Nvidia and Google Launch Alliance to Make AI Data Centers Use Less Power

TLDR

• Nvidia, Google, and Emerald AI launched the AI Energy Management Alliance (AEMA) on Wednesday
• AEMA aims to create AI data centers that can shift or reduce power use based on grid conditions
• Flexible data centers could move workloads, use stored energy, or cut demand during grid stress
• The alliance will set common standards for response speed, reliability, and performance measurement
• Global data center electricity use rose 17% in 2025 and could double to 950 TWh by 2030

---

Nvidia, Google, and energy startup Emerald AI have joined forces to tackle one of the biggest problems facing the AI industry: power.

Nvidia $NVDA, Google $GOOGL and Emerald AI are launching the AI Energy Management Alliance, or AEMA, to help AI data centers dynamically adjust electricity use based on grid conditions. pic.twitter.com/ggVpkVHWwc

— Wall St Engine (@wallstengine) September 16, 2026

The three companies launched the AI Energy Management Alliance, known as AEMA, on Wednesday. The group wants to make AI data centers smarter about how they use electricity.

Right now, data centers draw a fixed amount of power from the grid. AEMA wants to change that by making facilities that can dial their power use up or down depending on grid conditions.

How Flexible Data Centers Would Work

Under the model Nvidia described, a flexible data center could respond to grid stress in several ways. It could shift computing workloads to off-peak hours, tap stored energy, or simply reduce demand when the grid is under pressure.

This kind of flexibility could help utilities connect new AI facilities faster. Grid operators would have more confidence approving large connections if they knew a facility could back off when needed.

The alliance is technology-neutral. That means it will focus on what a data center can actually deliver, not on what specific hardware or software it uses.

Nvidia and Emerald AI had already been working on this concept before Wednesday’s announcement. Earlier tests showed that lower-priority computing workloads could be paused or rescheduled while higher-priority AI tasks kept running.

Why This Matters for Power Grids

The timing of the launch reflects real pressure on electricity infrastructure.

According to the International Energy Agency, global data center electricity consumption rose 17% in 2025 to around 485 terawatt-hours. AI-focused data centers alone saw a 50% jump in power use.

The IEA expects total data center consumption to roughly double by 2030, reaching around 950 terawatt-hours. That would be about 3% of all electricity used on earth.

In the United States, the pressure is already visible. Data centers accounted for around half of the entire increase in U.S. electricity consumption in 2025.

AEMA plans to bring together companies from across the AI and energy sectors. That includes AI platforms, data center operators, power producers, utilities, and regional grid operators.

The alliance also plans to work with utilities on grid connection solutions and push for policies that recognize and reward demand flexibility.

Both Nvidia and Alphabet shares were trading slightly higher at the open on Wednesday following the announcement.

AEMA’s work builds directly on real-time grid response demonstrations already completed by Nvidia and Emerald AI, with plans to expand that model across the U.S.