700.HK) on Wednesday reported better-than-expected profit in the third quarter ended Sept. 30, spurred by growth in games, advertising and cloud services.
Management highlighted that the company is strategically investing in AI, increasingly integrating AI across its product offerings, enhancing various services including marketing and cloud solutions. Although founder and CEO Pony (Huateng) Ma tried to ride the AI wave and opened up the Q&A floor with:
“We are increasingly seeing a tangible benefit of deploying AI across our products and operations, including marketing services and cloud. And we'll continue investing in AI technology, tools, and solutions that assist users and partners.”
All in all, it was a little lacklaster. Anyway, here are some key takeaways:
Cloud revenues mostly from GPU-powered services grew year on year and represented 10%+ of IaaS revenues.
Tencent cloud saw significant revenue increase in its international business, mostly attracting new customers from its gaming and live streaming touchpoints. Company noted that China does not have a big enterprise market and fewer AI startups when compared to the US, hence AI driven cloud revenue is not THAT significant.
A. Tencent Huyuan Turbo:
An upgraded foundation model utilizing a heterogeneous Mixture-of-Experts (MoE) architecture, significantly improved training and inference efficiency. This model reportedly doubles the efficiency of its predecessor while able to half inference costs.
B. Content & Marketing & Search:
In terms of future monetization of AI, the management team seem to believe it is essential to improve (Weixin/ WeChat) search, content and ad recommendation across platforms, to further improve user experience. Future AI adoption could potentially unlock more revenue.
C. User Engagement:
The gaming sector remains a strong focus for Tencent, with ongoing development of "evergreen" games that leverage AI for improved user engagement. The company is also exploring new game releases that can utilize advanced technologies to enhance gameplay experiences.
Alibaba Slow and Steady Wins the Race
Alibaba's latest earnings announcement mostly focused on its ecommerce business as the Double 11 Shopping Festival just concluded this week as well. The company saw a healthy high single digit growth (~5%) in revenue, mostly in line with analyst estimations. During the call, management highlighted a significant developments in the area of AI.
Eddie Wu, CEO of Alibaba (BABA) said that the company’s core business segments maintained steady growth guided by “user-first AI-driven strategy.” And emphaized that the firm will continue to optimize its integrated cloud plus AI development strategy. [See my deep dive on Alibaba’s AI strategy here]
“We believe the AI era is just beginning and we're just still in the early stage of AGI transformation. Looking ahead, AI's potential extends beyond mobile screens. It's poised to reshape the digital world and ultimately transform all industries in the physical world. As a leading cloud service provider for AI in China, we will continue to invest in advanced technology and AI infrastructure while optimizing operational efficiency. This will enable us to deliver more reliable and cost-effective AI technologies and products across industries. We believe that as AI penetration grows, Alibaba Cloud's cloud computing and AI-related products will become the foundational infrastructure that supports development across industries.“
— Eddie Wu, CEO of Alibaba Group
Key highlights related to AI propelled Cloud growth:
Source: Alibaba Group Q3 Earnings Material
A. Triple-Digit Growth in AI-Related Products:
Alibaba reported that revenue from AI-related products has experienced triple-digit growth for five consecutive quarters, reflecting the company's strong focus on enhancing its cloud services and AI capabilities.
Alibaba Cloud revenue grew 7% year-over-year, with adjusted EBITDA increasing by 89% to RMB 2.7 billion, driven by a shift towards higher-margin public cloud products and improved operating efficiency.
Enhanced Marketing Efficiency: The AI-powered marketing tool, Quanzhantui, saw increased merchant adoption, improving marketing efficiency and expected to drive higher spending on Alibaba's platforms.
B. Investment in AI Infrastructure:
The company launched GPU container services and upgrading AI servers. These enhancements aim to improve model training and inference efficiency, making advanced AI technologies more accessible to customers and supporting the growing trend of cloud adoption for AI applications.
Two key words to close: Integration and Commercialization
All three companies reported substantial increases in revenue from AI-related products. Baidu noted its Ernie model handling over 1.5 billion API calls daily, while Alibaba highlighted triple-digit growth in AI product revenue for multiple quarters. Tencent is also leveraging its Hunyuan model to enhance various applications, particularly in gaming and social media.
Each company is investing heavily in improving their AI capabilities. Alibaba is enhancing its cloud infrastructure to support AI applications, including launching GPU services and upgrading AI servers. Baidu is deploying AI across more than 100 industries, and Tencent is focused on integrating AI features into its existing services, showcasing a collective commitment to advancing their technological frameworks to support AI adoption.
Innovative AI-written articles. A wonderful share.