China's Intelligent Computing Reaches 2,185 EFLOPS
China's Intelligent Computing Reaches 2,185 EFLOPS
Anonymous
Key Points
1
China’s intelligent computing capacity grew 177% year over year by the end of June, underscoring the speed of the country’s infrastructure expansion.
2
Seventeen regional nodes have been approved to support interoperability across the national computing network, adding an institutional layer beyond the more than 70 transmission channels already built.
3
China is exploring “computing supermarkets” and “computing banks” as service models intended to lower access barriers for smaller companies and make computing resources available on demand.
4
Chinese authorities plan to develop high-speed, low-latency and resilient transmission links, including city-level access with millisecond response times, while building a multilevel and tiered computing architecture.
5
The AI infrastructure buildout is also being accelerated through retrofitting existing Bitcoin-mining sites and power assets, with industry leaders comparing the transition to earlier technology revolutions rather than viewing it as a short-lived speculative cycle.
China's national computing network reached 2,185 EFLOPS of intelligent computing capacity by June, a 177% jump from the year before. Bloomberg reported that the country is now shifting focus from building more facilities to connecting them efficiently and making computing power available to smaller businesses through new service models like 'computing supermarkets' and 'computing banks'.
Globally, the AI infrastructure boom is accelerating as companies convert Bitcoin mining sites into data centers to meet soaring demand. TrendForce projects that data center power demand will hit 161GW by 2026, up 31% year-over-year, with AI servers now accounting for over 30% of this growth. Industry leaders say this expansion reflects real, sustained demand — not a bubble.
China's 17 New Computing Hubs Boost Network Coordination
China approved 17 regional nodes to help different parts of the computing network work together seamlessly. These nodes sit on top of more than 70 existing transmission channels that already connect major cities and regions. The goal is to make it easier for data to move across the country without bottlenecks, ensuring computing resources reach businesses faster and more reliably.
Computing Supermarkets Bring AI Power to Smaller Companies
China is testing new service models called 'computing supermarkets' and 'computing banks' to lower barriers for small and medium-sized businesses. Instead of building their own expensive data centers, companies can rent computing power on demand — similar to how you buy electricity from a grid. The strategy reflects a shift in priorities: authorities now focus on utilization and accessibility rather than just adding more capacity.
To support this transition, China plans millisecond-speed, low-latency transmission links at the city level and is building a multilayered computing architecture. This means processing power will be available both locally and at the national level, giving businesses flexible options for where their work gets done.
Bitcoin Mining Sites Transform Into AI Data Centers
Across the globe, companies are repurposing Bitcoin mining operations into AI data centers, converting existing power infrastructure and real estate into valuable computing assets. AIB has launched a 'power-first strategy' to rapidly build data centers for mid-market AI and cloud firms, tackling one of the biggest bottlenecks in the industry: reliable power supply.
Industry experts reject claims of an AI infrastructure bubble, comparing the buildout to earlier technology revolutions like the railroads or internet. Goldman Sachs noted that global AI demand is actually outpacing infrastructure growth, meaning companies still struggle to secure enough computing capacity fast enough. Rising costs and regulatory scrutiny remain concerns, but sustained adoption and power availability suggest expansion will continue.
Data Center Demand Strains Global Power Supply
TrendForce projects a global data center power gap of 268GW by 2030 as AI adoption accelerates. By 2026 alone, data centers will need 161GW of power capacity — a 31% increase year-over-year — with AI servers driving the surge. Mississippi is already seeing construction booms as utilities and developers rush to build new facilities to meet demand.
Sentiment
Mostly PositiveNEGATIVENEUTRALPOSITIVE
Publishers
17
Articles
8
Reach
25
A
AnonymousAuthor
