Goldman Sachs reiterates Buy on MiniMax Group with a target price of HK$860 and 189.2% upside

Overview

This is a Goldman Sachs company research report on MiniMax Group (0100.HK). The core view is that MiniMax's risk-reward has become more skewed to the upside after the share unlock. Goldman Sachs believes that MiniMax has a differentiated position among Chinese AI model companies due to the M3 model's cost efficiency, commercialization ability, multimodal product setup, and the upcoming H3 video generation model, at a stage where Chinese AI models are near a globally scalable tipping point.

Core views

The report maintains a Buy rating based on four main points: first, the M3 model is in a favorable position on pricing, token volume, and cost efficiency; second, the imminent launch of the H3 video generation model may improve video generation quality and feature diversity and provide positive valuation mapping with Kling AI; third, compared with DeepSeek, Zhipu, and ByteDance, MiniMax's key advantage lies in cost efficiency and an independent multimodal positioning; fourth, the lock-up period ended on July 9, 2026, and Southbound Connect eligibility is expected to start in August 2026, which may improve supply pressure and financing conditions.

Analysis framework

Goldman Sachs analyzed using a competitive positioning framework, ARR quadrant, model performance roadmap, cost efficiency, peer valuation comparison, and DCF valuation. Valuation was done using DCF with a 12% WACC assumption and 2% terminal growth rate, while also referencing valuation multiples of Chinese and global AI model companies at similar ARR stages.

Methodology notes

• Valuation methodsDCF valuation

  Discounted cash flow valuation

  The report states the target price is based on DCF valuation, using a 12% WACC and 2% terminal growth rate.

• competitive analysisCompetitive Positioning framework

  AI model company competitive positioning framework

  Goldman Sachs used this framework to compare foundation models, multimodal capabilities, cost efficiency, pricing power, and financial strength, and concluded MiniMax stands out in cost efficiency and cost advantage.

• commercialization analysisARR maximizing quadrant

  ARR maximizing quadrant

  The report treats attractive pricing and high token volume as a favorable combination, arguing that the M3 model is in a better quadrant, which helps drive ARR growth.

• factor analysisGS Factor Profile

  Goldman Sachs factor profile

  The framework compares the stock's relative position versus the market and sector peers across Growth, Financial Returns, Multiple, and Integrated dimensions.

Asset mapping & comparison

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

• MiniMax Group (0100.HK)

  Coverage company and rated subject

  Strengths

  The M3 model's cost efficiency is strong, API gross margin is relatively high, multimodal product coverage is comprehensive, the 2026 year-end ARR target is high, and the H3 video generation model is expected to launch soon.

  Weaknesses

  The company remains loss-making, with 2026E EBITDA and net profit both negative, and earnings visibility still needs to be validated.

  Comparison

  The report compares MiniMax with AI model and video generation platforms such as DeepSeek, Zhipu, ByteDance, and Kling AI, concluding that while MiniMax stands out in cost efficiency, it still needs to improve pricing power and financial strength in overall foundation model competition.

  Risks

  Model performance may fall short of expectations, commercialization may slow, cash burn, IP and content-generation risks, and geopolitical as well as U.S.-China technology rivalry risks.

• Chinese AI model industry

  Company's industry and valuation reference group

  Strengths

  Chinese AI models are considered to be near a globally scalable inflection point in intelligence, with substantial TAM potential across text, coding, multimodal, agentic/co-worker, and digital labor.

  Weaknesses

  Foundation model competition is intense, technological iteration is fast, and pricing and cost pressures coexist.

  Comparison

  DeepSeek and Zhipu are viewed as relatively strong in foundation model positioning, ByteDance is stronger in multimodal, and MiniMax is differentiated by cost efficiency and independent full multimodal capability.

  Risks

  Rising global foundation model competition could weaken a single company's model performance, pricing power, and commercialization.

Key data

• 12-month target priceHK$860.00Goldman Sachs maintains a Buy rating associated with this target price.
• Current stock priceHK$297.40Price disclosed on the report cover page.
• Expected upside189.2%Calculated from target price versus current price.
• Market capitalizationHK$93.3bn / US$11.9bnMiniMax market cap as disclosed in the report.
• Enterprise valueHK$86.9bn / US$11.1bnMiniMax enterprise value as disclosed in the report.
• 2026E revenueUS$300.0mnGoldman Sachs revenue estimate.
• 2027E revenueUS$880.1mnGoldman Sachs revenue estimate.
• 2028E revenueUS$2,469.6mnGoldman Sachs revenue estimate.
• 2026E EBITDAUS$(452.5)mnGoldman Sachs expects EBITDA to remain negative.
• 2026 year-end ARR targetUS$1bnThe company remains confident in this target.
• April ARR run-rateabout US$300mnThe report says it doubled from February and reached this level before the M3 launch on June 1.
• API gross margin60%+The report says this is supported by high-margin multimodal model contributions.
• Text model gross marginabout 40%The report says gross margin remains around 40% even with relatively lower API pricing.

Impact & implications

If post-M3 upgrades, H3 launch, inclusion in Stock Connect, and the dissipation of share unlock pressure proceed as expected, MiniMax could benefit from undervaluation, high growth, and multimodal catalysts. The report also notes that the company remains loss-making and in a high investment phase, and upside valuation depends on sustained delivery in model performance, commercialization, cost control, and financing capacity.

Risks

• Model performance weaker than expected in the global foundation model competition.
• Slower-than-expected visibility on the path to profitability.
• Commercialization capability weaker than expected.
• IP and content-generation-related risks.
• Cash burn and self-funding capacity risk.
• Geopolitical risk from intensified U.S.-China technology rivalry.

What to watch

• The quality, feature diversity, and user feedback after the expected H3 video generation model launch in the next 1 to 2 weeks.
• The next update to M3 expected around late July to August 2026, with a focus on coding intelligence, post-training, and reinforcement learning improvements.
• Whether the larger-parameter M3 model in 2H 2026 can improve pricing power for premium AI coding models.
• The impact of Southbound Connect eligibility from August 2026 on liquidity and investor composition.
• Whether actual supply pressure progressively eases after share unlocking.
• Progress on delivering the US$1bn ARR target by end-2026.
• Trends in API and Token Plan revenue contribution and gross margin.