This reveals Tencent's multi-faceted approach to US chip restrictions, showing they've strategically stockpiled GPUs ahead of restrictions while simultaneously developing software optimizations, model distillation techniques, and alternative chip sourcing to address their AI computing needs.

“On the GPU front, it's actually a very dynamic situation right now... since the last earnings call we have seen an H20 ban and then after that there was the BIS new guidelines that just came in overnight... The good thing that we are in is that number one, I think we have a pretty strong pile of chips that we acquired previously and that would be very useful for us in executing our AI strategy... Secondly, in terms of the training of our large language models... we start to move off the concept or the belief of the American tech companies which they call the scaling law which required continuous expansion of the training cluster. Now we can see even with a smaller cluster you can actually achieve very good training results and there's a lot of potential that we can get on the post-training side.”

“There are a lot of ways through which we can fulfill the expanding and growing inference needs and we just need to keep exploring these venues and spend probably more time on the software side rather than just brute force buying GPUs.”

“If you can improve inference efficiency 2x, that basically means the amount of GPUs get doubled in terms of capacity. So that's actually a very good way of investing our resources to improve on the inference efficiency.”

“We can actually potentially make use of other chips, compliant chips available in China or available for us to be imported, as well as ASICs and GPUs in some cases for smaller models inferences.”

— Martin Lau, President

Tencent has clearly ranked AI monetization paths with advertising enhancement as the top priority, followed by transaction facilitation, with GPU rental and subscriptions considered lower priority or unsuitable for the Chinese market.

“If you look at advertising, it's directly augmented by AI because AI can actually help to improve the targeting capability of our ads... and when we deliver better results then it translates directly into additional advertising revenue and I think that is a big opportunity that we are already realizing in our performance ads but there's more opportunity to develop over time.”

“Transaction is actually very closely tied to advertising right... when you have advertising that leads to direct transactions and then advertising value actually goes up significantly and I think that's the way we are actually also trying to increase our advertising revenue.”

“GPU rental is sort of directly related to cloud business and that's more like a reselling business mostly and to a large extent right now we are putting it on a lower priority because especially when there's a short supply of GPUs.”

“Subscriptions I think is not the most likely business model for AIs in China right now. Everybody is actually providing AIs for free, so the subscription model which exists outside of China I think is not going to be mainstream.”

— Martin Lau, President

Tencent reports improving consumer credit quality despite macroeconomic uncertainties, attributing this partly to the substantial savings Chinese consumers have accumulated in recent years.

“What we have observed on the ground this year is that the credit quality for the loans that we facilitate through our platform has been gradually but consistently improving, and we think that's partly due to better selection of borrowers and so forth but it's partly also due to a more macro development which is that over the last several years, consumers in China in aggregate have built up unusually large savings balances.”

“All else equal one would rather be lending into a consumer with unusually large savings and a consumer with unusually limited savings.”

— James Mitchell, CSO