At this week's 2026 World Internet Conference Wuzhen Summit, Sohu founder, Chairman of the Board and CEO Charles Zhang showed rare candor in his speech. He publicly admitted that Sohu had "missed mobile internet" back then, and, against the current industry-wide frenzy over artificial intelligence large models, explained the deeper considerations behind Sohu's relatively "calm" stance.
Missing Mobile Internet Led to Limited Resources; the Large Model Track Is a Resource Battle Sohu "Can't Afford"
In response to outside discussions that Sohu's investment in AI does not seem aggressive enough, Charles Zhang made it clear that this seemingly calm posture is not deliberate, but a clear-eyed choice based on reality.
He pointed out bluntly that because the company missed the golden opportunity in mobile internet development back then, its overall resource reserves today are not abundant enough. On the large model track, which currently consumes massive amounts of capital and computing power, the essence is an asset-heavy game of resource competition. Sohu's current scale and resources mean it "can't afford to play," so it must make pragmatic judgments that fit its own pace.
Don't Get Blindly Swept Along by AI: Opportunities to Make Money Often Lie in Unnoticed Places
When talking about today's red-hot artificial intelligence technology, Charles Zhang poured cold water on the industry and entrepreneurs, reminding everyone "not to get blindly swept along by AI." He believes that real business opportunities and chances to make money are often hidden in unnoticed corners.
As large models and algorithm technologies spread extremely rapidly, today's Chinese-language internet is flooded with various kinds of algorithm-recommended content. This has caused some audiences to experience aesthetic fatigue and begin to crave and turn to real human social interaction for information. Charles Zhang said he does not oppose algorithms themselves, but in Sohu's long-term product design, it has not blindly followed the trend of pursuing an all-algorithm recommendation model; instead, it has retained a unique, differentiated perspective.
