Chinese open-weight models gain share, while ARR growth and new model launches at MiniMax and Z.AI provide near-term catalysts

Overview

This report previews MiniMax and Z.AI's first-half 2026 results, focusing on ARR, revenue mix, gross margin, R&D investment, and operating losses, while outlining new model launches over the coming months. Morgan Stanley believes that the rising market share of Chinese open-weight models, further upside potential for both companies' ARR, and near-term product catalysts support its positive view, and therefore reiterates its Overweight ratings on both companies.

Core views

The report's overall assessment is that recurring revenue at MiniMax and Z.AI still has upside potential as Chinese open-weight models gain greater market share through improved model capabilities. Morgan Stanley views ARR as the key metric in this results preview, estimating August 2026 ARR of US$600mn for MiniMax and US$1.2bn for Z.AI, and expects these figures to reach US$1bn and US$2bn, respectively, by year-end. Near-term model iterations are important catalysts for achieving this growth trajectory, while the two companies trade at 15.5x and 33.7x P/ARR, respectively, based on expected year-end ARR. The report reiterates its Overweight ratings on both companies while maintaining an In-Line view on the relevant Asia-Pacific industry. For MiniMax, recently disclosed ARR increased from US$150mn+ in February 2026 to US$400mn+ in May. Morgan Stanley estimates that ARR reached US$600mn in August and expects it to reach US$1bn by year-end; the ARR uplift from the recent H3 launch suggests that this forecast may have further upside. In terms of the product roadmap, the report looks forward to the launch of M3.1 in August and M3 Pro, with approximately 3T parameters, in September or October, and believes that successive model launches could sustain commercialization momentum. For MiniMax's 1H26 P&L, the report forecasts revenue of US$115mn, comprising US$55mn from MaaApp and US$60mn from MaaS. Gross margin is expected to improve from 25.4% in 2025 to 29.5%: greater inference efficiency and a higher contribution from the more favorable-margin MaaS business could offset the adverse impact of a larger share of the text business relative to the multimodal business. Non-IFRS operating loss is expected to be US$228.3mn, narrower than the US$290.5mn loss in 2025; 1H26 R&D expenses are expected to be US$230mn, versus US$253mn in 2025. The report therefore presents an outlook combining revenue growth, gross-margin improvement, and narrowing losses. For Z.AI, the report states that its most recently disclosed ARR was US$1bn in June 2026 and lists the March comparison figure in the original text as US$250bn+. Morgan Stanley estimates August ARR at US$1.2bn and expects it to reach US$2bn by year-end; the recent GLM 5.3 launch and expansion of third-party channels could provide additional upside. The next clearly identified product catalyst is the planned October launch of GLM 6.0, with approximately 3T parameters. For Z.AI's 1H26 P&L, the report forecasts revenue of Rmb967mn, comprising Rmb267mn from on-premises deployment and Rmb700mn from cloud services. Gross margin is expected to decline from 41% in 2025 to 31.9%, mainly because the contribution from the relatively lower-margin cloud business is expected to rise sharply from 26% in 2025 to 72% in 1H26; although price increases are expected to lift the cloud-services gross margin from 19% to 25%, this would still be insufficient to fully offset the shift in revenue mix. The report expects a non-IFRS operating loss of Rmb2.3bn, narrower than the Rmb3.2bn loss in 2025; R&D expenses are also expected to decline from Rmb3.2bn in 2025 to Rmb2.3bn in 1H26. The core logic is that cloud-business expansion will drive revenue and ARR, but the change in business mix will weigh on overall gross margin in the short term. On valuation, Morgan Stanley uses DCF for both companies. The MiniMax and Z.AI models both assume a 15% WACC and a 3% terminal growth rate; their respective target prices imply 32x and 42x 2027 P/S. The report also compares their pricing based on expected year-end 2026 ARR, with MiniMax at 15.5x P/ARR and Z.AI at 33.7x P/ARR. MiniMax's upside scenarios include launching a globally leading model and easing competition, while downside factors include geopolitical risks, intensifying competition and price wars, and model performance lagging peers; Z.AI's upside scenarios include global expansion through partnerships with overseas cloud-service providers and easing competition, while downside factors include computing-power constraints and geopolitical risks.

Analysis framework

The report first assesses the pace of commercialization using ARR, including recently disclosed figures, August estimates, and year-end forecasts. It then separately breaks down the two companies' 1H26 revenue sources, business mix, gross-margin drivers, R&D expenses, and non-IFRS operating losses. It subsequently links the launch schedules of products such as M3.1, M3 Pro, and GLM 6.0 to potential ARR upside, and finally evaluates valuation through P/ARR comparisons and a DCF target-price framework while outlining upside and downside scenarios for each company.

Methodology notes

• Valuation MethodDCF Discounted Cash Flow

  DCF Valuation

  The report discounts the companies' future cash flows to present value, applying a 15% WACC and a 3% terminal growth rate to both MiniMax and Z.AI; the resulting target prices imply 32x and 42x 2027 P/S, respectively.

• Valuation Method

  P/ARR Relative Valuation

  The report compares the pricing of the two AI model companies using market capitalization as a multiple of annual recurring revenue. Based on Morgan Stanley's year-end 2026 ARR forecasts, MiniMax and Z.AI trade at 15.5x and 33.7x, respectively.

