MiniMax ARR Significantly Beats Expectations; Near-Term Profit Pressure Does Not Change HK$900 Price Target

Overview

This report assesses MiniMax's first-half results, ARR growth, model launch plans, gross margin pressure, and valuation. Morgan Stanley believes commercialization is progressing materially faster than expected, but has adopted more cautious assumptions for near-term gross margins and R&D costs. Long-term revenue upgrades offset the increase in near-term losses, leading the bank to reiterate Overweight and maintain its HK$900 price target.

Core views

First, MiniMax's commercialization progress has significantly exceeded Morgan Stanley's previous expectations. Management stated that ARR had exceeded US$800mn as of August 2026, above the bank's US$600mn expectation, with toB accounting for 80% and toC for 20%. Morgan Stanley accordingly raised its year-end ARR forecast from US$1bn to US$1.3bn. The report also provides another set of figures attributed to management: ARR reached RMB800mn in August and is expected to exceed RMB1bn by year-end; the original report's US dollar and renminbi figures are inconsistent. Operating data likewise indicate rapid demand expansion: Token consumption increased 20-fold from January to July 2026, while the number of customers rose from 200,000 at the end of 2025 to 2mn currently, representing 10-fold growth. The acceleration in ARR was mainly driven by adoption of the M3 text model and H3 video model, as well as improved commercialization execution. The report summarizes four mechanisms: improved platform stability and transaction-processing capacity per second enable customers to deploy at production scale; use cases have expanded from coding and agents to broader office productivity applications; industry price increases have created a more favorable competitive environment; and H3 began contributing incremental growth in August. M3.1, M3 Pro, and H3.1 are the main product pipeline for the second half of 2026, and existing computing capacity is expected to support their iteration. MiniMax is also accumulating computing capacity and plans to train a 10-trillion-parameter model for 2027. Second, first-half gross margin was weaker than expected, but management attributed the pressure to one-off factors during the M3 launch phase rather than structural issues. AI-native products themselves already generate positive gross margins and did not weigh on overall gross margin. The main negative factors included investments to improve service stability and customer compensation during the model transition; lower inference efficiency early in the launch; a higher proportion of low-margin Token packages to accelerate adoption; and an increased mix of low-margin text models relative to higher-margin multimodal models. Management expects gross margin to improve sequentially in the second half of 2026, supported by continued declines in inference costs, stronger monetization discipline, and premium models such as M3 Pro enabling higher pricing and better unit economics. Regarding earnings forecasts, after incorporating the first-half 2026 results, Morgan Stanley raised its 2026, 2027, and 2028 revenue forecasts by 25.3%, 25.0%, and 24.6%, respectively, to reflect higher ARR expectations. The financial summary provides net revenue forecasts of US$500mn for 2026, US$1,497mn for 2027, and US$3,727mn for 2028, corresponding to revenue growth of 532.3%, 199.5%, and 149.0%. However, the bank also lowered its gross margin forecasts and raised its R&D expense assumptions due to large-parameter model training and higher computing costs, increasing its 2026 to 2028 non-IFRS operating loss forecasts by 29.6%, 14.4%, and 15.4%, respectively, and its loss-per-share forecasts by 34.6%, 16.2%, and 18.0%, respectively. This means stronger revenue growth will not immediately translate into narrower near-term losses. On valuation, the increase in near-term net loss forecasts offset the higher long-term revenue forecasts, leaving the price target unchanged at HK$900. Morgan Stanley uses a DCF valuation with assumptions of a 15% WACC and a 3% perpetual growth rate. The price target implies 25x 2027 P/S, below the company's historical average one-year forward P/S of 137x. Based on the bank's forecasts, the stock currently trades at 10x 2026 P/ARR and 9x 2027 P/S. Morgan Stanley therefore considers the valuation undemanding and believes the expansion of the global AI foundation model market, MiniMax's technological capabilities, full-modality products, application diversity, and scalable business model could provide valuation upside. The report also constructs three scenarios centered on the next-generation large-parameter model scheduled for release in October 2026. The bull case assumes that the new model becomes globally state-of-the-art and outperforms peers, boosting customer demand, market share, and pricing. It forecasts 2027 revenue of US$2.5bn and applies 40x 2027 P/S to derive HK$2,230. The base case assumes that the model matches the performance of globally state-of-the-art models, driving nonlinear 2027 revenue growth to US$1.5bn, and applies 25x P/S to derive HK$900. The bear case assumes that the model fails to outperform peers and revenue maintains only linear growth, forecasting 2027 revenue of US$1bn and applying 10x P/S to derive HK$220. Model performance and commercialization effectiveness are therefore the key determinants of the revenue and valuation ranges.

