Baidu reported its latest quarterly results with the same cash-flow profile as Tencent. Q2 revenue was RMB 31.3 billion, down 4 per cent year on year and 2 per cent quarter on quarter. Attributable net profit was RMB 2.3 billion and non-GAAP net profit RMB 2.6 billion, down 68 per cent and 46 per cent. US shares fell nearly 13 per cent overnight and Hong Kong shares dropped more than 13 per cent intraday, both one-year lows.
By segment, general business revenue was RMB 25.2 billion, down 3 per cent quarter on quarter and 4 per cent year on year. AI-driven business was the growth engine at RMB 12.5 billion, up 25 per cent year on year but down 8 per cent quarter on quarter, about 50 per cent of general revenue, two points lower than the prior quarter. Smart-cloud infrastructure kept growing, with revenue of RMB 7.3 billion and year-on-year growth accelerating to 50 per cent, while GPU cloud revenue rose 283 per cent, its fourth straight triple-digit quarter.
Baidu’s Robotaxi business is globalising, with public-road tests in Hong Kong, Switzerland, Dubai and London, covering 28 cities with over 350 million kilometres of autonomous driving, including more than 240 million fully driverless kilometres, though the gap from testing to routine commercial operation remains.
Like Tencent, Baidu’s free cash flow turned negative, with net outflow of RMB 7.954 billion as capex hit RMB 11.39 billion, up roughly 200 per cent year on year and 93 per cent quarter on quarter, directed at AI compute. Total cash and investments reached RMB 283.1 billion, up over 80 per cent, a buffer for the spending ahead. Baidu’s leadership said it remains in the AI investment cycle with unwavering resolve but equal focus on prudent spend and return on investment.
Kunlunxin, Baidu’s AI chip unit, is sprinting toward a Hong Kong IPO and, buoyed by AI demand, adapted in Q2 to new models including Kimi K3, GLM 5.2, MiniMax M3 and Hunyuan Hy3. Li Yanhong said Kunlunxin has shipped three generations of AI chips and will push the latest M100 for large-scale inference and a future M300.
On foundation models, Baidu has slipped among Chinese peers. Li said the market is still in fast flux and long-term competitiveness comes from sustained investment, application-driven thinking and patience, and expressed confidence in returning Ernie to the front rank of foundation models. This year Baidu reshuffled its org, created a model committee in May and brought in 1997-born Sun Tianxiang to lead base-model development, reporting directly to Li.
Editor’s note: This is an adapted translation of the original Sohu report. It has been trimmed and restructured for readability for an international business audience.