Emerald AI, Google and NVIDIA have launched the AI Energy Management Alliance, a new coalition focused on making AI data centers more flexible participants in the power grid by allowing facilities to dynamically adjust electricity consumption during periods of peak grid stress.
The founding members are joined by 18 launch partners spanning AI, semiconductors, utilities, energy infrastructure and grid software. Participants include Anthropic, Analog Devices, AES, National Grid, RWE, Constellation, NRG, Fluence, Generate Capital, GridUnity, Calibrant Energy, Camus, ClearPath, Encoord, PassKey, Splight, Verrus and Voltus.
The alliance is being formed around one of the most significant constraints facing the AI infrastructure buildout: access to electricity.
As data center development accelerates across the U.S., the companies argue that power availability, rather than capital or access to AI chips, is becoming a major bottleneck.
Traditional grid interconnection systems were largely designed around relatively predictable and static electricity demand. AI data centers, however, can require enormous amounts of power and may face years-long waits for grid connections or new generation capacity.
AEMA is proposing a different approach in which AI infrastructure can behave more dynamically.
Instead of assuming that a data center will consume its maximum power requirement continuously, operators could temporarily reduce or shift certain computing workloads when the grid experiences periods of high demand.
The alliance believes this type of demand flexibility can allow more AI infrastructure to connect to existing electrical systems while reducing pressure to immediately construct enough generation and transmission capacity to satisfy every facility’s theoretical peak demand.
AEMA plans to promote accelerated interconnection processes for data centers that demonstrate an ability to support the grid through technologies and operating strategies such as colocated power generation, battery storage and computational flexibility.
The group will also advocate for policies that recognize flexible AI demand as a potential grid resource rather than treating data centers exclusively as large, inflexible electricity consumers.
Computational flexibility is particularly relevant to AI workloads because some computing jobs can potentially be shifted across time or locations.
Certain training, inference or background workloads may not need to run at full capacity during the exact periods when regional electricity demand is highest.
If software and data center operators can coordinate those workloads with grid conditions, they may be able to reduce electricity consumption temporarily without materially disrupting customer services.
That could help utilities manage demand spikes while allowing data centers to secure power connections more quickly.
Emerald AI has made power-flexible AI infrastructure central to its strategy. The company argues that data centers capable of responding to grid conditions can become assets to utilities rather than simply additional sources of demand.
The launch comes alongside significant momentum for Emerald AI. The company recently announced a $150 million Series A at a $1.05 billion valuation to scale technology for power-flexible AI data centers.
Google is also increasingly focused on demand flexibility as it expands its global data center footprint.
The company has been investing in energy technologies and operating approaches intended to better coordinate data center electricity consumption with available grid resources.
Through AEMA, Google plans to help expand demand flexibility practices that could shorten the amount of time required to bring new computing infrastructure online while limiting the effect of data center growth on electricity costs for other customers.
NVIDIA brings another important part of the AI infrastructure stack.
As the leading supplier of GPUs used for advanced AI workloads, NVIDIA’s growth is directly connected with the rapid expansion of power-intensive computing facilities.
The company said AEMA will advocate for technology-neutral and performance-based standards that measure data centers according to their actual contribution to grid reliability.
Rather than requiring a specific technology or architecture, the alliance wants flexible data centers to be evaluated based on measurable performance.
The broader objective is to use existing electrical infrastructure more efficiently even as utilities and developers continue building new generation, transmission and storage capacity.
That could be increasingly important as AI companies race to construct massive computing clusters and utilities simultaneously face rising electricity demand from manufacturing, electrification and other industries.
AEMA’s membership reflects the need for coordination across several traditionally separate industries.
AI companies and data center operators control computing workloads. Semiconductor companies provide the hardware. Utilities operate electrical systems. Energy companies supply generation and storage. Grid technology companies provide the software and infrastructure required to coordinate those assets.
Bringing those groups together could make it easier to create common standards and operating models for flexible AI infrastructure.
Frank Lacey, an experienced energy industry executive, has been named Executive Director of AEMA.
The organization represents an evolution of the Advanced Energy Management Alliance, which was founded in 2014 to support energy technologies designed to improve grid efficiency.
The new organization is specifically oriented around AI infrastructure and the rapidly growing power requirements associated with it.
AEMA plans to convene companies across the full AI and energy value chains, including AI platforms, infrastructure providers, data center operators, technology companies, utilities and grid operators.
The alliance’s central thesis is that smarter energy management could simultaneously accelerate AI infrastructure deployment, strengthen grid reliability and protect electricity affordability.
If successful, the model could change how utilities evaluate data center projects.
Instead of assessing each facility primarily according to its maximum potential electricity demand, utilities could increasingly consider how dynamically that facility can respond to changing grid conditions.
That approach could become increasingly significant as power emerges as one of the defining constraints on the next phase of AI development.
KEY QUOTES:
“The most effective way to accelerate American AI is to make every data center a good citizen of the grid. Data centers that flex their power use in response to peak grid stress can connect faster and at far greater scale, while the communities that host them gain a more reliable grid and protection from rising electricity bills.”
Dr. Varun Sivaram, CEO of Emerald AI
“Google is committed to investing in energy solutions that benefit everyone, and pioneering smarter ways to manage our demand turns data centers into dynamic grid allies that help power systems operate more efficiently and reliably.”
Tyler Norris, Head of Advanced Energy Market Innovation at Google
“The fastest path to reliable, affordable energy is making smarter use of what already exists, even as we build the new infrastructure we need. NVIDIA is joining AEMA to advocate for technology-neutral, performance-based standards that evaluate flexible data centers on measurable grid reliability metrics.”
Josh Parker, Head of Sustainability at NVIDIA