• Industry/Sector Analysis FrameworkVolume-price decomposition

  Revenue and Business-Mix Breakdown

  The report breaks MiniMax's revenue into MaaApp and MaaS and Z.AI's revenue into on-premises deployment and cloud services, using changes in each business's contribution and gross margin to explain overall revenue and gross-margin trends.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• MiniMax (0100.HK)

  The report reiterates its Overweight rating, viewing the ARR uplift from H3, the launches of M3.1 and M3 Pro, and the relatively low P/ARR as positive catalysts.

  Strengths

  ARR increased from US$150mn+ in February to US$400mn+ in May; improved inference efficiency and a higher MaaS contribution are expected to lift the 1H26 gross margin to 29.5%.

  Weaknesses

  A non-IFRS operating loss of US$228.3mn is still expected in 1H26, with R&D expenses forecast at US$230mn.

  Comparison

  Trades at 15.5x P/ARR based on expected year-end 2026 ARR, below Z.AI's 33.7x.

  Risks

  Geopolitical risks, intensifying competition and price wars, and model performance lagging peers.

• Z.AI CO., LTD. (2513.HK)

  The report reiterates its Overweight rating, believing that GLM 5.3, third-party channel expansion, and the October launch of GLM 6.0 can support continued ARR growth.

  Strengths

  ARR is expected to reach US$1.2bn in August and US$2bn by year-end; the cloud-services gross margin is expected to improve from 19% to 25%, while the operating loss is expected to narrow.

  Weaknesses

  The cloud-business contribution rises from 26% to 72%, which is expected to reduce the overall gross margin from 41% to 31.9%; a 1H26 loss of Rmb2.3bn is still expected.

  Comparison

  Trades at 33.7x P/ARR based on expected year-end 2026 ARR, above MiniMax's 15.5x.

  Risks

  Computing-power constraints and geopolitical risks.

Key data

• MiniMax ARRFebruary 2026 US$150mn+; May US$400mn+; August estimate US$600mn; year-end forecast US$1bnThe report believes the year-end ARR forecast may have further upside following the H3 launch.
• MiniMax 1H26 RevenueUS$115mnMaaApp contributes US$55mn and MaaS contributes US$60mn.
• MiniMax 1H26 Gross Margin29.5%Up from 25.4% in 2025, supported by improved inference efficiency and a higher MaaS contribution.
• MiniMax 1H26 Non-IFRS Operating LossUS$228.3mnThe 2025 loss was US$290.5mn; 1H26 R&D expenses are expected to be US$230mn, versus US$253mn in 2025.
• MiniMax Model Launch ScheduleM3.1 in August 2026; M3 Pro with approximately 3T parameters in September or October 2026The report views successive model launches as near-term catalysts.
• Z.AI ARRJune 2026 US$1bn; August estimate US$1.2bn; year-end forecast US$2bnThe report's original text lists the March 2026 comparison figure as US$250bn+ and believes GLM 5.3 and third-party channel expansion could provide upside.
• Z.AI 1H26 RevenueRmb967mnOn-premises deployment contributes Rmb267mn and cloud services contribute Rmb700mn.
• Z.AI 1H26 Gross Margin31.9%Down from 41% in 2025; the cloud-business contribution rises from 26% to 72%, while the cloud-services gross margin increases from 19% to 25%.
• Z.AI 1H26 Non-IFRS Operating LossRmb2.3bnThe 2025 loss was Rmb3.2bn; R&D expenses are expected to decline from Rmb3.2bn to Rmb2.3bn.
• Z.AI Model Launch ScheduleGLM 6.0 with approximately 3T parameters in October 2026The next clearly identified product catalyst listed in the report.
• Expected Year-End 2026 P/ARRMiniMax 15.5x; Z.AI 33.7xBased on Morgan Stanley's year-end 2026 ARR forecasts.
• Key DCF AssumptionsWACC 15%; terminal growth rate 3%MiniMax's target price implies 32x 2027 P/S, while Z.AI's target price implies 42x 2027 P/S.

Impact & implications

The report believes that the rising share of Chinese open-weight models and rapid product iteration could continue translating into ARR expansion. MiniMax's inference efficiency and improved MaaS mix are expected to lift gross margin; Z.AI, meanwhile, is trading lower margins for faster revenue and ARR growth through cloud-business expansion, although the higher contribution from the lower-margin cloud business will weigh on overall gross margin in the short term. Both companies are expected to narrow their non-IFRS operating losses, while their ability to launch new models as scheduled, achieve year-end ARR targets, and manage competitive and computing-power risks will be critical to the realization of the report's positive view.

Risks

• MiniMax faces geopolitical risks.
• MiniMax may be affected by intensifying competition and price wars.
• MiniMax's model performance may lag peers.
• Z.AI faces computing-power constraints.
• Z.AI faces geopolitical risks.

What to watch

• Watch whether MiniMax launches M3.1 as scheduled in August 2026 and M3 Pro, with approximately 3T parameters, in September or October.
• Watch whether Z.AI launches GLM 6.0, with approximately 3T parameters, as scheduled in October 2026.
• Watch whether MiniMax and Z.AI's ARR reaches US$1bn and US$2bn, respectively, by year-end.
• Watch whether MiniMax's H3 continues to drive ARR above the report's forecast.
• Watch whether Z.AI's third-party channel expansion and partnerships with overseas cloud-service providers deliver additional growth.
• Watch the differences between both companies' actual revenue, gross margins, and non-IFRS operating losses and the 1H26 forecasts.