Tencent is prioritizing the use of its newly acquired AI servers for its own models and services. During an earnings call held on the evening of the 12nd, following the release of its Q2 2026 financial results, management faced repeated questions regarding their capital allocation priorities and the expected timeline for returns.

Capital expenditure on facilities and equipment in the second quarter reached CN¥52.784 billion (approx. ₩11.08 trillion), 2.8 times higher than a year ago. When factoring in actual cash outflows of CN¥59.3 billion (approx. ₩12.45 trillion), along with content costs and rental fees, total spending exceeded the CN¥52.7 billion (approx. ₩11.06 trillion) generated from operations, resulting in a negative free cash flow of CN¥13.8 billion (approx. ₩2.90 trillion).

Management explained that the cost of the equipment could be recouped immediately simply by leasing the servers to other companies. They added that if they were to resell the inventory they had pre-paid for months ago, they could realize a profit of over 30%. However, they emphasized that new incoming capacity is being prioritized for training their own AI models and scaling their internal services, with leasing to third parties taking a secondary role.

Losses from newly launched AI products increased from approximately CN¥8.8 billion (approx. ₩1.84 trillion) in the first quarter to about CN¥10.5 billion (approx. ₩2.19 trillion) in the second. The focus of investment shifted significantly between the two quarters; management explained that they prioritized investment in the AI productivity tool 'WorkBuddy' due to a notable increase in users, while delaying other AI products. They noted that there is a cap on total investment and did not provide specific figures on when they expect to see a return.

Management described this large-scale investment as a temporary expenditure concentrated in this year and the next. They stated that the costs associated with model development are one-time in nature, meaning they will not need to invest at the same scale every year. They also stated that they would only increase the number of servers used for running services as they confirm returns on those investments.

To fund these investments, the company is drawing on cash reserves, existing assets, and prudent borrowing, in addition to cash generated from operations. Decisions on whether to use funds for share buybacks or further investment will be based on performance and returns. They also announced that they would separate and disclose expenditures for new AI-driven businesses from their existing operations.