發生了什麼事
Axios reports that a coalition of AI startups, cloud providers, researchers, and investors has launched the National Grid. The initiative aims to pool AI computing capacity from various sources to address a supply crunch that is driving enormous infrastructure spending. The system uses a shared scheduler to match workloads with available capacity, showing members details like chip type, location, and pricing. The consortium states it currently has about 760 megawatts connected or in sight, with a goal of reaching 2 gigawatts by 2030. Access is opening to public-sector employees, including government, education, and national laboratory users.
Axios reports that a coalition of AI startups, cloud providers, researchers, and investors is launching the National Grid. The primary goal is to pool AI computing capacity to help address a supply crunch that is driving enormous spending on AI infrastructure.
The system is designed to show members available capacity, chip type, location, pricing, and utilization, then automate how workloads are matched to it. Grid members can contribute idle capacity and reserve larger clusters for planned training runs.
Anjney Midha, a leader of the effort and former venture investor who now runs AI holding-company Amp, stated that the project is a way to share computing for both commercial research and public-sector use. He emphasized the need for coordination around an open standard to scale AI efficiently in America.
Sam Sinha, Head of AI at 1X, noted that smaller AI companies often struggle to get access to computing resources because larger players like OpenAI and Anthropic can pay far more and sign long-term contracts. Sinha argued that a healthy AI ecosystem needs more than two companies to own all the .
The consortium says it has about 760 megawatts connected or in sight, with a goal of 2 gigawatts by 2030. The paper accompanying the announcement states that independent, single-tenant data centers average less than 15% net computing utilization, leaving expensive capacity unused.
為什麼這很重要
This development addresses a critical bottleneck in the AI industry: the high cost and scarcity of computing resources. By coordinating idle capacity, the grid could significantly alter the direction of the AI boom by making more accessible to smaller players who cannot afford long-term contracts with major providers. It also aims to increase the overall supply of scarce computing resources by bringing existing, underutilized capacity online.
Building AI infrastructure has become a multi-trillion-dollar bet largely because of the high cost of accessing scarce computing resources. However, many data centers run at low capacity, meaning extremely valuable chips sit idle while some startups and researchers struggle to get access.
The creators of the grid want to make it easier for idle computing capacity to be used. If successful, this could significantly alter the direction of the AI boom by reducing the barrier to entry for smaller companies and researchers.
The initiative aims to increase the supply of one of the scarcest and most precious resources in the world by finding ways to bring more existing capacity online. This could help mitigate the supply crunch that is currently driving up costs across the industry.
互動機制：它實際上是如何運作的
以互動方式探索這項發展背後的基礎技術。
Which component of an AI application is the machine-learning model itself?
接下來看什麼
Monitor the actual utilization rates of the grid and whether it successfully reduces costs for smaller AI companies. Watch for the timeline of reaching the 2 gigawatt goal by 2030 and the specific terms of access for public-sector users.
Watch for the actual utilization rates of the National Grid and whether it successfully reduces costs for smaller AI companies and researchers.
Monitor the timeline for reaching the 2 gigawatt goal by 2030 and the specific terms of access for public-sector users, including government, education, and national laboratory users.
Observe how the shared scheduler handles the matching of workloads to available capacity and whether it effectively addresses the issue of idle chips in data centers.