Tencent has confirmed it is passing up a fast, guaranteed profit from its massive AI infrastructure spending in favor of a longer term strategy built around its own models and applications. The disclosure came during the company’s Q2 earnings call, when a Bernstein analyst asked when Tencent expects a return on the $53 billion in capital expenditure it made during the quarter.
Chief Strategy Officer James Mitchell said demand for compute is strong enough that Tencent could recover its depreciation costs almost immediately if it simply rented out its infrastructure, following the neocloud model used by other AI infrastructure operators. Company president Martin Lau went further, saying Tencent has offers for its compute capacity at more than 30 percent profit compared to what it paid just a few months ago, and that behaving like a neocloud would deliver a decent return in an immediate timeframe.
Instead, Lau said Tencent is allocating a substantial share of its new compute toward building its own models to state of the art status, and toward pushing its own AI applications to market leadership in China. The bet is that superior intelligence, delivered through top tier models and market leading applications, will convert into stronger economic returns over the long term than short term infrastructure rental would provide.
Tokens, Models, and Products
Those longer term returns are expected to come from selling tokens for services such as WorkBuddy, described as an agent swarm capable of planning, executing, and running tasks in parallel to hand back complete deliverables. Tencent also offers CodeBuddy, a code generation tool that Mitchell said is accelerating cloud migration projects and generating more business for Tencent’s cloud unit.
On the model side, Tencent released its 295 billion parameter open weight Hunyuan-3 model in July. Lau called it a very small model and said the upcoming Hunyuan-4 will be larger and more capable than bigger models from rivals. He said Tencent is designing its products specifically around Hunyuan-4, with the expectation that mutual optimization between model and product will make both more powerful than relying on third party models. A fifth version of Hunyuan is also planned, with Lau saying Tencent will eventually deliver a genuinely state of the art model.
Strong Quarter, Mixed Investor Reaction
The strategy comes as Tencent posts solid financial results. Q2 revenue grew 11 percent to $30.3 billion, and net profit rose nine percent to $10.3 billion. Weixin and WeChat combined added seven million average monthly active users to reach 1.349 billion. Advertising revenue climbed 22 percent, and the gaming business grew 17 percent in China.
Despite the numbers, investors appear uncertain about the strategy. Tencent’s share price has trended downward since before the earnings call and dipped roughly three percent following the announcement, suggesting some skepticism about deferring guaranteed infrastructure profits in favor of a long term AI product bet.