Goldman Sachs: Upgrades Z.AI (02513) rating to "Buy", target price HK$1,560

Goldman Sachs released a research report stating that it has upgraded Z.AI's rating from "Neutral" to "Buy" and revised its DCF-based 12-month target price to HK$1,560. The reasons for the upgrade include: a clearer monetization paththe bank raised its end-2026E ARR forecast to US$3.2 billion (previously forecast at US$2.7 billion), driven by strong token demand and new commercial terms reached with Chinese and global hyperscalers since October; compute infrastructure expansion/balance sheet enhancement to drive scale expansion; better inference gross margins achieved through cost efficiency; and continued progress on harness/Co-Work products. Despite intensifying competition among Chinese AI models, recent concerns surrounding ZCode data retention, shareholder dilution from approximately US$9 billion in equity financing since July 2026, and the upcoming pre-IPO share lock-up expiry (from early January 2027), according to the bank's competitive positioning framework, Z.AI remains one of China's top AI model companies, with its ARR run-rate ranking first among Chinese peers, and it has simultaneously launched frontier models and smaller, more cost-effective flash models. The bank believes that at 12x end-2026E ARR/10x FY27E ARR (compared with MiniMax at 9x/6x), the risk-reward has turned favorable; the bank views Z.AI's valuation premium relative to MiniMax as justified based on the pricing power, cost efficiency, and financial strength framework, given its higher competitive positioning among Chinese AI models. The bank updated its base/bull/bear valuations to US$98 billion/US$157 billion/US$32 billion, corresponding to upside/downside of +149%/+299%/-19%, assuming Z.AI's revenue share among Chinese AI model vendors reaches 22%/33%/14% by 2030E. Due to the accelerated pace of ARR ramp-up, the bank raised Z.AI's 2026-28E revenue forecasts by 3%-12%. Based on DCF valuation, it assumes continued market share gains to 22% by 2030E, and a long-term adjusted EBIT margin of 26% by 2035E. The bank's bull case corresponds to a valuation of HK$2,500 per share, and the bear case corresponds to HK$510 per share. 1) Clearer monetization path: The bank raised its end-2026E ARR to US$3.2 billion (previously Goldman Sachs forecast US$2.7 billion; the company's latest target is US$3.0 billion), driven by strong token demand and new commercial terms reached with Chinese and global hyperscalers since October 2026, which can create incremental high-margin revenue sources. The bank believes that as Z.AI scales its upcoming GLM-5.5 and GLM-6 models to larger pre-training parameter sizes, its data flywheel and significantly increased compute capacity will become the DRIVE for GLM models' performance leap to the next frontier. 2) Compute infrastructure expansion/strong balance sheet to drive model training scaling/inference scaling: The bank believes that its nearly US$10 billion in new equity financing year-to-date (including IPO proceeds), plus (2H26/FY27E) annualized R&D spending intensity of US$1.3 billion/US$1.9 billion, will a substantial increase in compute capacity, which can be used for training (Goldman Sachs forecast: 50% contribution) and inference (50% contribution). The bank views Z.AI's financial strength as a key competitive advantage among independent AI labs. Given that 70%-80% of its revenue comes from domestic China by mid-2026, the bank also notes that Z.AI has a solid domestic growth foundation; at the same time, through overseas API revenue and new revenue-sharing arrangements with hyperscalers, it also has exposure to fast-growing international markets. 3) Better inference gross margins through cost efficiency/optimization: The bank estimates that Z.AI's model intelligence/pricing power, combined with its cost-effective inference architecture, can drive further expansion of inference gross margins, forming sustainable cash generation capability, and is expected to achieve group profitability turnaround by 2029E. 4) Continued progress on harness/Co-Work products: The bank believes that with expanding context windows and improved cost efficiency, Chinese AI models have reached a critical level of intelligence sufficient to deliver cost-competitive virtual employee/white-collar work tasks.