中国互联网:现在是坏消息......还是好消息?

China Internet: Is bad news now... good news? A bounce… or something more?In this note we outline some latest thoughts on theChina Internet sector, taking stock of the latest developments and data points across theindustry, recent discussions with the large caps we cover, and investor feedback we’d heardin a few dozen meetings with investors on the road in North America. Robin Zhu+852 2123 2659robin.zhu@bernsteinsg.com Charles Gou+852 2123 2618charles.gou@bernsteinsg.com If it can’t get worse…We started 2026 with a cautious top-down view. Some concerns onmacro and regulation persist, though compared with December, investor perceptions andsector valuations have caught down in recent months. Some of the narratives we’d heard onAI costs (e.g. unconstrained inference costs even while agentic AI engagement remains low)had become misinformed and overly negative. At the risk of sounding wishful (consumptionremains anemic, but non-growth in Q2 was at least partly influenced by high 6.18 comps),we do wonder if the NBS’ willingness to report below-target GDP growth might precede aperiod of less repressive macro policy. Min-Joo Kang+852 2123 2644minjoo.kang@bernsteinsg.com Hyrum Caesar+81 3 6777 6979hyrum.caesar@bernsteinsg.com Is AI capex good again?The combination of Tencent’s Hy3 release and Alicloud growthremaining on track (plus e-commerce profits proving better than feared) likely catalysedthe recent bounce in sector sentiment. For Tencent, Xiaowei feedback and the timing ofthe next Hy4 pre-train represent the next data points to watch. Alibaba moving T-head intoits Cloud segment felt a little cute, but should contribute to revenue growth. Chip salesto neocloud businesses (who then lease back compute?) feels circular, but are probablynecessary on the path to a possible spin-off T-head IPO. Beyond tokenmaxxing… the next leg of AI adoption.More cost-efficient use of AIshould ultimately be bullish for adoption. Open source progress is bullish for everyone…except closed source AI labs rushing to IPO. That said, the transition between the top 1%of power users spending heavily, and more broad-based adoption, bears monitoring. Incontrast with Western markets where investors have largely abandoned hope for consumerAI monetisation, trillions of RMB of Weixin Mini Program GMV represent the obviousopportunity for Tencent, with monetisation happening via merchants. Beneficiaries of cheaper, more capable AI.Google Gemini using Capcom to justify theusefulness of its AI product felt like a useful data point in the gaming disruption debate. Weremain of the view that AI solves for production, but not IP credibility or multiplayer networkeffects. Some give-back is likely in NetEase’s shares after Sea of Remnants launches, butlow H2 comps, higher buybacks, likely southbound inclusion, and eventual Ananta newsremains an attractive catalyst runway in our view. AI enabling better matching should helpBoss Zhipin improve ARPJ and enterprise ARPU. Tactical stock thoughts.Tencent remains modestly valued in our view. The “all-clear” forshorter-term investors will likely require operating profit revisions to bottom, a blockbusternew game (Delta Force continues to perform strongly), or perhaps Hy4 to truly settle theAI debate. But we doubt the stock stays on 11-12x forward PE as these things play out.We like NetEase, and JD and BZ feel too cheap to us, given double-digit profit growth.Meanwhile, we won’t have been the only ones to notice the seasonal trade in Alibaba’sshares into the Apsara cloud conference… or the value of its CXMT and Kimi stakes. BERNSTEIN TICKER TABLE INVESTMENT IMPLICATIONS Was that rock bottom? From a starting point where valuation multiples in the sector fell mostly to trough 2022-2023 levels,Tencent’s Hy3 model release and Alibaba Q1 earnings sounding better than feared - along with extreme semis positioningdetonating - have catalysed a rebound in our sector coverage, and investor sentiment. The ROI on hyperscaler capex willcontinue to be debated. But some of the bearish views we’d heard around agentic AI inference costs had become badlymisinformed, and almost certainly overly negative. Tactically, the set-up for the next few months feels constructive to us. Q2 reporting should be somewhat de-risked at this point.Tencent’s Xiaowei roll-out will be gradual, but early feedback has been solid, and it’s not hard to see how the functionality candovetail with agent2agent agreements signed with the main Android smartphone makers. The full Hy3 release has helped toimprove perceptions of the company’s model-building credentials. For Alibaba, the decision to move Freshippo and T-head intoChina E-commerce and Cloud segments should help to boost reported growth in these segments, on top of ongoing capex andpotentially more price increases for compute supporting the latter. The CXMT IPO and Apsara cloud conference line the catalystrunway in the next few months. Elsewhere, some give-back is likely in NetEase’s shar