The AI revolution is upon us, and Baidu, the Chinese tech giant, is standing front and center. Analysts are abuzz about Baidu's ERNIE family of large language models (LLMs), its potential to revolutionize search and advertising, and the breakneck race towards robotaxi profitability. But hidden beneath these headlines, a quieter, yet potentially more impactful story is unfolding: the unexpected surge of Baidu's CPU cloud business.

While the spotlight shines on the dazzling world of GPUs powering AI model training, a crucial fact has been overlooked. Baidu is leveraging the very success of its GPU cloud to drive a parallel growth engine: CPU cloud services. This isn't merely a footnote in Baidu's narrative. It represents a savvy business strategy, exploiting a symbiotic relationship between the two sides of its cloud offering.

Dou Shen, EVP in charge of Baidu AI Cloud Group, revealed this subtle shift during the Q1 2024 earnings call. He explained that as customers flock to Baidu for its superior GPU cloud capabilities for model training, they're increasingly entrusting Baidu with their CPU cloud needs as well. This cross-selling phenomenon is generating a virtuous cycle, solidifying Baidu's position not just as an AI powerhouse, but as a comprehensive cloud solutions provider.

Why does this matter? The CPU cloud market, while less glamorous than its GPU counterpart, is a massive and steadily growing sector. It forms the backbone of traditional cloud services, hosting websites, running applications, and managing databases. Baidu, by seamlessly integrating CPU cloud services into its AI-centric offerings, is tapping into this existing market with a potent value proposition.

The numbers tell a compelling story. Baidu's AI Cloud revenue, powered by both GPU and CPU services, is accelerating. It surged by 12% year-over-year in Q1 2024, building upon a strong 11% increase in Q4 2023. This acceleration is driven not only by the direct revenue from gen-AI and foundation models (6.9% of total AI Cloud revenue in Q1 2024) but also by the expansion of its traditional CPU cloud business.

Hypothesis: The Rise of CPU Cloud Driven by ERNIE Inferencing

Here's a hypothesis worth considering. As Baidu refines its ERNIE family of models, making them increasingly affordable and efficient, the demand for model inferencing will skyrocket. This shift towards inferencing will fuel the need for CPU resources, creating a massive tailwind for Baidu's CPU cloud business.

The company's strategic focus on cost optimization and efficiency further strengthens this hypothesis. Baidu has dramatically reduced ERNIE's inference cost to only 1% of its March 2023 level. This relentless pursuit of efficiency makes ERNIE-powered applications incredibly attractive to enterprises, driving adoption and, in turn, boosting demand for CPU cloud resources.

Baidu's success in attracting new customers to its AI Cloud, many of whom are tech companies looking to build their own AI models, further validates this trend. As these customers build their AI infrastructure with Baidu, they're more likely to utilize Baidu's entire suite of cloud services, including CPU offerings.

Visualizing Baidu's AI Cloud Growth

The following chart illustrates the growth trend in Baidu's AI Cloud revenue, driven by both GPU and CPU services.

The CPU Cloud: An Understated Powerhouse

Investors should pay close attention to this understated yet powerful driver of Baidu's cloud growth. The CPU cloud, often overshadowed by the AI frenzy, represents a significant and sustainable revenue stream for Baidu. It's a testament to the company's strategic foresight and its ability to capitalize on emerging market trends. In the grand narrative of the AI revolution, Baidu's CPU cloud business may not grab the headlines, but it's a crucial piece of the puzzle, quietly shaping the company's long-term success.

Baidu Q1 2024 Earnings Call Highlights

"Fun Fact: Baidu, founded in 2000, derives its name from a poem written over 800 years ago. The poem uses the phrase "hundreds of times" to describe the persistent search for one's ideal. This reflects Baidu's mission to provide users with the information they need, no matter how complex their search."