Knowledge Atlas Technology (2513.HK): Initiate with Neutral at US$110bn valuation; China’s top enterprise/coding AI model

Research evidence excerpt

Knowledge Atlas Technology (2513.HK): Initiate with Neutral at US$110bn valuation; China’s top enterprise/coding AI model

Goldman Sachs Knowledge Atlas Technology (2513.HK)

Key upside/downside risks

Upside risks:

n Stronger-than-expected model intelligence uplift from self-sustainable

positive flywheel: Breakthroughs in coding and multi-modal performances, offering

seamless integration and understanding across text, image, and audio, would unlock

new applications, physical AI and AGI potentials.

n Faster-than-expected path to profit visibility: If Zhipu shifts faster toward being

open-weight with commercial terms (from pure open source) and demonstrate

superior operational efficiency, such as optimizing model inference costs and

training infrastructure, or diversify monetization channels to achieve higher revenue

growth or better unit economics sooner, it would accelerate the timeline to positive

cash flow and net income.

n Additional non-inference take rate revenue streams: If Zhipu is able to train and

achieve SOTA model performance, an open-weight approach charging take

rates/revenue sharing on neoclouds/hyperscalers that house such models and sell to

customers may transform its business models to a high gross margin business

(without needing to conduct inference for each token sold, but earning via revenue

sharing with platforms that sell top performing Chinese models).

n Stronger-than-expected commercialization capability: Growth of new

applications, Code and Co-Worker platforms, coupled with ‘digital worker’ agent

models could provide multiple channels for revenue generation. If these channels

achieve higher-than-expected adoption rates, conversion efficiencies, or ARPU,

Zhipu could outperform current revenue forecasts.

Downside risks: