Have Chinese internet giants hit bottom? Bernstein: AI costs are being overly pessimistically interpreted; focus on Tencent.

Have Chinese internet giants hit bottom? Bernstein: AI costs are being overly pessimistically interpreted; focus on Tencent.

After a deep adjustment in the first half of the year, China’s internet sector is ushering in a wave of sentiment recovery.

According to Chasing Wind Trading Desk, Bernstein’s latest research report believes that the market’s pessimistic interpretation of AI inference costs has been seriously distorted. Coupled with catalysts such as the launch of Tencent’s Hy3 model and Alibaba Cloud’s better-than-expected growth, the sector’s valuation has rebounded from the historical lows of 2022–2023, and the risk-return ratio has improved.

The release of Tencent’s Hy3 model and Xiaowei assistant beginning to push to users is the core driving force behind the recent warming sentiment in the sector.

Bernstein analyst Robin Zhu pointed out in the report that previous AI cost narratives circulating in the market—such as platforms needing to indefinitely subsidize inference costs—are “seriously inaccurate and excessively pessimistic.” Meanwhile, Alibaba Cloud’s revenue growth accelerated to the mid-40% range in the second quarter, with e-commerce profits also beating expectations, jointly driving the sector’s overall rebound.

Bernstein maintains "outperform" ratings for Tencent, Alibaba, NetEase, JD.com, and Boss Zhipin. Tencent’s target price is HKD 780, representing about 69% upside from the current price; JD.com’s target price is USD 40, corresponding to a projected 2026 P/E ratio of just about 8.4 times.

The report explicitly points out that Tencent’s current forward P/E ratio of 11–12 times is undervalued. As various negative factors gradually dissipate, the stock price is expected to return to the historical average of 17–18 times since 2021.

AI Cost Panic Overblown, Inference Fees Not a Bottomless Pit

One of the main short theses around Tencent and Alibaba is that the launch of AI will lead to inference costs greatly eroding profits, while monetizing AI on the consumer side remains far off.

Bernstein refutes this: The explosive growth of inference costs, in fact, requires agentic transaction volume and GMV to take off simultaneously.

Data shows chatbot-style conversations usually only consume a few hundred tokens, whereas agentic transactions average about 50,000 tokens consumed. In other words, Tencent’s token consumption will only grow exponentially when agentic transaction volume truly scales up.

The report believes there is a time lag between user engagement and monetization growth, but historical patterns indicate the issue is “when” rather than “if.”

On capital expenditures, Tencent’s capex accounted for 10.9% of revenue in the past 12 months, lower than Alibaba’s 12.3%.

Relative to operating cash flow, Tencent has much greater room to expand capex than Alibaba. For Alibaba Cloud, second quarter revenue growth accelerated to the mid-40% range, and a low double-digit profit margin means 35–40% incremental cash profit margin. Alibaba also stated it will continue raising computing power prices this year.

Tencent AI Strategy Becoming Clearer, Hy4 Timeline Key Observation Point

Bernstein’s report details current progress in Tencent’s AI layout. Hy3 model has made significant advances based on the April preview, Xiaowei assistant has begun gray-push to users with positive early feedback, and agent2agent agreements signed with major Android phone makers create synergy.

The report notes that the Hy team and WeChat team have operated independently, with management explaining this is to avoid mutual dependency and prevent delays from one side affecting the other. During this period, the WeChat team has leveraged cutting-edge Chinese models like DeepSeek to drive product iteration, and monthly active users of Yuanbao and Workbuddy have reached 8–9 million, providing feedback sources for the Hy3 team.

Bernstein estimates that with continuous improvements in Hy4 and subsequent models, the degree of collaboration between the two teams will gradually deepen. Based on the pattern of Chinese AI labs expanding parameter scale by 2–2.5 times on average, the Hy4 preview may be released near the end of the year. The report believes most of the AI functions Tencent aims to realize within WeChat are relatively straightforward under existing agentic AI standards, meaning self-developed models will more quickly reach a “good enough” level.

