Open in interactive viewer → charts, metric popovers & call review

Sohu's Real Competitor Isn't Tencent. It's AI Short Drama.

Q3 advertising growth was project timing dressed as a recovery, Changyou's CEO spent the Q&A conceding a demand shock he can't measure, and the $150M buyback holding up a $360M market cap is down to its last $16M with no successor named.
SOHU · Earnings Call · 2026-11-09

A quarter that looked like a turn, and a management team that spent the call explaining why it wasn't

Sohu reported $133M of third-quarter revenue, down 26% year-over-year, a GAAP net loss of $15M against a $9M profit a year ago, and a Q4 guide of $17M to $27M of net loss. On the surface that is a company in managed decline. But the interesting tension is elsewhere: Sohu's marketing services line actually grew — $15M, up 10% year-over-year — and management spent the Q&A carefully dismantling the idea that it means anything.

The growth was a calendar accident, and the CEO said so

Three months ago, CFO Alicia Yap asked Charles Zhang whether the Q3 guide marked “the first time for a long time that you are now expecting a positive growth on the year-over-year” — Charles Zhang, Chairman and Chief Executive Officer · 2026-08-10, and whether it was “the first sign of potentially the more positive year-over-year trend going forward” — Charles Zhang, Chairman and Chief Executive Officer · 2026-08-10. Charles was cautious then; he was blunter now. The Q4 guide of a 6% to 13% sequential decline and an 18% to 24% annual decline got a one-line explanation: “It's just that some of these projects just happened to concentrate in Q3 instead of Q4.” — Charles Zhang, Chairman and Chief Executive Officer · 2026-11-09 That is the honest read, and it is also the reason not to extrapolate. Sohu no longer sells standard brand advertising; it sells bespoke event-and-KOL packages, which is why Charles noted “we changed the name for advertising to so-called marketing services revenue” — Charles Zhang, Chairman and Chief Executive Officer · 2026-11-09. Project accounting cuts both ways — big offline conventions and influencer festivals can be pulled forward across a quarter boundary, and Sohu's base is small enough that a couple of deals swing the whole line. Add a soft Double 11 events cycle on the e-commerce side and the Q3 beat evaporates. The vertical mix behind the number is worth flagging, because it moved violently. Auto was 19% of ad revenue in Q1, 33% in Q2, and now “the auto industry now in Q3 having 42% of the total, and FMCG decreased a little bit to 17%, and IT 12%” — Charles Zhang, Chairman and Chief Executive Officer · 2026-11-09. A doubling of auto concentration in two quarters is not diversification; it is a single, price-war-battered, export-focused industry deciding whether to keep buying Sohu's advertising revenues. Meanwhile, the Sohu social media platform still ran a $74M quarterly operating loss against roughly $19M of revenue, marginally worse than the $71M lost a year ago. That business loses about five dollars for every one it earns, and nothing on this call changed that arithmetic.

Changyou meets a competitor that isn't a game

The most striking shift in this call is lexical. In February, the AI question to Sohu was about tooling, and the answer — “pushing the application of AI, especially in terms of the creation of game design plans, is essential for the production of the gaming industry” — Yaobin Wang, Chief Financial Officer · 2026-02-09 — was the standard response. In November, the first question about AI was about something else entirely: AI drama as a competitor for user attention. That keyword did not exist in Sohu's vocabulary a quarter ago; now it is the second-strongest theme in the transcript. Changyou CEO Dewen Chen gave the most substantive answer of the day, and it was a concession: user research shows competition for entertainment time, and Sohu's response has been to make its games lighter, reducing the time burden on players. That is a product strategy dictated by an outside substitute, not by player preference. His structural defense is fair — “AI dramas offer the users passive and single-person experiences. While gaming offers players interactive and multiplayer experiences” — Dewen Chen, Chief Executive Officer · 2026-11-09 — but the first analyst follow-up, and the second, and the third all went back to the same place, which tells you the sell side does not find the coexistence argument fully convincing yet.

As for whether it impacts the casual or the hardcore games, well, to be honest, we are not sure. But intuitively, I think the AI dramas impact them both, and it will be a long-term impact... the average daily time that users for AI drama spend is almost 2 hours a day.

Joanna Lv, Chief Financial Officer · 2026-11-09
The bull case embedded there is that two hours a day is close to a ceiling, so the incremental damage is bounded. The bear case is that management is quantifying the ceiling but not the substitution, and the casual users cohort — the low-ARPU, high-churn tail that aging MMORPGs quietly depend on — is exactly the segment most likely to drift toward short-form video. Sohu's game business meanwhile posted its sharpest year-over-year decline in this dataset: $114M, down 30%, with segment operating profit falling to $59M from $88M. Q4 is guided to $111M to $121M, or minus 8% to plus 1% — i.e., management is calling the bottom somewhere in the next ninety days, with no new title to justify it. The card RPG promised for 2026 or early 2027 was not re-timed on this call, which is itself a data point. What also dropped out of the conversation matters. Two quarters ago game pipeline and game pipeline chatter, expansion-pack cadence, and TLBB server launches filled the Q&A; this quarter the TLBB roster was reduced to a list of holiday-themed events, and the CEO's forward-looking paragraph leaned on competitive edge in MMORPGs and "global appeal" without a single dated product milestone. The World Cup — the single loudest theme in the prior quarter, which management had spent a question insisting would be immaterial — vanished entirely, having indeed been immaterial.

The buyback cliff nobody priced

Sohu has repurchased 10.1 million ADS for about $134M through November 5, against a market capitalization of roughly $360M. That is a third of the company bought back by the company — the single most important support for the stock, and far more consequential than any advertising quarter. It is nearly exhausted.

We'll continue to finish that. And until when we finish soon, we'll think more. Now we still don't have a plan to have a new program to buy back yet.

Charles Zhang, Chairman and Chief Executive Officer · 2026-11-09
Recall the history. In August 2025, Charles argued against other capital returns because “we need a war chest to do the battle” — Chaoyang Zhang, Chairman and Chief Executive Officer · 2025-08-04 for the social network. In August 2026, the program was converted to buyback program open-ended status because daily trading volume made it impossible to finish $150M by the November deadline. Now the war chest is nearly spent, the social network still burns $74M a quarter, and management says there is no successor program. On the same call they guided to a $17M to $27M net loss for Q4 and declined to give any 2027 advertising outlook beyond "depends on the success of our social network." Against the August 2025 statement that “before that, I don't think we will be able to breakeven” — Chaoyang Zhang, Chairman and Chief Executive Officer · 2025-08-04, the silence on capital return reads as a company hoarding cash for a fight it has been fighting for years without a critical mass moment.

The setup

Strip it down and Sohu is three things: a ~$450M-a-year game cash cow decelerating at 30%, a social networks platform burning ~$300M a year to chase a user base it cannot yet monetize, and a shrinking buyback that has been doing the work of the equity story. Q3's advertising headline flattered the first; the Q4 guide and the AI-drama admission expose the second; the last $16M exposes the third. The most valuable thing management did on this call was refuse to dress up a project-timing quarter as a trend. The least valuable was not having an answer for what happens to the stock when the bid disappears. advertising campaign demand in China is not the variable to watch here. The variable is whether two hours of daily short drama is a plateau or a beachhead — and Sohu's own CEO does not know.