Shares of China AI 'tiger' Zhipu surge 35% after revenue doubles in first earnings report

Revenue rose ~132% to 724 million yuan in 2025 but missed the Reuters-polled mean estimate of 760 million yuan; net adjusted loss widened to 3.18 billion yuan, up 29.1% year-over-year. Shares jumped as much as 35% (about +30% in afternoon trading) after the report; Zhipu listed in Hong Kong in January after a $558 million IPO and recently released its GLM-5 model. CEO Zhang Peng said the firm is accelerating use of domestic Chinese chips to meet a surge in computing demand since February, a key point given prior U.S. export restrictions and the company’s placement on the U.S. Entity List.

Analysis

The market pop signals a willingness to price China-native AI as a distinct growth bucket, which will reallocate near-term enterprise AI capex toward domestic clouds, integrators, and server OEMs. Expect tighter wafer and assembly demand for domestic foundries over the next 6–18 months as model training budgets shift from foreign suppliers to locally sourced compute — that creates a vector for pricing power and longer lead times for vendors selling mature-node GPUs and accelerators.

Two catalyst bands matter: sentiment-driven re-ratings in days-to-weeks (earnings/earnings-call replay, peer prints) and hardware supply/demand dynamics over months (chip procurement announcements, subsidy disbursements, fab capacity expansions). Major reversals will come from either concrete evidence of sustainable ARR from model commercialization or fresh export-control moves that materially tighten access to advanced nodes; watch these on a 3–12 month horizon.

Second-order winners include stack integrators (software+hardware) that reduce reliance on bleeding-edge silicon by co-optimizing models with inference chips and middleware; losers are standalone foreign vendors who lack China-compliant channels or onshore partners and will face slower unit growth. Finally, data-localization and certification regimes will raise switching costs for multinational AI players, advantaging well-capitalized domestic incumbents able to offer end-to-end, compliant solutions.

Key Decisions for Investors

• Buy a 12-month call spread on SMIC (0981.HK): buy a 20% OTM call, sell a 40% OTM call. Size as a 2–3% portfolio bet. Rationale: captures upside from accelerated domestic capex while capping premium spend; target 3x+ payoff if onshore wafer demand tightens within 9–12 months, max loss = premium.
• Buy Alibaba (9988.HK) on any >15% pullback within 3 months; target 30–50% upside over 6–12 months, stop at 12% loss. Rationale: China cloud/AI adoption reallocation should lift cloud ASPs and incremental margin; use equity size 3–5% and tighten stops if macro risk rises.
• Short the HK AI pure-play momentum via buying 3-month ATM puts on the cohort (or short individual names after >20% post-earnings spikes). Size small (1–2% equity) — expect 20–40% mean reversion in 1–3 months if commercial monetization signals disappoint. Max loss = premium; asymmetric payoff if sentiment collapses.
• Pair trade: long Inspur (000876.SZ) equity (6–12 months) and buy 3-month 10% OTM puts on NVIDIA (NVDA) as macro/volatility hedge. Use 1:0.25 notional (equity:hedge). Rationale: capture China-specific server demand while protecting portfolio from a global AI hardware sell-off; target 25–40% upside on Inspur with limited downside via NVDA put protection.