On monetization pathways, the report highlights that several trillion RMB GMV of WeChat mini-programs is Tencent AI’s core monetization opportunity, realized on the merchant side. This forms a sharp contrast with the pessimism of Western market investors, who have largely given up hope for consumer-side AI monetization.

Alibaba Business Restructuring Boosts Valuation, Seasonal Catalysts Worth Watching

Bernstein adopts a cautiously optimistic attitude towards Alibaba’s recent organizational restructuring. Incorporating Hema into China e-commerce, and PingTouGe chip business into cloud computing, will help boost growth rates in businesses investors focus on, with positive implications under a sum-of-the-parts (SOTP) valuation framework.

The report points out that Neocloud customers’ arrangement to sell PingTouGe chips and lease back computing power helps reduce capital intensity reported for Alibaba Cloud, while also supporting the narrative of external chip sales revenue. Also, Alibaba owns around 5% stake in Changxin Technology, whose potential value cannot be ignored.

Historical patterns deserve attention as well: In both 2024 and 2025, Alibaba’s stock price reached stage highs around the September Yunqi Conference. Bernstein believes progress of Changxin Technology’s IPO and the Yunqi Conference will be critical catalyst windows in the coming months.

NetEase, JD.com, and Boss Zhipin: Low Valuations Combined with Earnings Growth

Regarding other key picks, Bernstein is positive on NetEase, JD.com, and Boss Zhipin, with core logic based on low valuations and earnings growth.

For NetEase, “Sea of Remnants” will be launched domestically on July 23. The report expects some stock price pullback by then, but low second-half revenue base, accelerated buybacks, expectation of inclusion in Southbound flows, and news on “Ananta” constitute a clear catalyst path. Bernstein’s backtesting shows a high correlation between NetEase’s accelerated buybacks and subsequent stock performance over the past decade.

JD.com and Boss Zhipin are currently trading at low single-digit forward P/E—about 4 times after excluding cash and investments for Boss Zhipin—with both companies maintaining double-digit earnings growth. Bernstein raised JD.com’s earnings forecast, projecting non-GAAP net profit for 2026 to be up about 8.8% over previous estimates, mainly reflecting faster-than-expected narrowing of losses in the food delivery segment and ongoing margin improvement.

By contrast, the report is more cautious about Meituan and Pinduoduo. Meituan’s upside is limited after the recent rebound, and its AI investment is still at an early stage; Pinduoduo’s management continues to avoid communicating capital returns with investors, which the report views as a core obstacle for long-term investors.

Valuations Still Near Historical Lows, But Conditions for Recovery Are Building

In terms of overall valuation, Chinese internet stocks covered by Bernstein are down about 16% this year on average (including recent rebound), currently trading at average projected P/E of 14.3x for 2026 and 12.1x for 2027, the latter only slightly above the historical lows of 2022–2023, when fundamentals were severely suppressed by China’s “zero-COVID” policy.

On free cash flow/enterprise value yield, Tencent and JD.com are about 5%, NetEase about 10%, Boss Zhipin close to 20%. JD.com, Boss Zhipin, and Pinduoduo all have net cash exceeding 40% of market cap, with management at JD.com and Boss Zhipin actively repurchasing shares.

Bernstein believes that Tencent’s current low valuation reflects a series of “presently reasonable but ultimately transient” concerns, including slowing game growth and expense pressure from AI investment. Clear reversal signals include operating profit forecasts bottoming and turning up, breakout new games (such as “Delta Force” currently performing strongly), or the release of Hy4 truly settling AI capability debates. The report believes, as these factors materialize, Tencent’s stock is unlikely to remain at an 11–12x forward P/E level for long.

 

 

~~~~~~~~~~~~~~~~~~~~~~~~

The above excellent content comes from Chasing Wind Trading Desk.

For more detailed interpretation, including real-time coverage, frontline research and more, please join [Chasing Wind Trading Desk Annual Membership]

Risk Warning and DisclaimerThe market involves risks, and investment requires caution. This article does not constitute personal investment advice and does not take into account the special investment objectives, financial situation, or needs of individual users. Users should consider whether any opinions, viewpoints, or conclusions in this article are suitable for their particular circumstances. Invest at your own risk.