An AI data center can be a bet on the future and an argument over the utility bill at the same time. OpenAI has promised to cover electricity and related infrastructure costs for its facilities, while Google plans a $15 billion expansion of Malaysian data centers and cloud services that includes funding for water infrastructure. These are separate announcements, but both confront a practical problem for the computing boom: the resources are local, and so is much of the opposition.
The backlash has reached the ballot box
The construction of large, power-hungry data centers has brought disputes over electricity grids, utility bills, water supplies and the effects on nearby communities. Those disputes became a bipartisan issue ahead of the 2026 US midterm elections, with voters largely opposing expansion in the local measures tracked.
An analysis covering 41 local data center measures across 10 states found that voters supported restrictions or rejected proposals favoring construction in all but one case. That is a tally of local measures, rather than a nationwide vote on AI, but it gives developers a fairly direct reading of resistance where projects are proposed.
Following the election, Democratic leaders called for further restrictions and slower development. Companies including OpenAI and Amazon have promised closer engagement with communities and commitments to cover their data center energy costs.
The argument has moved beyond whether more computing capacity would be useful. Communities are asking how the costs and demands of that capacity will be shared. An investment announcement can describe the size of a project without settling either issue.
OpenAI promises to pay for power and project risk
OpenAI's National Community Compact sets out commitments to people living near the data centers it builds and operates. Its central financial promise is to pay fully for the electricity those facilities use and the related infrastructure.
The compact also addresses what happens when a project does not proceed as planned. OpenAI says it will provide financial guarantees if construction is delayed, a development is reduced in scale or a project is canceled.
Those guarantees are intended to protect utility customers from costs associated with unfinished developments. Paying for a running data center and protecting customers when a project stalls are different obligations. The compact promises both.
The financial commitments cover three distinct areas:
- Electricity use. OpenAI says it will pay the full electricity costs of its data centers.
- Associated infrastructure. The promise extends to the infrastructure connected with supplying that power.
- Changes to projects. Financial guarantees are intended to protect utility customers when developments are delayed, reduced or abandoned.
The distinction matters because a promise to cover normal operating bills does not, by itself, explain what happens to costs linked to an incomplete project. OpenAI is explicitly including that possibility in its commitments.
The company is also reported to be planning roughly $600 billion in computing infrastructure spending by 2030. That is an expected investment program, rather than money already spent. It provides a sense of the scale over which the company's community commitments would need to work.
A Georgia agreement addresses peak demand
OpenAI has already reached an agreement with Georgia Power concerning a planned data center in Effingham County, Georgia. The arrangement would allow electricity supplied to the facility to be reduced by up to 1,000 MW during periods of high demand.
The purpose is to ease pressure on the grid when households and businesses particularly need electricity. It gives a specific example of how a data center's power arrangement can address the needs of other customers, rather than dealing only with the facility's bill.
The 1,000 MW figure describes the possible reduction in electricity supply. It should not be read as the site's total capacity or as a new source of generation.
There are two separate issues here. One is who pays for electricity and infrastructure. The other is how much demand the grid must serve at a given moment. OpenAI's broader compact addresses the first, while the Georgia arrangement also addresses the second.
The facility is planned, so the agreement is evidence of a proposed operating arrangement, rather than a demonstration of how the site has performed during a period of grid stress. Its practical value will depend on how the arrangement works when the facility is operating and demand is high.
More disclosure, with a confidentiality exception
OpenAI's commitments extend beyond energy costs. The company says it will seek residents' views before construction, disclose development plans and publish information about water and electricity consumption.
It also intends to commission independent noise studies. That adds another local concern to a compact otherwise dominated by power and infrastructure costs.
The proposed transparency measures cover several stages of a project:
- Before construction. Residents would be consulted while the development is still being planned.
- During project planning. The company would disclose its plans for the facilities.
- During operation. Published water and electricity information would give communities a view of resource use.
- For local impacts. Independent studies would examine noise levels.
OpenAI also says it will stop using nondisclosure agreements with government officials, except where they are needed to protect confidential information. That exception matters. The commitment is a retreat from such agreements, not an unconditional promise that every discussion or document will be public.
Taken together, these measures would give residents more information with which to assess a project. Their usefulness will depend on what is actually disclosed and whether the reporting continues after construction has ceased to be the main event.
