Analysis
Baidu reported second-quarter revenue of 31.33 billion yuan ($4.62 billion), down 4% year-over-year and short of analyst expectations, as a steep decline in online advertising outpaced growth in its AI cloud business, Reuters reported via Yahoo Finance. Online marketing revenue -- historically Baidu's core business -- fell 19% year-over-year to 13.1 billion yuan, which the company and analysts attributed to China's prolonged real-estate slump and soft consumer demand pushing advertisers to pull back spend.
The one clear growth line was AI. Baidu's Core AI-powered Business segment, spanning cloud infrastructure, AI applications and AI-native marketing services, grew 25% year-over-year to 12.5 billion yuan. Shares fell as much as 14% on the report, per MarketScreener, and net profit plunged 68% year-over-year as Baidu continues heavy AI infrastructure investment against a shrinking legacy revenue base.
The arithmetic problem every legacy tech company doing an AI pivot faces
Baidu's quarter is a clean illustration of a math problem showing up across incumbent tech companies attempting an AI transition: the AI segment is growing fast in percentage terms (25%) but is still small in absolute terms (¥12.5B) relative to the legacy business it needs to offset (¥13.1B in advertising alone, down ¥3.1B year-over-year). A 25%-growing ¥12.5B segment adds roughly ¥2.5B in new revenue; a 19%-declining ¥13.1B+ segment loses more than that. The AI business has to compound at its current growth rate for several more years before it's large enough to outrun the legacy decline in absolute dollar terms, not just percentage terms -- and that's before accounting for the capital being spent to build it.
CEO Robin Li's public commitment that Baidu's Ernie model line will "return to first tier" is a direct acknowledgment that the company's foundation models have fallen behind the pace set domestically by DeepSeek and Alibaba's Qwen, and globally by OpenAI, Anthropic and Google. Baidu was an early and aggressive mover in Chinese AI -- Ernie launched well before most domestic rivals -- but a company admitting its flagship model needs to "return" to a competitive tier is conceding it lost ground it once held.
Not an isolated case
Alibaba and Kuaishou are reportedly addressing similarly mounting AI infrastructure costs as competition intensifies across Chinese tech, per the same wire coverage -- this is a sector-wide capital-intensity problem, not one specific to Baidu's execution. The read for US investors and founders watching the China AI market: capital is being committed at scale by multiple large Chinese platforms simultaneously, funded increasingly by cannibalizing legacy ad and e-commerce revenue rather than by fresh growth capital, which is a structurally different and more fragile funding model than the venture-backed buildout happening in the US.
