Google and Nvidia Launch Alliance to Enhance Data Center Efficiency
Google, Nvidia, and startup Emerald AI have formed the AI Energy Management Alliance (AEMA) to revolutionize data center power usage. The coalition aims to encourage data centers to reduce their energy consumption during grid strain, in exchange for faster connections.
16 September, 2026
Credit: Ahab Sun on Unsplash
Google, Nvidia, and Emerald AI are the founding members, with 18 launch partners including Anthropic, National Grid, AES, RWE, Constellation, and NRG. Other partners include Calibrant Energy, Camus, ClearPath, Encoord, Fluence, Generate Capital, GridUnity, PassKey, Splight, Verrus, and Voltus. Frank Lacey will serve as AEMA's executive director.
Alliance Objectives
AEMA argues that power constraints, not capital or chips, are the primary barrier to expanding data centers in the US. They propose faster connections for data centers capable of adjusting their demand when needed by the grid. This could be achieved through workload shifts, battery usage, or on-site power generation.
AEMA's Key Principles:
- Setting rules for curtailment and emergency response before connection.
- Standardizing technical requirements, performance metrics, and data sharing.
- Providing faster connection routes based on credible flexibility commitments.
- Equitably allocating connection costs according to each facility’s grid impact.
Driving Change
AEMA plans to advocate for its principles with federal and state regulators, regional grid operators, and utilities.
Varun Sivaram, CEO of Emerald AI, stated: "The most effective way to accelerate American AI is to make every data center a good citizen of the grid." He further claimed that power-flexible data centers could connect faster and at greater scale, benefiting host communities with a more reliable grid and protection from rising bills.
Josh Parker, head of sustainability at Nvidia, expressed Nvidia's support for "technology-neutral, performance-based standards", emphasizing evaluation based on measurable grid reliability metrics.
Google has pledged to reduce 1GW of power demand when needed, through utility agreements across the US, as reported by Axios.
Sivaram emphasized the importance of verifying and enforcing data centers' ability to cut demand for faster connections, acknowledging the "wonky" nature of the topic.
Potential Impact
AEMA estimates that flexible AI could unlock an additional 100GW from the existing grid and points to a 5-10 year backlog for connecting new AI facilities. They suggest that improving grid utilization by just 10% could lower electricity rates by approximately 3.4%, according to a Brattle Group estimate cited in Sivaram's Fortune commentary.