1 Tencent’s earnings report put direct downward pressure on the Hang Seng Index. No wonder the U.S. listed Chinese concept stocks were also disappointing a few days ago—looks like there was a leak of information in advance.
Overall, the biggest highlight is that the gaming industry remains stable. The biggest downside is that free cash flow turned negative, which is almost the first time in all these years.
Due to the earnings impact, the stock fell by nearly 4 points.
Tencent Q2 revenue was RMB 204.8 billion (+11%): it looks steady, but GAAP net profit of RMB 56.0 billion was basically flat—most of the money went into AI.
Three key figures: Capital expenditures were RMB 52.8 billion, up 176% year over year (Q1: RMB 31.9 billion, showing an accelerating trend). Excluding AI products, operating profit was RMB 86.1 billion, up 19% year over year (indicating the core business hasn’t weakened; it was dragged down by AI). AI products consumed RMB 10.5 billion in profit this quarter (Q1: RMB 8.8 billion, still accelerating).
Brightest segment: Marketing services +22% (AI ad recommendation model fully rolled out). Domestic games +17% (Delta Action and Fearless Contracts hit new all-time highs in daily active users). Cloud business growth accelerated to ~21%.
Management’s stance: They clearly said AI capital expenditures are “concentrated investments in the next two years,” not linear growth. Even though they invested RMB 52.8 billion, during the call they also said “compute is capacity-constrained, so we need to continue buying.” WorkBuddy monthly visits were 20.97 million—already 1.6 times the number for ByteDance’s TRAE.
2 In the AI era, everyone is most concerned about how agents perform. The big question is whether WorkBuddy can keep growing steadily, and how its subsequent monthly active users compare with ByteDance’s AI products. But besides Doubao, ByteDance also has two very different products—Coze and TRAE—that target and compete with WB in different ways. Combined with the current dominance of its SD video model, ByteDance still has a slight edge.
ByteDance’s biggest advantage is that it captures people’s time. Tencent’s biggest advantage is that it firmly uses games to retain the middle-tier user segment with the strongest willingness to pay.
PS: I think Tencent could consider investing more in minimax. If that happens, the “SD monopoly” issue would be halfway resolved.
PS2: Personal view—there’s basically no major issue with building a position under 500. Now you can sell some near 420 the ATM monthly put; the returns should be quite good.
3 Before WorkBuddy took off, I was fairly bearish on Tencent. But after using it deeply, I found that because large models in agents sit in a position that can be replaced at any time, their internal competition pressure is far higher than front-end or user-facing c-end products. This is truly AI consumption. Ironically, the large model serves more professional domains like coding—what it sells is more like production tools rather than consumer-life services.
4 For the U.S. stock market: many sideways-trading sectors are not recommended for new involvement. Sectors with no volatility are a dead pond. Retail investors can’t hold on; they’ll enter a slow-distribution bear pattern, which is extremely dangerous for stocks.
PS: Still optimistic about copper, aluminum, and uranium. Friends with asset allocation opportunities in the A-share market can consider taking a quasi-exposure to these three sectors.
5 CRS has started charging taxes for offshore trusts. Every amount of money won’t be rootless water—who charges also protects and who is responsible. As long as there’s no double taxation, a 20% collection at the capital gains level across the global range is generally reasonable.
6 Known footballer Luka Dončić was extorted by his fiancée for a $50 million breakup fee. The whole world blew up—lawsuits in the U.S. can’t be won, so he went to his home country, Slovenia. But given the fiancée’s level of recognition and reputation in the home country (and that they never officially registered the marriage), I estimate she might receive at most $10 million...
7 Another related news item: the Lakers. They sold the Lakers for $12 billion, previously bought for $10 billion from the Buss family. It’s said the owner sold in a hurry because of a financial investigation, but he still made $2 billion in a year—and likely through a leveraged acquisition.
The new buyer is a relative of the Trump–Ivanka family. It’s a story of political dynasty people turning into a sports powerhouse—an approach that can please the public and win votes.
8 Recently, Taiwan’s chip stocks have seen a certain rebound. With Korea’s rebound as well, could we get another delivery cycle for selling?
9 Samsung is going to make a 1nm lithography machine. Be alert to the strategy of continuing equity financing at high levels... Currently, storage capacity is being expanded across a wide area. I don’t think highly trapped friends should wait; they can quickly break even and get out.
10 Recently more and more people around me ask how long it will take for the big A to recover their break-even. But none of them have cut yet. I think it’s probably not there yet... If this range-bound phase continues for about half a year, prices could be entered. Otherwise, keep waiting—until margin trading and financing falls below 1.5 trillion, then we’ll look again.
