Summary of Key Points
This news article focuses on the new trends in artificial intelligence (AI) computing infrastructure: As AI evolves from large-scale model training to more complex applications involving high-concurrency reasoning, the competition in computing power is no longer solely about chip performance. Instead, it has shifted to a "system-level collaboration" that encompasses computing, networking, power supply, storage, and software. At the forum held on that day, Inner Mongolia and the Yangtze River Delta region collaborated to establish a computing power research institute, launched a cross-regional computing power transmission technology platform, and an operating system for coordinating power consumption with computing resources. They also expanded the domestic computing power ecosystem alliance and initiated cooperation to provide computing services overseas under the "Belt and Road" initiative, all aiming to move towards a more systematic integration and internationalization of domestic computing power.
1. The Competition in AI Computing Power Has Changed: From "Chip-by-Chip Battles" to "Teamwork"
In the past, the competition in AI computing power was mainly about how fast individual chips could perform. However, AI applications have become increasingly complex—large models need to respond to millions of users simultaneously (high-concurrency reasoning), and intelligent systems may need to connect multiple systems (such as robots handling vision, speech, and decision-making tasks). In such cases, having a fast chip alone is not enough; other factors are critical, such as the speed of data transmission over the network, the stability of power supply (since AI consumes a lot of energy), and the compatibility of different software and storage solutions. It's like running a restaurant: previously, the focus was on the chef's skills, but now it's also essential to have good kitchen equipment, a reliable supply chain of ingredients, and well-coordinated staff—any one of these elements can be a bottleneck. Therefore, the competition in computing power is about "system-level collaboration," where all components must be optimized together.
2. Cross-Regional Computing Power Cooperation: The Complementary Partnership between Inner Mongolia and the Yangtze River Delta
Inner Mongolia and the Yangtze River Delta have signed a strategic cooperation agreement to establish an advanced computing network research institute. Why these two regions? The Yangtze River Delta is a major consumer of AI applications (with many AI companies in cities like Shanghai and Hangzhou), but the cost of electricity is high there; Inner Mongolia, on the other hand, has abundant green energy sources (wind and solar power) and lower electricity costs, making it suitable for building large-scale computing centers. Cross-regional connectivity allows Inner Mongolia's computing power to be transported to the Yangtze River Delta, similar to the transmission of electricity from the west to the east. The "100-Terabyte, Thousand-Kilometer Fiber Optic Direct Connection Platform" launched that day is designed to test how to achieve fast and stable long-distance data transfer, addressing the issue of insufficient computing power in the east while having excess capacity in the west.
3. Power Consumption Matters: Computing Power Operating Systems Help Save Money
AI computing power is highly energy-intensive—a large computing center can spend hundreds of millions on electricity annually. The "Zhongke Lèinǎo Computing-Power OS" (Chinese for "Brain-like Computing-Power Operating System") released that day acts as an intelligent manager for both computing and power consumption. It determines when it's cheaper to use electricity (e.g., at night) and schedules tasks to make the most of available power. This system is similar to a smart home energy meter, automatically adjusting the operation times of appliances to save energy without compromising performance.
4. Domestic Computing Power Comes Together: Alliances to Overcome Incompatibility Issues
Domestic AI chips (such as Ascend and Moore Threads) are being used more frequently, but new challenges have emerged: software developed for one chip may not work with another, just as apps from different platforms (e.g., Android and iOS) are not compatible. The newly formed integrated computing power ecosystem alliance includes chip manufacturers, software companies, and application developers, who are working on three main goals: ensuring that software can run on various chips, unifying the management of different computing centers (like using a single app to control all smart devices at home), and finding practical use cases (e.g., in healthcare and autonomous driving) to make domestic computing power more reliable and widely applicable.
5. Computing Power Services Go Global: The "Belt and Road" Initiative Uses Tokens as a Medium of Exchange
A strategic alliance for using tokens in the "Belt and Road" region was also established. These tokens are not cryptocurrencies but more like digital credentials for computing power services. For example, Chinese computing centers can provide services to companies in Southeast Asia, using tokens for settlement and authentication, which is secure and convenient for cross-border transactions. This initiative helps expand China's computing power capabilities into international markets and makes AI services more affordable for those countries.
Why Are These Developments Important?
AI is at the heart of future technological competition, and computing power is its essential fuel. By addressing issues such as cross-regional resource allocation, reducing costs, and improving compatibility between domestic chips and software, China aims to create a more stable, cost-effective, and autonomous AI ecosystem. This will give China a competitive advantage in the global arena. For ordinary people, this means that future AI services (such as intelligent customer service and autonomous driving) are likely to become more accessible and affordable due to lower computing power costs.