MiniMax has started to generate revenue through external partners.

MiniMax's first half-year financial report after going public, the most interesting figure is by no means its revenue.

Its revenue in the first half of the year reached 117 million U.S. dollars, which has exceeded the full-year revenue of 2025. However, its R&D investment is 2.5 times its revenue.

What is more notable is its revenue structure. Revenue from the open platform and other AI enterprise services hit 73.93 million U.S. dollars, accounting for 63.4% of the total revenue; revenue from AI-native products reached 42.64 million U.S. dollars, with the proportion dropping to 36.6%.

This large model company, which was first known to the public for its C-end products, now has a completely different main source of revenue.

MiniMax is embarking on a brand new development path.

1. Hailuo Steps Back

MiniMax was first recognized by users for Hailuo AI. With text, speech, and video generation capabilities, plus a series of AI-native products, MiniMax has built a very clear C-end brand image. The business feedback of this type of operation is also very straightforward: As long as users are willing to pay, the product is valuable.

In this financial report, the revenue of AI-native products still increased by 100.9% to 42.64 million U.S. dollars. It just no longer occupies the dominant position.

In the same period last year, revenue from AI-native products accounted for 69.7% of MiniMax's total revenue. This year, the proportion is only 36.6%.

On the other side, revenue from the open platform and other AI enterprise services surged 703.1% year on year, rising from 9.2 million U.S. dollars to 73.93 million U.S. dollars.

AI-native products: 42.64 million U.S. dollars (+100.9%), revenue proportion shifted from 69.7% to 36.6%

Open platform and enterprise services: 73.93 million U.S. dollars (+703.1%), revenue proportion shifted from 30.3% to 63.4%

With these two curves moving in opposite directions and crossing each other, MiniMax's business focus is no longer on Hailuo.

Users are still paying for products, while enterprises and developers are starting to pay for model capabilities. These are two completely different types of business, and the latter obviously has a much higher ceiling.

2. B-side Takes Over

Within one year, the proportion of revenue from B-side and the open platform rose from 30.3% to 63.4%. The significance of this change is more worth discussing than the 703% growth rate.

C-end users purchase a product. Enterprises purchase a set of capabilities. Developers call the model. Agents call for Tokens.

When the model is embedded in customer service, coding, marketing, knowledge base, search and Agent workflows, it is no longer "an AI product", but becomes a pipeline in the business process. At this point, what customers buy can hardly be defined by the term "software".

MiniMax has begun to gain something it rarely owned in the past: Other parties use its model to create value. This means its revenue no longer depends on how good it is at making products, but on how much other parties need its capabilities.

This is probably the most important change in this financial report.

Hailuo requires users to come to MiniMax. The open platform allows MiniMax to enter other parties' products. It shifts from "drawing users in" to "integrating itself out".

With this step, MiniMax's identity begins to change.

3. The Model Starts to Generate Profits

Hailuo sells products. APIs sell capabilities. Enterprise services sell the results generated after the model is integrated into business operations.

The difference between these three layers lies in: Products are one-time transactions, capabilities are continuous supply, and results follow the logic of profit sharing.

MiniMax is shifting from developing applications on its own to allowing others to build applications with its models.

This is also the truly noteworthy point behind the rapid growth of open platform revenue.

The increase in API calls means that the model has begun to become a means of production for others. The growth of enterprise clients means that the model has begun to enter real workflows. The rise of Agentic workloads means that the model has evolved from answering questions to completing tasks.

Once the model is embedded in other parties' business processes, its commercial value will no longer be calculated by "number of openings", but priced by "degree of dependency".

How many times users open Hailuo every day matters. How many Tokens enterprises call every day also matters. But the most critical thing is how much business developers build on top of your model.

The upper limit of a model company depends on how many people are willing to build their own business on top of your capabilities.

The position MiniMax is striving for is to move down from the application layer and become the underlying infrastructure of other parties' businesses.

4. Capital Is Still Not Sufficient

Commercialization is advancing, but profitability has not yet been achieved.

Revenue: 117 million U.S. dollars (exceeding the full-year revenue of 2025)

R&D investment: 297 million U.S. dollars (about 2.55 times the revenue)

Adjusted net loss: 293 million U.S. dollars (+111.2% year on year)

Seeing this set of figures, people may easily come up with a question: It has made quite a lot of sales, why is it still burning cash?

The answer lies in the cost structure of large models. After traditional internet products reach a certain scale, the marginal cost of adding a new user can drop rapidly. But it is not that easy for large models. Model training costs money, reasoning costs money. The more users call, the more computing power is consumed; the more enterprise clients there are, the higher the infrastructure cost will rise accordingly.

Therefore, the revenue growth of large model companies will never automatically turn into profit growth. While commercialization is advancing rapidly, costs are also rising fast.

What MiniMax is really facing now is no longer "whether there are clients". Instead, it is another more difficult problem:

Whether the growth of clients can eventually outpace the cost of the model. The former is a sales problem, while the latter is a survival problem.

5. Gross Margin Takes the Lead

The good news also lies here.

In the first half of the year, MiniMax's sales cost increased by 258.1%, while its revenue increased by 283.1%. Revenue is growing faster than cost. Gross profit rose from 3.69 million U.S. dollars to 20.81 million U.S. dollars, and gross margin increased from 12.1% to 17.9%.

