Goldman Sachs has issued a research report raising its rating on Z.AI (02513) from Neutral to Buy, and revising its DCF-based 12-month target price to HK$1,560.

The upgrade is supported by several factors: a clearer monetization path, as the bank lifts its end-2026E ARR forecast to US$3.2 billion (previously US$2.7 billion), driven by strong token demand and new commercial terms reached with Chinese and global hyperscalers since October; expansion of computing infrastructure and a strengthened balance sheet 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 over ZCode data retention, shareholder dilution from roughly US$9 billion in equity financing since July 2026, and the upcoming pre-IPO share lock-up expiry starting in early January 2027, the bank notes that under its 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 alongside 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; it views Z.AI's valuation premium over MiniMax as justified based on pricing power, cost efficiency, and financial strength frameworks, given its higher competitive positioning among Chinese AI models.

The bank updates 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 faster ARR ramp, the bank raises its Z.AI 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. Its bull case corresponds to a per-share valuation of HK$2,500, and its bear case corresponds to HK$510 per share.

1) A clearer monetization path: the bank raises its end-2026E ARR to US$3.2 billion (previously Goldman Sachs forecast US$2.7 billion, with the company's latest target at US$3 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 streams. 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 substantially increased computing capacity will become the driving force behind the GLM models' leap to the next frontier.

2) Computing infrastructure expansion and a 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), combined with annualized R&D spending intensity of US$1.3 billion/US$1.9 billion in 2H26/FY27E, will translate into a substantial increase in computing 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 base; at the same time, through overseas API revenue and new revenue-sharing arrangements with hyperscalers, it is also exposed to fast-growing international markets.

3) Better inference gross margins through cost efficiency optimization: the bank estimates that Z.AI's model intelligence and 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 profit turnaround by 2029E.

4) Continued progress on harness/Co-Work products: the bank believes that as context windows expand and cost efficiency improves, Chinese AI models have reached a critical level of intelligence sufficient to deliver cost-competitive virtual employee and white-collar work tasks.