Nvidia and Google have formed an AI Energy Alliance, aiming to transform data centers from "major electricity consumers" into "grid resources."

Nvidia, Google, and Emerald AI have joined forces to form an industry alliance, attempting to fundamentally change the relationship between data centers and the power grid, transforming AI infrastructure from a passive electricity consumer into a dispatchable power grid resource.

On September 16, 2026, the three companies announced the formation of the AI Energy Management Alliance (AEMA) , with the core objective of enabling data centers to dynamically adjust their power consumption based on real-time grid conditions. NVIDIA and Emerald AI are collaborating with partners in the energy and infrastructure sectors to develop AI factories capable of responding to grid conditions in real time. AEMA is also committed to paving faster, risk-adjusted grid connection paths for facilities that make verifiable flexibility commitments.

The alliance's formation directly addresses a core bottleneck facing the expansion of AI infrastructure—power access. In key US markets, connecting new AI data centers to the power grid often takes five to seven years or even longer. AEMA believes that improving the flexible scheduling capabilities of data centers could free up to 100 gigawatts of capacity from the existing power system, while saving approximately $733 million in power system costs per gigawatt of new AI data center.

Electricity becomes a core constraint on the expansion of AI

In a statement released by the alliance, Nvidia pointed out that electricity has become a decisive constraint on the expansion of AI infrastructure in the United States. Traditional grid connection processes, designed around stable and static electricity demand, are unable to adapt to computing infrastructure with real-time response capabilities.

Flexible data centers can adjust the amount of electricity drawn from the grid in a variety of ways, including offloading computing workloads, releasing energy storage, activating supporting power generation facilities, or responding to system contingencies. These capabilities allow large electricity customers to act as a controllable resource rather than a stationary load. Nvidia states that effectively utilizing flexibility can improve the efficiency of existing grid capacity utilization, reduce demand during periods of system stress, avoid or postpone infrastructure upgrades, and give utilities and grid operators more confidence to connect AI facilities to the grid in a shorter timeframe.

AEMA, citing research data from Duke University's Nicholas Institute, stated that approximately half of the power system's capacity is idle throughout the year; a Brattle report shows that a 10% increase in grid utilization can reduce utility rates by 3.4%. Furthermore, demonstration projects have already achieved a one-third reduction in power consumption within one minute in emergency scenarios.

Technology-neutral, performance-oriented alliance principles

AEMA operates on a technology-neutral, performance-based principle, focusing on quantifiable services that facilities can provide, including response speed, duration, predictability, and performance in emergency situations, rather than specific hardware or software solutions.

The alliance principles require that obligations regarding cross-operation, power curtailment, and emergency response be clearly defined before facilities are connected to the grid, and that technical requirements, performance indicators, and operational data sharing mechanisms be standardized. Regarding cost sharing , grid connection costs will reflect the actual system impact and benefits, including avoided upgrade expenditures and improved ramp-up capabilities .

AEMA stated that the measures aim to reduce uncertainty for developers while providing system operators with the information and control mechanisms needed to maintain reliability.

Covering the entire value chain and actively participating in policy advocacy.

AEMA brings together the entire value chain in computing and power, with members including AI platforms, infrastructure providers, data center operators, technology companies, power generators, utilities, and regional grid operators. The alliance will collaborate with utilities to develop grid connection solutions and advocate for policies that recognize grid response needs.

At the policy level, AEMA stated that it maintains long-term partnerships with the Federal Energy Regulatory Commission, the Department of Energy, federal agencies, and state regulators, and can provide members with policy coordination on grid connection, transmission, and high load, as well as early warnings of policy developments.

Based on existing cooperation

This alliance is an extension of the existing collaboration between NVIDIA and Emerald AI. According to a press release issued during CERAWeek in March 2026, the two companies have partnered with AES, Constellation, Invenergy, NextEra Energy, Nscale Energy & Power, and Vistra to develop AI factories that can be connected to the grid more quickly and operate as flexible energy assets.

The plant utilizes NVIDIA's Vera Rubin DSX AI Factory reference design, which includes the DSX Flex software library for connecting the AI factory to grid services. Emerald AI's Conductor platform coordinates computational flexibility with on-site power generation, batteries, and other off-meter resources. In the aforementioned statement, Emerald AI founder and CEO Varun Sivaram said, "AI factories are so valuable that they should not be viewed as passive loads or permanent islands."

The two companies have reportedly completed AI power flexibility demonstration tests in five commercial data centers globally over the past year. DSX Flex is expected to achieve commercial-scale deployment later in 2026 at NVIDIA's AI Factory Research Center in Virginia, which plans to become one of the world's first power-flexible AI factories running Vera Rubin infrastructure. NVIDIA stated that AEMA will further expand upon this work to promote the full implementation of this model in the United States.

Risk warning and disclaimerInvesting involves risk; please exercise caution. This article does not constitute personal investment advice and does not take into account the specific investment objectives, financial situation, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article are suitable for their specific circumstances. Any investment decisions made based on this information are at your own risk.