The figures are still not high, but the trend has begun to change.

This means MiniMax has begun to enter the second stage of large model commercialization: From proving market demand to pursuing economies of scale.

Strong model capabilities can only solve the first problem. The next problem is whether the same one-dollar revenue can be generated at an increasingly lower cost. The first problem tests technology, while the second tests engineering, system and operation capabilities.

Behind this, the competition is no longer only about model parameters. Training efficiency, reasoning efficiency, model routing, chip adaptation, system scheduling, caching, compression, and Token usage efficiency at the product level will all be counted into the cost.

MiniMax mentioned in its financial report that it will improve training and reasoning efficiency through full-stack collaborative design of models, infrastructure, systems and products.

This direction is very critical. Because the real cost competition in the large model industry is shifting from who has a stronger model to who can deliver the same intelligence at a lower cost.

Model leaderboards determine who can grab attention. The cost curve determines who can survive longer.

6. Global Revenue

MiniMax has another distinctive feature: its commercialization has not been limited to the Chinese market from the very beginning.

In the first half of the year, revenue from regions outside mainland China reached 70.83 million U.S. dollars, accounting for 60.8% of the total revenue. Revenue from mainland China was 45.75 million U.S. dollars, accounting for 39.2%. At present, MiniMax's products and services have covered more than 230 countries and regions.

This revenue structure is very distinctive among Chinese large model companies.

AI products are naturally capable of global distribution. One model API can serve developers from different countries, and one AI-native product can directly reach overseas users.

When traditional internet companies go global, they need channels, localization, supply chains and a large number of offline infrastructures. The globalization path for AI companies is much shorter: models are developed in China, products are distributed globally, and APIs are directly connected to overseas developer systems. The business can cross borders almost without spending extra money on long-distance trips.

MiniMax has already taken this path. Of course, global revenue also brings new complexities such as exchange rate, compliance, payment and local market operation.

But at least from the perspective of revenue structure, it already has a rare feature: Develop in China, monetize globally.

If this path continues to succeed, its business ceiling will be further lifted.

7. After the Hundred-Billion Market Cap

The capital market has placed very high expectations on MiniMax.

After going public, the company's market value once surged, and now it has fallen back to around 100 billion Hong Kong dollars. The market is not pessimistic about it, but after learning about its performance, the market is looking for more new growth stories.

This means that simply talking about "the huge potential of the AI industry" can hardly support a new valuation story anymore.

Revenue has already proven part of its commercialization capabilities. The fact that B-side revenue accounts for more than 60% of total revenue also proves that enterprises and developers are willing to use MiniMax's models. Listing and subsequent financing have also significantly improved the company's cash reserve.

The only remaining core question is: Can this business make a profit?

This question is far more difficult than expanding revenue. Because a stronger model does not necessarily mean higher profits; more clients do not necessarily mean better cash flow. What really determines the sustainability of this business is whether the gap between revenue growth and cost growth can be gradually widened.

The 17.9% gross margin is much better than the previous 12.1%, but there is still a long way to go to reach the profitability structure of mature software businesses.

MiniMax is now in a very special position: its commercialization curve is advancing rapidly, while the profitability curve is still at the origin. The gap between these two lines is the problem it needs to solve next.

8. The Last Stage Remains

Putting MiniMax into the context of China's large model industry, the change will be clearer.

In the first stage, everyone competed on models. Model parameters, Benchmark, reasoning capabilities, and release speed became the focus of competition. In the second stage, everyone competed on applications. Whoever can turn the model into a product that real users use will get the first batch of real users.

MiniMax has completed these two stages. Hailuo proved its C-end product capability, and the open platform proved the demand from enterprises and developers.

Next, the core points that large model companies compete for begin to change: clients, channels, Tokens, reasoning costs, business models, and globalization.

Model capability is no longer the end point, but the starting point.

A sufficiently strong model can only get you the admission ticket; a product that is widely used can get you the second card; the willingness of enterprises and developers to pay can get you the third card. The last card is profitability.

MiniMax has got the first three cards. The last one is still on the way.

9. Where Has MiniMax Arrived?

Looking back at this financial report, three figures are most worth putting together.

117 million U.S. dollars in revenue: The model capability has begun to form large-scale commercial demand

63.4% B-side and open platform revenue proportion: Shifting from an AI product company to a model capability company

293 million U.S. dollars in adjusted net loss: Reminds everyone that commercialization is far from over

These three figures together represent the current MiniMax.

It has gone past the stage of "whether people will use its products", and is now passing through the stage of "whether people are willing to pay for the model". The next question to answer is very specific: Can the money others pay for it eventually support its own operation?

This may also be the signal that China's large model industry truly enters the next stage. Revenue can keep rising, the number of clients can keep growing, but in the end everything will return to the most basic income statement:

For every additional dollar of revenue, how much profit can be retained in the end?

Words Beyond the Layout:

The hardest step for a large model company is never to develop the model. The real test is how to turn intelligence itself into a sustainable business.

MiniMax has already proved that there are people willing to pay for its products and services.

The next step is to prove that it can support its own operation and make profits.

The gap between "people are willing to buy" and "the business can support itself" covers the entire deep water zone of large model commercialization.

This article is from the WeChat Official Account "Beyond the Layout", written by Ban Jun, and published with authorization from 36Kr.