Morgan Stanley evaluates MiniMax as a 'scarce asset of globally leading foundational models,' with the core logic of its high valuation being 'technology determines the ceiling, globalization determines the valuation.'

Morgan Stanley initiated coverage of MiniMax (00100.HK) with an Overweight rating and a HKD 930 target (base case), arguing the company is a global leader in AI foundational models and assigning a 54x 2027 P/S on a USD 700m revenue forecast. MS projects revenue rising from USD 75m in 2025 to USD 700m in 2027 (9–10x), gross margin expanding from 12% (2024) to 32% (2027), but widening non‑IFRS operating losses (~USD 484m in 2027) and average monthly cash burn of ~USD 27.9m in 2025. Key drivers are top‑tier model performance (M2.5 usage dominance, 1.97 trillion tokens) and rapid globalization (73% overseas revenue YTD 9M2025; APAC 61%/Americas 24%/EMEA 15%), while main risks include GPU supply/geopolitical restrictions, competition from hyperscalers, pricing pressure from commoditization, and dependence on a mid‑2026 next‑generation model to trigger stepwise revenue growth.

Analysis

Market structure: If MiniMax (00100.HK) sustains top-tier SOTA performance and >75% inference efficiency, winners are MiniMax, API-centric cloud partners, and GPU makers (NVDA) as demand for inference hardware and cloud capacity rises; losers include incumbents selling regional-locked models and low-efficiency commoditized LLM providers whose gross margins compress. Expect pricing power for high-efficiency APIs to allow gross margins to move from ~12% (2024) toward MS’s 32% (2027) projection if enterprise API share rises to ~40% by 2027, shifting revenue mix overseas (current ~73% through 9M25). Cross-asset: stronger MiniMax narrative supports EM/China equity risk-on, USD demand for GPU imports could pressure CNH; higher growth expectations lift HY spreads in tech if financing risk perceived low, while GPU scarcity/controls increase implied vol on NVDA options.

Risk assessment: Tail risks include US/EU export controls that cut advanced GPU access (days-weeks), a failed mid-2026 model launch (binary event within 0-3 months of release), or cash-runway shock (monthly burn ~$28M implies runway strain if liquidity <6–9 months). Short-term catalysts (weeks–months) are model release benchmarks and incremental enterprise API contracts; long-term (quarters–years) risks are commoditization and competition from hyperscalers requiring >$1B+ capex. Hidden dependencies: MoE/linear-attention efficiency assumes uninterrupted access to high-end GPUs and talent — a geopolitical or hiring setback magnifies dilution/cash burn.

Key Decisions for Investors

• Establish a 2–3% portfolio long position in MiniMax (00100.HK) via a mix of stock (1–1.5%) and Sep–Dec 2026 call spreads (remaining allocation, 20–30% OTM) to target the mid-2026 next-gen model catalyst while capping premium spend.
• Implement a hedge by shorting 1–1.5% notional of KWEB (US ETF) to isolate MiniMax-specific upside from China-internet beta; increase hedge to 3% if MiniMax's overseas revenue share falls below 60% or monthly cash burn exceeds $30M for two consecutive months.
• Buy NVDA 2026/2027 LEAP calls (0.5–1% portfolio) or a call spread to capture GPU tailwinds; reduce position if NVDA implied vol rises above 80% or if major export-control headlines emerge.
• Set hard stop/cut triggers: reduce MiniMax equity by 50% if post-release benchmarks (within 30 days) show model ranks fall outside global top-10 or if disclosed cash runway <6 months; take profits if revenue guidance moves toward or above MS’s optimistic USD1B 2027 path.