According to MoneyDJ, Tencent executives stated yesterday (November 18) that the company expects AI infrastructure investment costs and operational pressure to ease by 2028 as efficiency improves. The infrastructure built in 2026 and 2027 will continue to support model training and token generation in subsequent years, executives said.
The company anticipates hardware procurement costs will decline significantly by 2028 compared to current high prices, while computing efficiency for model training and inference continues to improve rapidly. This combination of infrastructure reuse, lower hardware costs, and higher computational efficiency is expected to substantially improve the return on investment for AI operations by 2028.