On December 29, Greenwoods Asset Management, a Chinese multi-billion-dollar private fund, issued its annual letter to investors.

Greenwoods fund manager Gao Yuncheng stated in the letter that "2026 is likely to mark the true beginning of widespread AI Agent adoption."

Gao Yuncheng pointed out in the letter that AI large language models utilizing hardware like smartphones as underlying operating platforms could potentially challenge existing iOS and Android systems. Many traditional competitive advantages might be overturned during this technological wave; companies without an AI strategy risk being marginalized. This is an area requiring vigilance!

Key takeaways from the letter are summarized below for readers.

Core Views: 1. The fact that 50% of AI researchers are ethnic Chinese actually affirms the strength of China's basic education system; China possesses competitive and cost-effective advantages in the generative artificial intelligence race. 2. Against this macroeconomic backdrop, we observe international capital flowing back into Chinese assets, a process that has only just begun. 3. As investors, we should not hastily draw conclusions about AI, but we must continuously scrutinize the cash flow situations of companies in our portfolio and recalculate valuations using stricter accounting assumptions. 4. Companies in our core holdings are valued within reasonable historical ranges, boast business models with pricing power, operate in favorable competitive landscapes, and are led by sufficiently rational and clear-minded management teams. 5. Our core holdings remain largely stable, showing little change from the portfolio shared mid-year. In summary: two systems, each with its stars. 6. Only companies with exceptionally strong business models and unique competitive positions possess the ability to control their profit levels. Such companies are extremely rare.

China's Competitiveness in Generative AI The emergence of DeepSeek in 2025 made the world recognize that China holds competitive and cost-effective advantages in the generative AI race.

It's not just DeepSeek; the global adoption of Qwen among open-source models demonstrates that even without access to the most advanced GPUs, domestic models can leverage available resources to pressure leaders and provide the world with a non-US alternative.

The fact that 50% of AI researchers are ethnic Chinese affirms the strength of China's basic education system. The vast pool of STEM talent genuinely supports continuous upgrades from basic industries to aerospace and military technology.

The iterative development of generations of engineers behind China's industrial growth fully illustrates the historic progression from learning advanced Western manufacturing to participating in complex global supply chains, and ultimately achieving R&D and manufacturing superiority in certain sectors.

It is this strong talent reservoir and the emergence of globally competitive leading companies within developed industrial chains that have endowed China with sufficient strength.

International Capital Returning to Chinese Assets Against this macroeconomic backdrop, we observe international capital flowing back into Chinese assets, a process that has only just begun.

In cutting-edge fields of new economic development like AI, new energy, intelligent driving, and humanoid robots, both China and the US are actively investing. Related companies are growing rapidly within this major wave, with their revenues and market capitalizations continually setting new records, presenting a future as vast as the stars and sea.

Cautious View on Overly Optimistic Expectations However, alongside rapid advancement, increasingly skeptical voices are emerging.

Discussions about whether future cash flows can support trillions in AI-related infrastructure investments intensified towards the end of 2025. The on and off-balance-sheet capital expenditures of large tech companies have grown so substantial that many doubt the sustainability of this cycle.

As investors, we should not readily draw conclusions. But based on such risk warnings, continuously scrutinizing the cash flow of portfolio companies and recalculating valuations with stricter accounting assumptions is necessary.

Certainly, some companies' market capitalizations are built on overly optimistic future expectations, but these are not the companies in our portfolio.

Companies in our core holdings are valued within reasonable historical ranges, possess business models with pricing power, operate in favorable competitive landscapes, and are led by sufficiently rational and clear-minded management teams.

Future Giants = Nascent AI-Native Companies A regret this year was not actively participating in the storage and information transmission segments of the AI industry chain. The reason for missing the former was an excessive focus on cyclicality while overlooking structural demand changes and the difficulty of adding new capacity; for the latter, it was concern over excessively high profit margins and potential future changes in technology路线.

While remaining vigilant about potential bubbles or overinvestment, we also clearly see that AI's penetration and transformation of various industries have only just begun. Future giants might currently be just-starting AI-native companies. Because once this efficiency-boosting tool is adopted, its use becomes indispensable.

The difference lies mainly in which AI tool to use and at what cost? The research efficiency and cost-effectiveness demonstrated by China's frontier companies once again prove this to the world.

Beyond the aforementioned areas, the global economy actually faces growth pressures, including employment pressures. Maintaining low unemployment rates in the AI era is a global challenge.

There is hope, there are problems, there are areas bustling with resource concentration, and there are deep-seated issues requiring gradual resolution—this complexity, combined with human greed and fear, manifests in the stock market as high volatility in individual stocks—making steadfast holding difficult.

Many investors lament that holding onto their early-year investments until year-end would have yielded far higher returns than active trading. This encapsulates the difficulty of investing.

Two Systems, Each with Its Stars Our core holdings remain largely stable, showing little difference from the several businesses held in the portfolio shared mid-year.

In summary: two systems, each with its stars.

Within the two primary G2 systems, we have selected companies based on their respective comparative advantages, focusing on those with reassuring valuations and operating cash flows.

Companies in the core holdings share a common characteristic: strong customer stickiness and pricing power, distinct product differentiation, and the ability to control desired profit margin levels according to their own pace.

As mentioned in previous letters, the vast majority of companies operate in a state of frantic competition, lacking the composure to control their rhythm.

Only companies with exceptionally strong business models and unique competitive positions possess the ability to control their profit levels. Such companies are extremely rare, and we might hold them until they begin to lose this privilege. This is also a key characteristic of businesses with genuine moats.

Additionally, companies serving as important AI application gateways or platforms deserve attention.

2026: Likely the True Inflection Point for AI Agent Adoption There are only a handful of companies that can be considered world-class gateways or platforms, such as Google, Meta, Apple, ByteDance, Tencent, OpenAI, etc. 2026 is likely to mark the true beginning of widespread AI Agent adoption.

AI Agents at the software level and AI Agent devices combining software and hardware will be continuously launched. AI large language models utilizing hardware like smartphones as underlying operating platforms could potentially challenge existing iOS and Android systems. Many traditional competitive advantages might be overturned during this technological wave; companies without an AI strategy risk being marginalized. This is an area requiring vigilance!

Internal "Evolution" at the Private Fund After more than 20 years of development, Greenwoods can report to our investors that we have preliminarily established research and investment capabilities in some important overseas markets.

While remaining rooted in China-background listed companies, we have spent 7-8 years building teams and systems, gradually developing the ability to identify the world's best companies along the industrial chain in certain sectors.