This funding will primarily be used for the research and development of the next-generation GLM base model and the fully self-training system
By Emeka Uche
·
Published on September 13, 2026
·
2 min read
Zhipu announced today the completion of a financing round of approximately US$5 billion, including a share placement of approximately US$2 billion and a convertible bond issuance of approximately US$3 billion.
This funding will primarily be used for the research and development of the next-generation GLM base model and the fully self-training system, as well as the deployment and upgrading of large-scale training, production inference, computing resources, and related technical infrastructure.
In its announcement, Zhipu described fully self-training as the next-generation GLM training within the environment built by the previous-generation GLM, forming a recursive self-improvement loop.
Specific investments include automated generation and selection of training data, construction of task environments, improvement of long-range inference capabilities, as well as adaptation to domestic chips, operator development, and inference optimization.
Zhipu stated that this R&D plan simultaneously targets two key variables: model capability and computing efficiency.
Full self-training explores how the model participates in building training resources, expanding data and task environments for subsequent iterations; chip adaptation and inference optimization focus on improving effective output under the same hardware conditions.
Investing in these two paths concurrently allows Zhipu to explore more efficient ways to improve models while expanding the scale of training.
