China Internet & New Media: Takeaways from an LLM expert call

Research evidence excerpt

China Internet & New Media: Takeaways from an LLM expert call

Global Markets Research

13 July 2026China Internet & New Media

EQUITY: INTERNET & NEW MEDIA

Takeaways from an LLM expert call Research Analysts

China Internet & New Media

Open source is not an open playbook JialongJialong.shi@nomura.comShi - NIHK

+852 2252 1409

Key highlights Rachel Guo - NIHK

The Nomura China Internet team held a group call with an expert from a Chinese AI lab rachel.guo@nomura.com

affiliated with a research institution. The lab currently has its proprietary foundation +852 2252 1400

models deployed across more than 100 enterprise customers.

The expert noted that persistent supply tightness and price inflation for high-end Nvidia

(NVDA US, Not rated) chips are making domestic AI accelerators increasingly attractive.

While Chinese model providers are broadly reducing token prices, the expert highlighted a

clear bifurcation between basic and advanced models. Providers are aggressively cutting

prices for basic models – which process commoditized tasks – to acquire customers,

while maintaining premium pricing for their advanced models. The expert thinks that

model providers can enhance enterprise customer loyalty by embedding their models

deeply into corporate workflows. This integration makes switching models highly time-

consuming and financially burdensome, effectively deterring customer churn.

The expert believes DeepSeek (unlisted) possesses a distinct cost advantage driven by

superior operational efficiency during inference workloads. This efficiency is highly difficult

for other platforms hosting the same open-source DeepSeek model to replicate. The

expert views financial services, office productivity, coding, manufacturing, legal, and