Overall, the biggest highlight is that the gaming industry remains stable. The biggest downside is that free cash flow turned negative, which is almost the first time in all these years.
Due to the earnings impact, the stock fell by nearly 4 points.
Tencent Q2 revenue was RMB 204.8 billion (+11%): it looks steady, but GAAP net profit of RMB 56.0 billion was basically flat—most of the money went into AI.
Three key figures: Capital expenditures were RMB 52.8 billion, up 176% year over year (Q1: RMB 31.9 billion, showing an accelerating trend). Excluding AI products, operating profit was RMB 86.1 billion, up 19% year over year (indicating the core business hasn’t weakened; it was dragged down by AI). AI products consumed RMB 10.5 billion in profit this quarter (Q1: RMB 8.8 billion, still accelerating).
Brightest segment: Marketing services +22% (AI ad recommendation model fully rolled out). Domestic games +17% (Delta Action and Fearless Contracts hit new all-time highs in daily active users). Cloud business growth accelerated to ~21%.
Management’s stance: They clearly said AI capital expenditures are “concentrated investments in the next two years,” not linear growth. Even though they invested RMB 52.8 billion, during the call they also said “compute is capacity-constrained, so we need to continue buying.” WorkBuddy monthly visits were 20.97 million—already 1.6 times the number for ByteDance’s TRAE.
2 In the AI era, everyone is most concerned about how agents perform. The big question is whether WorkBuddy can keep growing steadily, and how its subsequent monthly active users compare with ByteDance’s AI products. But besides Doubao, ByteDance also has two very different products—Coze and TRAE—that target and compete with WB in different ways. Combined with the current dominance of its SD video model, ByteDance still has a slight edge.
ByteDance’s biggest advantage is that it captures people’s time. Tencent’s biggest advantage is that it firmly uses games to retain the middle-tier user segment with the strongest willingness to pay.
PS: I think Tencent could consider investing more in minimax. If that happens, the “SD monopoly” issue would be halfway resolved.
PS2: Personal view—there’s basically no major issue with building a position under 500. Now you can sell some near 420 the ATM monthly put; the returns should be quite good.
3 Before WorkBuddy took off, I was fairly bearish on Tencent. But after using it deeply, I found that because large models in agents sit in a position that can be replaced at any time, their internal competition pressure is far higher than front-end or user-facing c-end products. This is truly AI consumption. Ironically, the large model serves more professional domains like coding—what it sells is more like production tools rather than consumer-life services.
4 For the U.S. stock market: many sideways-trading sectors are not recommended for new involvement. Sectors with no volatility are a dead pond. Retail investors can’t hold on; they’ll enter a slow-distribution bear pattern, which is extremely dangerous for stocks.
PS: Still optimistic about copper, aluminum, and uranium. Friends with asset allocation opportunities in the A-share market can consider taking a quasi-exposure to these three sectors.
5 CRS has started charging taxes for offshore trusts. Every amount of money won’t be rootless water—who charges also protects and who is responsible. As long as there’s no double taxation, a 20% collection at the capital gains level across the global range is generally reasonable.
6 Known footballer Luka Dončić was extorted by his fiancée for a $50 million breakup fee. The whole world blew up—lawsuits in the U.S. can’t be won, so he went to his home country, Slovenia. But given the fiancée’s level of recognition and reputation in the home country (and that they never officially registered the marriage), I estimate she might receive at most $10 million...
7 Another related news item: the Lakers. They sold the Lakers for $12 billion, previously bought for $10 billion from the Buss family. It’s said the owner sold in a hurry because of a financial investigation, but he still made $2 billion in a year—and likely through a leveraged acquisition.
The new buyer is a relative of the Trump–Ivanka family. It’s a story of political dynasty people turning into a sports powerhouse—an approach that can please the public and win votes.
8 Recently, Taiwan’s chip stocks have seen a certain rebound. With Korea’s rebound as well, could we get another delivery cycle for selling?
9 Samsung is going to make a 1nm lithography machine. Be alert to the strategy of continuing equity financing at high levels... Currently, storage capacity is being expanded across a wide area. I don’t think highly trapped friends should wait; they can quickly break even and get out.
10 Recently more and more people around me ask how long it will take for the big A to recover their break-even. But none of them have cut yet. I think it’s probably not there yet... If this range-bound phase continues for about half a year, prices could be entered. Otherwise, keep waiting—until margin trading and financing falls below 1.5 trillion, then we’ll look again.