Baidu shares fall 9.8% after launch of Ernie 5.0 AI model

Baidu shares dropped by up to 9.8% in Hong Kong after the company unveiled Ernie 5.0, the latest version of its AI model, at a conference on November 14.

The drop marks Baidu’s largest single-day fall in seven months, following a year-to-date rise of over 40%.

Investors were disappointed with the new model, and Robert Lea, an analyst at Bloomberg Intelligence, said it is not sufficiently differentiated from competitors.

Baidu has faced increasing competition from companies such as Alibaba and DeepSeek in China’s AI sector.

Alibaba shares also slipped as much as 3.5%, as the company is said to be preparing an overhaul of its main mobile AI app to more closely resemble OpenAI’s ChatGPT.

The recent pullbacks come amid broader concerns over AI stock valuations and rising investment costs.

🔗 Source: Bloomberg

🧠 Food for thought

Implications, context, and why it matters.

Investors question Ernie 5.0’s edge and missing outside tests

Baidu shares fell 9.8% after unveiling Ernie 5.0 as third-party head-to-head benchmarks for ERNIE X1 (Baidu’s new reasoning model within the Ernie 5.0 lineup) versus DeepSeek R1 (a competing Chinese reasoning model) remain missing, so performance claims stay unverified 1.
Third-party tests place Alibaba’s Qwen3-235B-A22B (an AI model in Alibaba’s Qwen family) cost-efficient on coding, math, and writing 2. Baidu says ERNIE X1 targets DeepSeek R1 level reasoning at about half price on standard rates, though DeepSeek’s off-peak discounts can flip that edge 1.

Infrastructure suppliers will benefit from China’s $98 billion AI buildout in 2025

China’s AI capital expenditures (capex) will reach $98 billion in 2025, with up to $55 billion from government while major internet companies add $24 billion, which opens room for data center and energy providers 3.
Chinese AI hyperscalers (large cloud and internet platforms that operate massive data centers) use home-grown chips with custom hardware 4. Cloud providers plan a 65% capex jump, so more spend goes to data centers and energy over semiconductors 3.
Suppliers of liquid cooling systems (coolant-based systems that remove heat from densely packed chips and servers) plus power transmission solutions (equipment that moves electricity from the grid to data centers) see demand as operators move to higher-density workloads (more chips and servers per rack) 3 5. Experts say China’s grid keeps 80–100% reserve margins (the capacity above expected peak demand), giving headroom for data centers 6.