Last month, Aljazeera reported that Ernie Bot has 200 million users compared to Chat-GPT’s 180 million users. In the earnings call, Baidu management stated Ernie Bot handles 200 million queries per day, up from 50 million in December. According to the call: “This considerable growth indicates that increasing adoption of ERNIE can point to strong future revenue potential from model inferencing […] We believe ERNIE ecosystem will over time contain millions of applications […]”
On the fundamentals, Baidu reported 1.2% growth in RMB yet this translates to a YoY revenue decline of (-3.75%) USD on a weaker exchange rate. AI cloud revenue increased 12% as Ernie Bot users and API requests continue to rise. Margins improved sequentially, with Baidu’s management emphasizing a near-term focus on optimizing operational efficiency while maintaining AI-related growth.
Revenue and EPS:
FX exchange rates between the RMB and dollar require nuanced approach for Baidu’s growth rates – the company reported YoY growth in local currency, but given that the RMB has weakened more than 5% against the dollar, US$ revenue declined.
- Revenue in Q1 was $4.37 billion, representing growth of 1.2% for RMB 31.51 billion. This represents a YoY decline of (3.7%).
- Baidu Core revenue was RMB 23.80 billion for an increase of 3.5% YoY. This equals $3.30 billion USD for a decline of (1.5%) YoY.
- GAAP EPS was $2.07 in Q1, a YoY decline of (10.8%). Adjusted EPS of $2.76 increased 17.9% YoY.
- Adjusted EBITDA was $1.14B for a margin of 26%. This represents a nominal decline from $1.19B in the year ago quarter.
Margins:
Margins improved sequentially, as Baidu reiterated a focus on maintaining a healthy operating margin while still investing in its AI growth engine.
- Gross margin of 51.5% was flat YoY yet up 130 basis points QoQ
- Operating margin of 17.4% was up 140 basis points YoY and up 200 basis points QoQ from 15.4%. This equals operating profit of $760 million. Adjusted operating margin was 21.2%, up from 20.6% in the year ago quarter.
- Net margin of 17.3% was down 140 basis points YoY yet is up 990 basis points QoQ from 7.4%. This equals a net profit of $755 million. Adjusted net margin was 22.3% up from 18.4% last year and flat QoQ. According to the call, weighing on overall net margin, Other Income was down 52% to RMB 2 billion due to “a decrease in favorable gain from long term investments, partially offset by the increase in net foreign exchange gain.”
Cash Flow and Debt:
Cash flow is in line YoY yet down QoQ.
- Operating cash flow of $861 million this quarter for a margin of 19.70% is up 100 basis points YoY yet is down considerably QoQ from 30.4% in Q4.
- Free cash flow of $579 million this quarter for a margin of 13.3% was down 130 bps from last year and down 660 bps QoQ.
- The company has $26.6 billion in cash and $11.3 billion in debt.
Stock based compensation was 3.6% of revenue. Baidu returned US$229 million to shareholders since the beginning of Q1 2024, bringing the cumulative repurchase to US$898 million under the 2023 share repurchase program. Per management: “we have consistently repurchased our shares on the market over the past four years, averaging around $1 billion annually. In total, we have allocated around 37% of our free cash flow towards the share buyback progress.”
Key Segments:
Revenue from Baidu Core grew 4% YoY to RMB 23.8B, which equals USD $3.3 billion. In March of 2024, Baidu’s monthly active users (MAU) on the search engine reached 676 million, up 3% YoY. This compares to 4% growth last year and was up 1.3% QoQ. Revenue from QiYi declined (-5%) YoY to RMB 7.9B, which equals $1.1 billion.
AI Cloud Revenue grew 12% YoY to RMB 4.7 billion and delivered “operating profit on a Non-GAAP basis.” AI cloud revenue has accelerated from a YoY decline in the third quarter of (-2%) to an increase of 11% in the fourth quarter, to a 100 bps acceleration QoQ to 12% this quarter. As stated, the revenue growth was “mainly driven by gen-AI and foundation models. In the first quarter, such revenue accounted for 6.9% of total AI Cloud revenue.”
Regarding scarcity of GPUs, the company stated they are creating GPU clusters of hundreds or even thousands of GPUs across various vendors into a unified computing cluster to train LLMs.
Online marketing revenue grew 3% YoY to RMB 17B, which equals USD $2.36 billion. According to management, this will eventually accelerate with AI: “we have been pushing hard to transform the user experience from a traditional mobile product to a generative-AI experience. This transition is ongoing and monetization has not yet started.” It was also stated macro weighed on the results: “In the first quarter, advertiser sentiment in some verticals, such as real estate and franchising, remained weak. Specifically in the real estate industry, not only was ad spending from developers and agencies muted, but the impact also extended to both upstream and downstream sectors.”
