AI Series #42: Connecting The Dots on Recent Earnings and Industry Developments

研报英文原文证据摘录

AI Series #42: Connecting The Dots on Recent Earnings and Industry Developments

2.7Pro 816B745B

triggered by market watching closely on ARR, we view the latest updates by BABA on ARR at Source: OpenRouter, Jefferies

USD1.5bn (RMB10bn) in Jun Q and USD4.5bn (RMB30bn) with inhouse models from Bailian

API as key contributor is an important metrics to be watched out in coming quarters. The

shift from conversation to autonomous execution in coding/agents is a huge addressable

market globally. Token price is expected to be on the uptrend considering value created to

customers amid supply constraints. For Tencent, productivity tool WorkBuddy/CodeBuddy

has fast user growth and high retention rate. The increase in paying user conversion drives

token consumption in Tencent Cloud; (2) ROI. BABA considers AI is like manufacturing sector

with two factories on AI training and AI inference, with both based on data centre. BABA

monetizes through different models such as IaaS other than MaaS. Chips are fully utilized. For

Tencent, products are measured on full lifecycle basis which are not the same for each one.

For example (a) chips on ad tech is a relatively short lifecycle investment; (b) model training is

a long term investment; (c) cloud has clear margin profile and ROI. Both companies support

the investment from solid cash flow of core business; (3) Compute. Tencent is expected to

step up capex spending on domestic GPU or ASIC due to increase supply of fabs in domestic

and neighbouring countries. On the other hand, it has been a big buyer for CPU and networking

chips for many years. This is positive to cloud revenue in 2H26. For BABA, it expects capex

over time to exceed 3-years guidance of RMB380bn. There is a number of ways to set up data