Analysis framework

Morgan Stanley first assesses the pace of commercialization using management-disclosed ARR, Token consumption, and customer numbers, and then analyzes how model performance, platform stability, application expansion, and industry pricing translate into revenue. The bank subsequently breaks down the temporary factors that weighed on gross margin during the M3 launch and adjusts its 2026 to 2028 forecasts for revenue, gross margin, R&D spending, and losses. Finally, it uses DCF to determine the base-case price target and presents a valuation range through bull, base, and bear scenarios based on next-generation model performance, 2027 revenue, and P/S multiples.

Methodology notes

• Valuation MethodDCF Discounted Cash Flow

  Discounted Cash Flow Valuation

  The report discounts the company's future cash flows to present value and uses a 15% WACC and 3% perpetual growth rate to derive a price target of HK$900.

• Valuation MethodPS valuation

  P/S Valuation and Scenario Multiples

  Because the company remains loss-making, the report uses revenue multiples as a supplementary valuation check: current 2027 P/S is approximately 9x, the base-case price target implies 25x, and the bull and bear cases use 40x and 10x, respectively.

Asset mapping & comparison

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

• MiniMax (0100.HK; standardized ticker 00100.HK)

  The report believes the ARR beat, model iterations, and accelerating application commercialization support long-term revenue growth and valuation potential.

  Strengths

  Advanced technology, full-modality capabilities, diverse AI applications, a scalable business model, and rapidly growing Token consumption and customer base.

  Weaknesses

  First-half gross margin was below expectations, with inference optimization, low-margin packages, and a relatively high mix of text models; large-model training and computing investment are increasing near-term losses.

  Comparison

  The stock currently trades at 9x 2027 P/S, the DCF price target implies 25x, and the historical average one-year forward P/S is 137x.

  Risks

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

Key data

• ARR as of AugustOver US$800mnDisclosed by management and above Morgan Stanley's US$600mn expectation; a detailed section separately states RMB800mn.
• Year-End ARR ForecastUS$1.3bnRaised by Morgan Stanley from US$1bn; a detailed section separately states that management expects more than RMB1bn.
• ARR Business MixtoB 80%; toC 20%Based on the ARR composition disclosed as of August.
• Token Consumption Growth20-foldFrom January to July 2026.
• Number of Customers2mnUp from 200,000 at the end of 2025 to the current level, representing 10-fold growth.
• 2026 to 2028 Revenue Forecast Upgrades25.3%, 25.0%, 24.6%Reflecting higher ARR expectations.
• 2026 to 2028 Revenue ForecastsUS$500mn, US$1,497mn, US$3,727mnCorresponding to the annual net revenue forecasts in the report's financial summary.
• Changes in 2026 to 2028 Non-IFRS Operating Loss ForecastsIncreased by 29.6%, 14.4%, 15.4%Driven by lower gross margin forecasts and higher assumptions for R&D and computing costs.
• Changes in 2026 to 2028 Loss-Per-Share ForecastsIncreased by 34.6%, 16.2%, 18.0%Corresponding to the three forecast years, respectively.
• Core DCF AssumptionsWACC 15%; perpetual growth rate 3%Used to derive the HK$900 price target.
• Current Valuation10x 2026 P/ARR; 9x 2027 P/SBased on Morgan Stanley's forecasts.
• Scenario Price TargetsBull HK$2,230; base HK$900; bear HK$220Corresponding to 40x, 25x, and 10x 2027 P/S, respectively.

Impact & implications

The report believes MiniMax's demand growth and commercialization pace have already exceeded prior expectations, prompting meaningful upgrades to medium- and long-term revenue forecasts. However, service investments during the initial model launch, inference efficiency, product mix, and training-related computing costs are pressuring near-term gross margins and losses. Whether the next-generation model can match or exceed globally leading levels will determine whether 2027 revenue achieves nonlinear growth and will be the central variable defining the price-target scenario range.

Risks

• Geopolitical risks may affect the company's operations or model development.
• Intensifying market competition and price wars may reduce pricing and gross margins.
• If model performance lags peers, customer demand, market share, and 2027 revenue growth may fall below the base case.

What to watch

• Monitor the launches and commercialization contributions of M3.1, M3 Pro, and H3.1 in the second half of 2026.
• Monitor whether the next-generation large-parameter model scheduled for release in October 2026 can match or exceed globally leading levels.
• Monitor whether year-end ARR can reach Morgan Stanley's US$1.3bn forecast.
• Monitor whether declining inference costs, improved monetization discipline, and premium model launches can drive sequential gross margin recovery.
• Monitor the company's progress in accumulating computing capacity and bearing training costs for its 10-trillion-parameter model planned for 2027.