Emerald AI, Google and NVIDIA have formed the AI Energy Management Alliance, a coalition the companies describe as the first of its kind, aimed at letting data centers manage electricity dynamically in response to grid conditions. The group's stated purpose is to speed up and enlarge the grid connections that AI infrastructure needs, while getting more usable watts out of the power infrastructure already in place and lowering the environmental impact of each watt consumed.
The alliance says it will stay technology-neutral and judge participants on performance rather than on particular hardware or software. Its members plan to focus on services that can be measured, including how quickly a data center responds, how long it can sustain that response, how predictable it is, and how it behaves during an emergency.
Read more: Nvidia Claims 100 Percent Water Use Reduction Inside Data Centers With Liquid Cooling
Under the principles the group has set out, operators would take on ride-through, curtailment and contingency-response obligations before they connect to the grid. The principles also call for standardized technical requirements, shared performance metrics and operational data sharing among parties. The alliance wants faster, risk-adjusted pathways for flexibility commitments it considers credible, and it wants interconnection costs to reflect the actual impacts and benefits a project brings to the system.
Membership is meant to reach well beyond the three founders. The alliance says it will convene AI platform companies, infrastructure providers, data center operators and technology firms, alongside power producers, utilities and regional grid operators. Founding members will be joined by launch partners drawn from that broader ecosystem, according to the announcement.
The group's work will include developing technical and operational approaches, cooperating with utilities and advocating for policy changes. NVIDIA and Emerald AI say they are already working with energy and infrastructure leaders on grid-responsive AI factories, and the alliance is intended to widen that effort across the United States.
The announcement does not name any other members, set a timeline for the alliance's work or attach numerical targets for flexibility.
The launch follows NVIDIA's recent push on the same theme. At the AI Infra Summit in Santa Clara, the company presented Vera Rubin and DSX platform work framed around optimizing tokens per watt for AI factories, and it has separately described how its systems aim to turn megawatts into tokens. NVIDIA has previously described Emerald AI as a partner behind its Conductor grid orchestration platform, and Silicon Valley Power has already sent a power-adjustment signal to an AI factory.
That summit drew more than 8,000 attendees, up from 3,500, with Ian Buck, NVIDIA's vice president of hyperscale and high-performance computing, among those presenting. The alliance now tests whether those technical efforts can be paired with grid rules that reward flexibility.













