China’s Zhipu Gets Lone ‘Short’ Call After 1,100% Rally

Felix Wang of Hedgeye said Zhipu's fair value on the short side is HK$407, arguing the company lacks pricing power after DeepSeek's V4 launch triggered a price war among Chinese AI firms. The call is bearish for Zhipu and highlights intensifying competition in China's AI sector. The article is mainly analyst commentary, so the broader market impact is likely limited.

Analysis

The key issue is not the headline bear case, but the transition from a scarcity-driven AI equity trade to a utility-style pricing environment. Once model capability becomes more reproducible, value shifts from raw performance to distribution, latency, compliance, and enterprise integration; that structurally compresses margins for standalone model vendors before it helps the application layer. In other words, the market is likely underestimating how quickly “AI winner” multiples can mean-revert once buyers realize they can multi-source comparable inference at lower cost.

The second-order winner is the broader ecosystem that can arbitrage falling model costs: cloud infrastructure, data tools, and vertical software with embedded workflow lock-in. If model pricing turns into an auction, the near-term losers are firms relying on inference revenue to subsidize R&D, because they face a brutal fixed-cost problem with no easy offset. That dynamic can also accelerate consolidation, as smaller players either get acquired on the cheap or forced into a race to the bottom that burns cash faster than public markets currently tolerate.

Catalyst timing matters: this is more of a months-long valuation reset than a days-long trade, unless there is a near-term sign of customer churn or pricing cuts. The main upside risk to the short thesis is a step-change in enterprise adoption that more than offsets lower unit pricing, but that usually takes multiple quarters to show up in revenue and rarely prevents near-term multiple compression. On the contrary, any evidence of aggressive discounting or guidance tied to lower average selling prices should widen the gap between market price and fair value quickly.

The contrarian view is that investors may be extrapolating a price war too mechanically and missing platform power. If one vendor can become the default distribution layer for Chinese enterprise AI, monetization may migrate from model fees to bundle economics, making headline price cuts less important than seat expansion and retention. But until there is proof that usage growth can outrun pricing pressure, the setup still favors being short the most expensive pure-plays and long the picks-and-shovels names.

Key Decisions for Investors

• Short the most expensive China AI pure-plays on any post-rally strength; treat this as a 3-6 month multiple-compression trade with a stop if management shows accelerating enterprise bookings or a major platform partnership.
• Rotate long China cloud/infrastructure beneficiaries versus standalone model vendors over the next 1-2 quarters; the risk/reward favors businesses selling compute and tooling into the price war rather than financing it.
• If liquid instruments are available, buy downside protection on high-beta AI names with 6-9 month tenor; the convexity is attractive because revenue deceleration can re-rate stocks faster than consensus models.
• Use a pair trade: short a premium-valued AI model developer against a long position in a software or cloud beneficiary with recurring revenue; this isolates the pricing-power reset while reducing market beta.
• Avoid chasing any relief rally until there is evidence that lower model prices are converting into materially higher usage and retention; absent that, rallies are likely sellable over the next several weeks.