[Zhipu discloses the computing power scaling model: 30 billion yuan investment can build 100,000 P, with annual inference revenue of up to 40 billion yuan] September 16th, Zhipu said in a telephone exchange conference on September 16 that large model companies need to be firm in scaling. An investment of 30 billion yuan in computing power can build approximately 100,000 P, with 40% reserved for training and R&D and 60% for business reasoning. Thanks to GLM-5.3's high inference cost-effectiveness, theoretical inference gross profit can reach up to 80%. 60% of the inference computing power can earn 40 billion yuan a year, covering computing power investment and model training R&D expenses. The actual situation is affected by computing power utilization, business discounts, and fluctuations in computing power prices. (Kechuangban Daily reporter Li Mingming)