Google plans a larger Malaysian footprint
Google's announcement concerns a different market and a different package of commitments. The company plans to invest $15 billion in Malaysian data centers and cloud services during 2027-2028.
Alphabet chief investment officer Ruth Porat announced the plan at an event in Kuala Lumpur. It includes an expansion of Google's existing computing campus at Elmina Business Park in Selangor, as well as new facilities in Selangor and the neighboring state of Negeri Sembilan.
The Elmina site originated in a $2 billion investment project announced in May 2024. The new plan expands that existing location while adding others. The earlier and later figures describe separate announcements and should not automatically be added together as a single spending total.
Part of the planned investment will go toward electricity and water provision, along with workforce training. Those are supporting requirements for the computing facilities, and Google is including them in the investment program rather than presenting the campuses in isolation.
The Malaysian announcement is not part of OpenAI's compact. Its relevance is the shared problem it addresses: adding large amounts of computing capacity also requires decisions about the infrastructure around it.
Water infrastructure gets its own funding
Within the Malaysian plan, Google will fund a reservoir and water treatment plant in Negeri Sembilan costing $160 million. The facilities are intended to provide 65 million liters of drinking water a day for the data center and nearby communities.
That figure is a planned supply volume for both uses. It is not a claim that the data center alone will consume 65 million liters daily, nor is it a measurement of current consumption.
The distinction is useful because the project combines an industrial requirement with a service for surrounding communities. Google's plan puts money into the water supply infrastructure itself, alongside the computing facilities that will need it.
Water concerns have already accompanied other developments in the region. A separate $15 billion Google data center project in India prompted protests over fears about water supplies. On the Indonesian island of Batam, residents have complained of water interruptions amid data center expansion.
Those cases do not establish what will happen in Malaysia, and the complaints in Batam should not be treated as proof that data centers caused every interruption. They do show why a promise of investment can coexist with anxiety about basic services.
Funding a reservoir and treatment plant is a concrete response to the supply question. It does not, on its own, establish the eventual water consumption of the completed campuses. The infrastructure plan and the operating figures answer different questions.
Google's economic estimates cover different phases
Google expects the campus construction program to support more than 37,000 jobs annually in Malaysia and add about $1.1 billion to the country's economy. Once the facilities are operating, they are expected to support more than 9,000 jobs.
These are company estimates, not measured outcomes. The construction and operating figures also describe different phases, so they should be read separately rather than added together.
| Phase | Google's stated estimate |
|---|---|
| Campus construction | More than 37,000 jobs supported annually |
| Economic contribution from construction | About $1.1 billion |
| Facilities in operation | More than 9,000 jobs supported |
The figures describe jobs supported by the projects. They should not be presented as an announcement of an equivalent number of direct Google hires.
Malaysia has become a major center for data center development in Southeast Asia, helped by comparatively low costs, developed infrastructure and proximity to Singapore. Oracle, Amazon, Alibaba and ByteDance are also expanding their presence in the country.
Across Malaysia, the combined announced capacity of planned data center projects exceeds 6 GW. That figure concerns the wider development pipeline, not Google's facilities alone, and it describes planned capacity rather than capacity already in operation.
The scale helps explain both the investment appeal and the resource concerns. Google's expansion joins a broader construction wave, making the supporting power and water arrangements part of the story rather than a footnote to the spending figure.
OpenAI wants voluntary commitments to become industry rules
Alongside the National Community Compact, OpenAI presented its Compute Campus Policy Blueprint. It called on federal, state and local authorities to turn comparable commitments into mandatory standards across the industry.
That request draws a line between what the company is promising now and what it wants governments to require more broadly. The compact is a voluntary company commitment, not a single law or a compulsory standard applying to every data center operator.
The Georgia Power agreement is a specific utility arrangement. The proposed industry standards are a policy request. Neither should be confused with the scope of the compact itself.
For communities, the practical test will be the terms of utility agreements, the protection provided when projects change and the consistency of public reporting. OpenAI's stated commitments address those issues, but their effectiveness will depend on execution.
Google's Malaysian plan presents a different set of tests: whether the promised facilities and supporting infrastructure are built, and how their eventual resource use compares with the supply arrangements made for them. The investment and employment estimates remain forward-looking.
Both announcements make room for a less glamorous part of AI expansion. Power infrastructure, water treatment and financial guarantees now sit beside computing capacity in the case companies make to communities. The cloud has reached the part of the project where somebody has to pay for the plumbing.