Global AI, Malaysia Angle / AI news for Malaysia
Google, NVIDIA and Emerald AI launch an alliance for grid-flexible AI data centres
Three AI heavyweights want data centres treated as loads that can ease off when the grid is stressed, in exchange for faster power connections. The pitch is American, but the power problem is familiar in Malaysia.
In brief
- Emerald AI, Google and NVIDIA launched the AI Energy Management Alliance on 16 September 2026 to promote data centres that adjust their grid power use when the system is under strain.[1]
- The alliance wants performance rules for flexibility, standard metrics and faster grid connections for facilities that make verifiable commitments. Its stated focus is the United States.[1][2]
- MIDA has projected Malaysian data centre demand above 5,000 MW by 2035, but no source links AEMA to Malaysia or names a Malaysian member.[4][1]
Three AI companies form a lobby for grid-responsive data centres
Three companies at the centre of the AI build-out want data centres to stop behaving like fixed, always-on loads. On 16 September 2026, Emerald AI, Google and NVIDIA announced the AI Energy Management Alliance, or AEMA, a coalition for facilities that can change how much electricity they draw from the grid.[1]
The bargain on offer is simple to state. A data centre that eases off when the power system is stressed could, in the alliance's view, be connected sooner and at a larger size than one that demands full power at every hour.[1][2]
What the alliance is asking grid operators to accept
NVIDIA's announcement describes several ways a flexible facility can cut what it takes from the grid: shifting computing workloads, discharging on-site storage, running paired generation or responding to system emergencies. The alliance says it is technology-neutral and judges facilities on the service they deliver, such as response speed, duration, predictability and behaviour during an emergency.[1]
Its principles call for ride-through, curtailment and contingency obligations to be defined before a facility connects. It also wants standard technical requirements, performance metrics and data sharing, faster risk-adjusted connection pathways for operators that make credible and verifiable commitments, and connection costs allocated according to actual impact on the system.[1]
The numbers behind the pitch are the alliance's own
AEMA's website makes large claims. It says half of the US power system's capacity goes unused through the year, and that moderately flexible data centres could unlock 100 gigawatts from the existing grid. It also says connecting a new AI data centre takes five to seven years or more in key US markets.[2]
The site adds that demonstrations have cut power use by a third in under a minute in emergency scenarios, and that each GW of new AI data centres with flexibility and capacity resources could avoid about USD733 million in power system costs. These are advocacy figures. The announcement cites no independent study for them, and no utility has said how it would verify flexibility.[2][1]
Who is behind it, and what they gain
Emerald AI sells the software at the heart of this idea. In August 2026 it announced a USD150 million Series A at a USD1.05 billion valuation, co-led by Energize Capital and DCVC, and said its Emerald Conductor platform now flexes power commercially at multi-megawatt, full data centre scale after five global demonstrations.[3]
NVIDIA says it and Emerald AI already work with energy and infrastructure firms on AI factories that respond to grid conditions in real time. Google appears as a founding member on the AEMA site. The alliance says launch partners will join, but the announcement names none, and it publishes no membership fee, governance structure or timetable.[1][2]
Why Malaysia should care
Malaysia faces a version of the same squeeze. In a March 2025 analysis, MIDA said data centre energy consumption could pass 5,000 MW by 2035, about 40% of Peninsular Malaysia's current power capacity, and that energy supply applications already exceeded 11,000 MW. MIDA pointed to efficiency metrics in incentive rules and to green power schemes as responses. Neither AEMA nor its founders mention Malaysia, and no Malaysian utility or operator is named as a member.
Data centre operators and developers in Malaysia
Grid flexibility is becoming a bargaining chip for faster, larger power connections in the US.[1][4]
Practical move: Ask how much of your planned load could be curtailed or shifted without breaching customer contracts.
SMEs and households paying electricity bills
The alliance argues flexible data centres ease grid strain and support affordability, but offers no Malaysian evidence.[2]
Practical move: Treat affordability claims as unproven here until a Malaysian regulator or utility publishes numbers.
Companies buying AI cloud capacity
Flexible sites may pause or move some workloads when the grid is stressed.[1]
Practical move: Check service terms for curtailment clauses before running time-critical AI work.
What Malaysians can do now
- Read AEMA's figures as advocacy claims about the US grid, not as measured results for Malaysia.
- If you plan or lease data centre capacity, ask the provider whether its power supply depends on curtailment commitments.
- Watch for any Malaysian utility, regulator or operator joining the alliance or publishing its own flexibility rules.
What we still do not know
The principles are public. The members and the proof are not.
- Which companies are the launch partners, and whether any utility has agreed to faster connections in exchange for flexibility.
- How curtailment would be measured and verified, and by whom.
- Whether the alliance plans to work outside the United States, including in Southeast Asia.
Sources
- 1.Emerald AI, Google and NVIDIA Launch Alliance to Advance Flexible AI Data Centers NVIDIA, 16 September 2026
- 2.AI Energy Management Alliance (AEMA) AI Energy Management Alliance
- 3.Emerald AI Raises $150M Series A at $1.05B Valuation Emerald AI, 25 August 2026
- 4.Powering Up: Navigating the Energy Crossroads of Malaysia's Data Centre Boom Malaysian Investment Development Authority, 24 March 2025
- 5.Google Data Center, Council Bluffs Iowa (49062863796) Wikimedia Commons, 18 February 2017
- 6.Google data center and wind turbines in Eemshaven Wikimedia Commons, 6 January 2023
- 7.Bandar Utama TNB Transmission Main Intake Wikimedia Commons, 28 November 2021
