MiniMax Hits $300M ARR as Self-Evolving M2.7 Model Captures Global Demand

Analysts project the company’s reported revenue will hit $219 million in 2026, up from $79 million in 2025 . Longer-term consensus estimates from Visible Alpha show a ramp to $5.8 billion by 2030, fueled by its AI-native portfolio: Hailuo AI (video generation), Talkie (chatbot), the core MiniMax platform, and MiniMax Audio .

Despite the top-line growth, MiniMax continues to report widening total losses—a common pattern for compute-intensive AI firms pouring capital into infrastructure and R&D ahead of profitability .

M2.7: a self-evolving flagship that changed the unit economics

MiniMax released M2.7 on March 18, 2026, positioning it as the first model to actively participate in its own development cycle . The model generates its own evaluation data, identifies capability gaps, and produces synthetic training examples to close them . This recursive self-improvement approach represents one of the earliest production-scale demonstrations of AI-assisted AI training .

Architecture and cost

M2.7 uses a sparse mixture-of-experts architecture with 230 billion total parameters, but activates only ~10 billion per token . This efficiency lets it achieve tier-1 benchmark performance competitive with Claude Opus 4.6 and GPT-5 while costing approximately 1/50th of mainstream flagship models . A "highspeed" variant launched simultaneously boosts output speed by 66% to 100 tokens per second .

At release, M2.7 ranked first on the Artificial Analysis Intelligence Index, topping all 136 evaluated models with a composite score of 50 .

International revenue dominance

Over 70% of MiniMax's revenue now comes from outside China . The international skew is partly structural: M2.7 was designed to power third-party agent harnesses such as Claude Code, Kilo Code, and OpenClaw, and gained rapid traction among developers in North America and Europe .

Stock surge and Hang Seng Tech Index inclusion

MiniMax debuted on the Hong Kong Stock Exchange on January 9, 2026, raising HK$4.8 billion ($619 million) at HK$165 per share . Retail investors subscribed to more than 1,830 times the shares available, and the stock closed its first day at HK$345—up 109% . Since then, shares have surged nearly 500% .

On May 22, 2026, the Hang Seng Indexes Company announced that MiniMax would be added to the Hang Seng Tech Index and Hang Seng Composite Index, effective after market close on June 5, 2026 . Inclusion is expected to unlock Stock Connect access and could attract over HK$100 billion in southbound capital inflows . Passive fund flows linked to the rebalance are estimated at $1.25–1.75 billion for MiniMax and Zhipu combined .

M3 and what comes next

The next-generation M3 large language model is planned for the second half of 2026. MiniMax has disclosed that M3 will incorporate a custom sparse attention mechanism to substantially accelerate both prefilling and decoding at one‑million‑token context lengths . Morgan Stanley has flagged the M3 launch as a key catalyst and responded by raising its target price for MiniMax to HK$990 . Whether M3 can maintain M2.7's cost-performance advantage while handling significantly longer contexts will be the critical technical question for the second half of 2026.

MiniMax’s current position reflects a model strategy that pairs aggressive efficiency with self-improving training, a business model tilted decisively toward international developers, and a stock market reception that few Chinese AI companies have matched. The company’s next test is turning revenue acceleration into sustained profitability, without sacrificing the technical momentum that M2.7 has created.