Zhipu, MiniMax extend rally as AI optimism persists By Investing.com

Zhipu AI (HK:2513) jumped as much as 15% to a record HK$999 after releasing its GLM-5.1 model and raising model prices for the second time this year; Minimax (HK:0100) rose 5.4% to HK$1,053 while the Hang Seng fell 0.3%. Zhipu and Minimax have massively outperformed YTD, up roughly 650% and 205% respectively, as Chinese AI operators roll out agent tools and capitalize on shifts like Anthropic ending third-party tool support. Broader sector sentiment was also boosted by Meta unveiling a new AI model and Anthropic releasing more agentic tools, indicating continued sector momentum but with heightened volatility risk.

Analysis

Large incremental model deployments and agent frameworks are creating a discrete, front-loaded demand shock for inference-optimized compute that will show up in next-quarter hardware bookings and follow-through server refresh cycles over 3–12 months. That favors vertically integrated, channel-focused server OEMs and system integrators who can convert GPU scarcity into pricing power and faster revenue recognition; conversely, it compresses margin optionality for cloud hyperscalers where capital intensity and latency requirements push toward capex, not just software margins.

At the application layer, better AI-driven targeting and agentic automation should lift monetization per user, but the timing of that lift is uneven: meaningful ARPU gains are likely to materialize over 2–4 quarters as products embed models, not instantly. This creates a staging trade: hardware and infra beneficiaries see near-term revenue re-rating, while ad/engagement beneficiaries face a multi-quarter execution cliff where product rollout, measurement attribution, and privacy/regulatory frictions must clear.

Key tail risks write through geopolitics and input economics — renewed export controls on accelerators or a second wave of GPU allocation to cloud providers could reverse hardware pricing and inventory assumptions within 1–3 months. Equally, aggressive model pricing by incumbents or rapid open-source compression of inference costs could cap vendor margin expansion over 6–18 months, making current small-cap sentiment-sensitive names the most vulnerable to a pullback.

Key Decisions for Investors

• Go long SMCI via a 9–12 month call spread sized to 1.5–2.5% of portfolio notional to express near-term server/GPU box demand; target 30–50% upside if bookings beat and gross margins expand, with premium loss limited to 100% of option spend as downside risk.
• Establish a constructive hold on META with a 3–9 month sell-to-open cash-secured put (size 2% portfolio) at a strike ~8–12% below spot to collect yield while committing capital only if ad monetization lags; reward is 10–20% total return if engagement/ARPU improves, tail risk is regulatory or platform slowdown.
• Buy APP 12-month call spread sized to 1% portfolio to capture multi-quarter adtech ARPU improvement from better targeting and LTV modeling; expect 25–40% upside on execution, downside limited to premium paid — watch CPI/IDFA-like measurement shifts as a catalyst to re-evaluate within 6 months.
• Hedge market crowding and China-specific execution risk by allocating 1% to an inverse small-cap China tech hedge or buying 3–6 month protection (puts) on a concentrated domestic AI small-cap basket; this limits drawdowns from sentiment reversals triggered by export controls or weak monetization runs over 1–3 months.