Perhaps most importantly, it was stated this will remain soft in the near-term: “As we enter the second quarter, we have not seen improvement in advertiser sentiment. Given the limited visibility for sentiment improvement and paired with tough comps in Q2, our online marketing revenue should remain fundamentally solid but from a growth perspective soft over the next few quarters.”
Non-online marketing revenue grew 6% YoY to RMB 6.8B for USD $935 million.
Apollo Go Rides were up 25% YoY to 826,000 rides. According to a recent statement, the AV unit will be break even by the end of this year and profitable next year. The robotaxis go for $27,697 USD and operating costs will be reduced by 30% once a robotaxi network is established.
For the outlook across key segments, management stated the following: “We expect our cloud business to accelerate and the loss of our robotaxi business to narrow for the rest of the year. We expect mobile business to be soft in the near term and start to recover when gen-AI becomes the new core of our existing products next year.”
More on ERNIE:
Similar to Chat-GPT, ERNIE has a family of models with the flagship models being ERNIE 3.5 and 4.0. Since March of last year, ERNIE’s training efficiency has improved 5X and the inference cost has been lowered drastically to 1% compared to the March 2023 models. The company offers a set of tools, such as AppBuilder and ModelBuilder, to entice developers to use Baidu’s models. Most recently, the company has released AgentBuilder, which helps developers create AI agents, which by using natural language programming, even those who can’t code can build an AI agent. The long-term goal for ERNIE (and Chat-GPT) is that millions of applications are built on top of these foundational models, ultimately locking AI development into the ecosystem. To do so, these companies will have to create very affordable models to compete with open-source models. Baidu discussed the following efforts: “In addition, our mixture of experts, or MOE approach can partition a user query into distinct tasks, assigning the most suitable models to handle each task and use ERNIE 3.5 or 4.0 only for the most complex tasks. This approach allows for faster responses and lower inferencing cost while maintaining similar performance levels to using more advanced models.”This approach allows for faster responses and lower inferencing cost while maintaining similar performance levels to using more advanced models.”
Being the largest search engine in China, Baidu Search with ERNIE improves search results similar to Chat-GPT/Bing and Gemini/Google. ERNIE will launch across Samsung, Oppo, Vivo and Xiaomi smartphones. APIs are also used across PCs and electric vehicles including Nio. The company has saw “double-digit YoY increase in paying users in the first quarter” for Baidu Wenku, which helps with document creation.
Brand advertisers and SMEs (small to medium sized enterprises) are also adopting ERNIE agents to serve as virtual storefronts, service desks and 24/7 assistance.
China’s GPU Restrictions:
Export controls prevent China from buying Nvidia’s most powerful GPUs. The company has created less powerful GPUs specifically to export to China, such as the A800 and H800. According to SemiAnalysis, Nvidia is also exporting the H20, L20 and L2 GPUs. Naturally, this brings up questions around Baidu’s ability to progress it’s ERNIE family models if the GPUs are less powerful than what’s available globally.
According to Baidu: “For the AI infrastructure layer, a key factor contributing to our high efficiency in model training and inference is our superior capability in GPU cluster management. We have recently made a breakthrough by integrating GPUs from different vendors into one large scale, unified computing cluster, allowing us to use less advanced chips for highly effective model training and inference. Our deep learning framework, PaddlePaddle, has through continuous innovation and enhancement helped reduce the cost of model training and inferencing on a constant basis. PaddlePaddle is compatible with over 50 different chips, many domestically designed, and the developer community has grown to 13 million […]”
“In the long run, I think China will form an ecosystem of its own, probably with less powerful chips but most efficient home grown software stack. There is ample room for innovation in the application layer, model layer, and framework layer […] We believe that the chips we have in hand are sufficient to support earnings for the next one or two years, and because of an ability of high performance chips in China in this year, 2024, we expect our capex to be smaller versus last year.”
Conclusion:
Per our last write-up, the I/O Fund team believes that Baidu could make for an interesting momentum play. We define a momentum position as one where technicals lead, where we respect the stops, and where the fundamentals may not be perfect for one reason or another. For Baidu, risk from China is too high for the stock to be considered for anything more than momentum. Given the emphasis on AI in the markets combined with China pushing for its domestic tech to be the predominant tech used by its citizens, we foresee a scenario where Baidu emerges as a decent choice for those who want to participate in lower valuations.
Recommended Reading: