MiniMax (00100) Report Interpretation

Overview

This conference-takeaways report assesses MiniMax's model pipeline, commercialisation progress, API economics and computing strategy following HSBC's 2026 China Conference and China internet and AI tour. HSBC remains constructive on product and margin developments but retains Hold because its unchanged HKD330.00 target price is below the then-current share price.

Core views

HSBC identifies the near-term model pipeline as the central operating catalyst. MiniMax expects to launch M3.1 in the coming weeks, followed by M3 Pro from late September to early October; H3.1 has no specified timeline. M3.1 uses the same base-model parameter size and architecture as M3 but is post-trained to improve performance and cost efficiency. M3 Pro could reach 2.7-2.8 trillion parameters and target frontier capability. The report says its upgraded MiniMax Sparse Attention mechanism, compute efficiency improvements and KV-cache compression are intended to expand model scale materially from M3's 428 billion parameters without a commensurate increase in inference cost. On multimodal competition, the report argues that MiniMax is differentiated in real-time text rendering and stylised video generation, supporting e-commerce and livestreaming applications. It contrasts this positioning with Seedance's strength in short drama and AI comics. HSBC also points to H3's open-weight strategy as a commercialisation lever: community feedback has been positive, and Fal.ai's H3 Max derivative is available on MiniMax Design. MiniMax can receive a share of revenue from such deployments, and the report states that H3 ARR grew solidly after launch, helping total ARR. The report views API unit economics as healthy, with further upside potential. ARR excludes non-recurring and one-off items and is described as reasonably representative of revenue; Token Plan accounted for a low-teens percentage of both August ARR and 1H26 revenue. Coding and agentic applications make up most use cases, while the customer base remains diversified. Consumer gross margin is positive, although Token Plan's gross margin is negative because of token subsidies at M3's initial launch. Multimodal API gross margin exceeds 50%, partly because its smaller parameter size permits cheaper chips. For text APIs, M2 generated c.40% gross margin, while M3 initially lagged but has surpassed that level after token-throughput optimisation. HSBC distinguishes MiniMax's distribution models. First-party API sales account for most ARR because they offer service quality, inference efficiency and faster model updates. Customers such as cloud service providers may purchase APIs directly and pay by volume, sometimes with bulk discounts; MiniMax records this as gross revenue. Alternatively, cloud providers and users such as Fal.ai can deploy open-weight versions, under which MiniMax records net revenue based on licence terms and can earn revenue sharing of up to c.30%, described as capped by industry practice. This open-weight route has much higher gross margin than first-party APIs. To meet computing needs, MiniMax is self-building clusters, securing capacity from cloud service providers and supplementing inference capacity with discounted, flexible token-factory procurement. HSBC maintains Hold and an unchanged HKD330.00 target price based on a 10-year DCF. The target compares with a HKD342.60 share price at the 02 September 2026 close, implying -3.7% downside. The valuation uses an 11.9% WACC, comprising a 4.25% risk-free rate, 4.75% market-risk premium, 1.71 beta and 3.9% debt-to-total-capital ratio, plus a 3% terminal growth rate. The report's financial forecasts show revenue rising from USD79m in 2025A to USD564m in 2026E, USD1,667m in 2027E and USD2,435m in 2028E, while HSBC net profit remains negative at USD666m, USD473m and USD373m for 2026E-2028E respectively. It therefore frames execution on model launches, monetisation and progress toward breakeven as key determinants of the investment case.

Analysis framework

HSBC combines conference and company-tour takeaways with product-roadmap analysis, competitive positioning, ARR and API gross-margin discussion, distribution-economics analysis and forecast financials. It values MiniMax using a 10-year DCF and assesses upside and downside catalysts against the target price.

Methodology notes

• Valuation methodsDCF (Discounted Cash Flow)

  10-year discounted cash flow valuation

  HSBC derives its HKD330.00 target price from a 10-year DCF using an 11.9% WACC and 3% terminal growth rate.

• Industry AnalysisVolume-price decomposition

  API monetisation and gross-margin analysis by product and distribution route

  The report separates first-party API revenue, open-weight revenue sharing, token subsidies and model-specific gross margins to explain MiniMax's monetisation and profitability path.

Asset mapping & comparison

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

• MiniMax (00100.HK)

  Primary covered company; AI-model launches, API monetisation and compute availability drive the report's outlook.

  Strengths

  M3 pipeline, improving text-model API margin, multimodal API gross margin above 50%, differentiated video capabilities and H3 commercialisation.

  Weaknesses

  Negative Token Plan gross margin at launch, continuing losses, cash burn and need to reach breakeven.

  Comparison

  MiniMax is described as stronger in real-time text rendering and stylised video, while Seedance is highlighted for short drama and AI comics.

  Risks

  Competitor model launches, weaker model performance, computing bottlenecks, slower breakeven, refinancing needs, lock-up expiry, and geopolitical, IP and content-generation risks.

Key data

• Target priceHKD330.00Unchanged; based on HSBC's 10-year DCF.
• Share priceHKD342.60As of the close on 02 Sep 2026.
• Implied upside/downside-3.7%Difference between the HKD330.00 target price and share price.
• M3 Pro parameter size2.7-2.8trnPotential size; launch expected from late September to early October.
• M3 parameter size428bnReference point for the planned increase in M3 Pro scale.
• Multimodal API gross margin50%+Partly supported by smaller models using cheaper chips.
• M2 text-model API gross marginc.40%M3 has exceeded this level after token-throughput optimisation.
• 2026E revenueUSD564mHSBC forecast, up from USD79m in 2025A.
• 2028E revenueUSD2,435mHSBC forecast.

Impact & implications

HSBC sees the next model launches, H3 monetisation and improving API gross margins as potential supports for MiniMax's revenue trajectory and eventual path to breakeven. However, the report's valuation remains constrained by an unchanged target price below the market price and by material competitive, operational and financing risks.

Risks

• New competitor-model launches could dilute MiniMax's token-usage growth.
• Other AI-company IPOs could reduce MiniMax's scarcity premium.
• A greater computing bottleneck could limit top-line growth upside.
• M3 Pro or other model iterations could perform below expectations.
• Progress toward breakeven could be slower than expected.
• Heavy cash burn and refinancing needs could pressure the company.
• The lock-up period ends in January 2027.
• Geopolitical, intellectual-property and content-generation risks remain.

What to watch

• Launch timing and performance of M3.1 in coming weeks.
• M3 Pro launch from late September to early October and its ability to deliver frontier capability at manageable inference cost.
• H3 ARR growth and further commercialisation through open-weight deployments.
• Text-model API margin after M3 token-throughput optimisation.
• The pace of MiniMax's progress toward breakeven and its computing-capacity availability.