MiniMax management meeting takeaways: model efficiency, in-house computing power, and multimodality are the core highlights

Overview

This report is an HSBC company research meeting note on MiniMax (0100.HK), focusing on key takeaways from a management meeting. On July 23, 2026, HSBC met with MiniMax's CEO, COO, and capital markets team in Shanghai, with discussions centered on the company's progress in the open platform, model strategy, multimodal models, and AI infrastructure.

Core views

The core view is that MiniMax is well positioned to drive solid gross margins for its open platform against a backdrop of strong demand and high inference efficiency, with a long-term open platform gross margin target in the high double-digit percentage range. On the model side, the company plans to launch M3 Pro and continue researching models with larger parameter scales; management also views the multimodal direction as an important avenue for growth and capability enhancement. On the infrastructure side, the company's self-developed computing power and cluster facilities help reduce reliance on cloud service providers and improve cost efficiency amid rising chip and memory prices.

Analysis framework

The report mainly uses management interviews and business progress tracking, organizing the meeting takeaways around three dimensions: objectives, model strategy, and AI infrastructure strategy. Valuation, full risk analysis, and more detailed discussion are referred to HSBC's previous report published on June 2, 2026.

Methodology notes

• Company fundamentalsManagement meeting note

  By meeting with the CEO, COO, and capital markets team, the report distills the company's strategy, product roadmap, and progress in infrastructure buildout.

  This method is more focused on event updates and qualitative tracking, making it suitable for supplementing investors' understanding of the company's near-term product cadence and long-term strategic direction.

• Rating frameworkHSBC equity rating framework

  The upside or downside of the target price relative to the current share price is used to support rating judgments such as Buy, Hold, and Reduce.

  The report states that HSBC ratings are usually based on the gap between the target price and the current share price, with the expectation that the market price will reflect the target price within 6 to 12 months; in this report, MiniMax is rated Hold.

Asset mapping & comparison

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

• MiniMax (0100.HK)

  Core covered name

  Strengths

  Strong inference efficiency, a long-term gross margin target for the open platform, a self-developed computing cluster, and a cross-disciplinary R&D organization; M3 Pro and multimodal products provide visible product iteration catalysts.

  Weaknesses

  The report does not provide a full valuation breakdown or detailed financial forecasts, and the company remains in a highly competitive AI environment.

  Comparison

  Management emphasized reducing dependence on cloud service providers through self-developed clusters and improving relative cost efficiency amid rising chip and memory prices.

  Risks

  Intensifying AI model competition, rising computing power and memory costs, product launches falling short of expectations, and the open platform gross margin target being achieved more slowly than expected.

Key data

• Management meeting date2026-07-23HSBC met with MiniMax's CEO, COO, and capital markets team in Shanghai.
• Current priceHKD199.30The pricing date is 2026-07-23.
• Target priceHKD760.00The report discloses a target price of HKD760.00 for MiniMax.
• RatingHoldThe report's rating on MiniMax is Hold.
• M3 Pro launch timingSeptember-October 2026Management said M3 Pro is on track for launch.
• M3 Pro activated parametersc60bnCorresponding to a c3trn parameter scale and a 2% activation ratio.
• Training data efficiencyThe intelligence level previously achieved by M3 with 40trn data is now reached with 8trn dataReflects improved training efficiency.
• Scale of self-developed computing cluster10k+ chip clusterIt can currently run model training tasks and continues to expand.

Impact & implications

From an investment perspective, the meeting takeaways reinforce MiniMax's medium- to long-term competitive narrative in model efficiency, open platform commercialization, and AI infrastructure self-sufficiency. If M3 Pro, multimodal products, and self-built clusters continue to deliver, they could improve inference costs, gross margins, and product competitiveness; however, the report still maintains Hold, indicating that these positive factors still need to be weighed together with valuation, competitive dynamics, and execution risks.

Risks

• Competition in the AI industry is intense, and model capabilities and cost efficiency need continuous iteration.
• Rising chip and memory prices may affect training and inference costs.
• There is uncertainty around the launch timing and commercialization performance of products such as M3 Pro and Hailuo 3.
• Expansion of in-house AI infrastructure requires continued investment, which may put pressure on expenses and cash flow.
• The report notes that full valuation and risk details should be referenced in the previous report dated June 2, 2026; this meeting note does not elaborate on all risks.

What to watch

• Whether M3 Pro launches on schedule in September-October 2026.
• The actual performance of M3 Pro in token throughput, training efficiency, and KV cache optimization.
• The launch timing of Hailuo 3 and the commercialization progress of multimodal products.
• The pace at which the open platform's long-term high double-digit gross margin target is achieved.
• The actual impact of 10k+ chip cluster expansion, utilization, and reduced dependence on cloud service providers.