Zhipu achieved a revenue of 954 million yuan in the first half of the year, marking a significant year-on-year increase of 399.7%, though it fell slightly short of market expectations. The company’s revenue structure has undergone notable transformations, with revenue from cloud-based model APIs now constituting over 85% of the total. Conversely, revenue from localized deployments has decreased, resulting in an overall decline in gross margin to 26.4%. Research and development expenditures surged to 2.131 billion yuan, reflecting a 33.6% year-on-year increase.

By the end of August, Zhipu’s monthly annualized recurring revenue (ARR) had soared to $1.6 billion, with weekly annualized revenue surpassing $2 billion. Monthly revenue was nearing the total achieved in the first half of the year, a testament to substantial user growth. The market remains cautious about Zhipu’s ability to sustain its state-of-the-art (SOTA) model advantage. At this juncture, Zhipu is focusing on post-training techniques, asserting that they yield higher returns. Simultaneously, the company is advancing the training of its next-generation foundational model, transitioning from coding to collaborative working stages, and exploring areas such as cybersecurity.

However, aside from this domain, Zhipu has yet to generate significant scaling revenue. Consequently, its revenue performance in the latter half of the year will still heavily depend on leveraging its model advantages.