Morgan Stanley has initiated coverage on MiniMax with an "Overweight" rating and a target price of HKD 930, positioning the company as a "global leader in AI foundational models." The report indicates that the investment thesis is not based on short-term profitability but hinges on two core aspects: whether the company's model capabilities rank among the global top tier, and if its revenue structure possesses the flexibility for international expansion.

According to the covering analyst, Gary Yu of Morgan Stanley, MiniMax has entered the global State-of-the-Art (SOTA) model cohort, boasting comprehensive multimodal capabilities and a highly scalable commercialization path. Based on this, the firm projects the company's revenue could surge from USD 75 million in 2025 to USD 700 million in 2027, representing a nine to tenfold increase within two years. Once a generational technological advantage is established, the revenue trajectory is expected to exhibit a "step-function" leap.

The report offers a straightforward explanation for the high valuation: this is an asset where "technology determines the revenue ceiling and globalization dictates the valuation framework." If model performance consistently ranks among the global elite, its potential is constrained only by the global Total Addressable Market (TAM). Furthermore, a revenue structure predominantly derived from overseas markets would naturally align its valuation anchor closer to international peers rather than domestic comparables.

Technology capability forms the starting point for valuation. Morgan Stanley identifies three core competitive strengths for MiniMax: sustained iterative capability, multimodal strategy, and cost efficiency.

In independent benchmark tests, the MiniMax-M2 model initially ranked fifth globally upon release; the latest flagship model, MiniMax-M2.5, holds the sixth position overall and ranks fourth among open-source models. As of mid-February 2026, M2.5 ranked first on OpenRouter by token usage volume, reaching 1.97 trillion tokens, and commanded a 58.8% market share in programming scenarios.

These metrics suggest the model is being utilized in high-frequency, real-world applications, moving beyond mere laboratory benchmarks.

More critically, the cost structure is highlighted. The company employs a Mixture of Experts (MoE) architecture and Linear Attention mechanisms, achieving a Model Flop Utilization exceeding 75% during inference, significantly higher than the industry average of 40%-50%. Inference efficiency directly impacts API pricing tiers and gross margin flexibility, determining whether profitability can improve alongside scale expansion.

Morgan Stanley forecasts the company's gross margin will rise from 12% in 2024 to 32% by 2027. However, operating losses are expected to widen concurrently, with a projected non-IFRS operating loss of approximately USD 484 million in 2027. This reflects a strategy focused on "expanding technology and scale first, with profitability following," rather than anticipating a near-term inflection point.

Technological leadership does not guarantee profitability but does dictate the potential revenue ceiling.

Revenue structure determines the growth trajectory. MiniMax's business model is not driven by a single product but operates on three parallel fronts:

* Consumer (2C): Agent and companion products like Talkie/Xingye. * Prosumer/Partner (2P): Products such as Hailuo AI and MiniMax Audio. * Business (2B): Open Platform API.

As of the first nine months of 2025, the company's Monthly Active Users (MAU) grew to 27.6 million from 3.1 million in 2023, with 1.77 million paying users. The revenue structure is becoming more diversified, with the share from the Open Platform continuously increasing.

Morgan Stanley projects the Open Platform's revenue contribution will increase from 29% in 2024 to 40% by 2027, with a three-year compound annual growth rate exceeding 200%. Following a breakthrough in model capability, enterprise API demand is more likely to experience a "jump" in volume.

The report emphasizes an industry characteristic: growth for foundational model companies is often triggered by key generational model releases, rather than following a smooth, gradual incline. Examples include OpenAI's ChatGPT 3.5 and Anthropic's Claude 3.5 Sonnet, both of which spurred significant revenue jumps post-upgrade.

Whether MiniMax replicates this pattern depends on the success of its next-generation model, slated for release in mid-2026.

Globalization is a prerequisite for valuation. Morgan Stanley particularly emphasizes MiniMax's "Born Global" approach.

The proportion of overseas revenue has increased from 19% in 2023 to 73% in the first nine months of 2025. The regional distribution is: Asia-Pacific 61%, Americas 24%, and EMEA 15%.

Against the backdrop of a global foundational model market projected to grow from USD 10.7 billion in 2024 to USD 206.5 billion by 2029 (a CAGR of 80.7%), the company's current global market share is only about 0.3%. Even a modest increase in market share would result in significant revenue elasticity.

More importantly is the valuation framework. If revenue is primarily derived from overseas markets and customers are primarily API and subscription-based, the valuation logic aligns more closely with international AI peers than with traditional Chinese software companies.

This reasoning underpins Morgan Stanley's core valuation basis of 54 times projected 2027 Price-to-Sales (P/S).

Valuation divergence centers on the "next-generation model." The scenario analysis is clearly defined:

* Base Case: 2027 revenue of USD 700 million, corresponding to a 54x P/S multiple and a target price of HKD 930. * Bull Case: 2027 revenue of USD 1 billion, target price of HKD 1240. * Bear Case: 2027 revenue of USD 400 million, target price of HKD 300.

The sole variable determining the valuation differential is whether the next-generation model, launching in mid-2026, meets or surpasses global SOTA standards.

Risks are similarly focused: GPU supply and geopolitical constraints, the resource gap compared to OpenAI and other hyperscalers, pricing pressure from model commoditization, and ongoing cash burn.

This is a valuation based on "ability to deliver on technology." Morgan Stanley does not avoid the reality: currently, no pure-play AI foundational model company has achieved stable profitability. MiniMax is projected to have an average monthly cash burn of approximately USD 27.9 million in 2025, with limited near-term profit visibility.

However, the report's core conclusion is that competition in the foundational model industry is not about marketing prowess but about generational breakthroughs. Technological capability determines the revenue ceiling, while the global market dictates the valuation anchor.

If model upgrades lead to non-linear revenue expansion, the current valuation merely represents a discounting of future scale. Conversely, if the model fails to maintain a position in the global top tier, valuation contraction could be equally rapid.

This investment represents a bet on the pace of technological execution. Morgan Stanley has chosen to side with what it terms a "scarce global foundational model asset."