In the first four months, China's software industry revenue reached 4,668.6 billion yuan, a year-on-year increase of 10.9%

China's software industry delivered a substantial yet structurally nuanced performance in the first four months of 2026. Business revenue of 4.67 trillion yuan and a year-on-year growth rate of 10.9% outline the contours of a continuously expanding industry. However, two parallel data points point in diverging directions: total profit grew by merely 2.2%, significantly lagging behind revenue growth; while exports surged by 13.0%, revealing a strong pull from external markets. This set of figures is not simply a mixed bag; it reveals a deeper reality: China's software industry is undergoing a pivotal transformation from scale expansion to value deepening, with technological breakthroughs and business model innovation becoming the core tools to pierce through the fog of "revenue growth without profit growth."

The sluggish profit growth is a typical characteristic of an industry entering its mid-stage. When market development and basic penetration reach a certain scale, the marginal benefits gained solely from increasing user or project numbers inevitably diminish. Intensifying competition, rising labor costs, and compressed gross margins on traditional businesses collectively erode profit margins. This is not a signal of decline, but the growing pains of a model switch. The industry's gaze must shift from "how big we've grown" to "how refined our work is" and "how high our efficiency is." It is precisely against this backdrop that innovation at the technological level appears especially urgent and valuable.

The technical roadmap disclosed by the Xiaomi MiMo-V2.5 large model team provides exactly such a micro yet profound case study. Their five core breakthroughs—dual-pool KVCache optimization, distributed caching, scheduling strategies, and inference acceleration—are fundamentally an "efficiency revolution." These optimizations directly target the core pain point of large model services: the exorbitant cost of computing power. Through extreme optimization at the algorithm and system engineering levels, they achieve a permanent price reduction for APIs while maintaining a break-even balance. This is not only proof of technical prowess but also a successful validation of a business model: in the field of AI infrastructure services, technological innovation can directly translate into sustainable competitive advantages and commercial viability. It breaks the惯性思维 that "price cuts inevitably lead to losses," demonstrating the possibility of absorbing cost pressures through technological dividends.

Even more noteworthy is the logic behind Xiaomi's "Trillion-Token Creator Incentive Program." Over 540,000 applicants and the issuance of over 65 million yuan worth of free tokens represent far more than a marketing campaign. This is essentially cultivating an ecosystem and locking in developers. When a technological platform lowers the barrier to use through price reductions, while simultaneously guiding developers to build applications on it through incentive programs, it evolves from a single "model service provider" to an ecosystem core that fosters a positive cycle of "application prosperity - data feedback - model optimization." This strategy aligns with the early logic of cloud computing, which subsidized to capture the developer ecosystem, and is now reappearing in a new form in the era of AI large models.

Combining macro-level data with micro-level practices, a clear path emerges. The macro phenomenon of "revenue growth without profit growth" compels the industry to seek new value growth points. Enterprises like Xiaomi, through hardcore technology that reduces costs and enhances efficiency, and by actively building an ecosystem, are precisely providing an active response to this demand. The robust 13% growth in software exports also suggests that Chinese software companies (including AI services) are leveraging cost-performance advantages and localization to capture new increments in international markets. Future competition will no longer be at the level of simple applications, but a deep contest in underlying algorithms, system efficiency, and ecosystem-building capabilities.

Therefore, what this industry report depicts is far from a static snapshot, but rather a dynamic transformation landscape. The foundation of scale remains solid, but the source of growth is shifting. Transitioning from "extensive growth" dependent on market dividends to "lean growth" reliant on core technological breakthroughs and ecosystem synergy will be the main thread of China's software and AI industry development for the coming years. Companies that, like the Xiaomi MiMo team, can effectively translate technological breakthroughs into commercial sustainability and ecosystem appeal will likely be the ones leading the charge on this new main track.