What Meta's RE100 exit means for tech's clean energy commitments
Meta left the RE100 renewable initiative as AI data center demand exceeds renewable supply, forcing reliance on natural gas while using questionable clean energy claims.
As pressures mount on hyperscalers to power energy-hungry AI data centers, it is becoming increasingly difficult to meet previously announced energy commitments.
In July 2026 (July 23), multiple media outlets revealed that Meta had left the Climate Group's RE100, an initiative focused on committing organizations to use renewable energy sources. The RE100 was started in 2014, with Meta (then called Facebook) joining the effort in 2016. The initiative has over 400 members, including Apple, Google and Microsoft. The key goal of the RE100 is to get all member organizations to be using 100% renewable energy and provide reporting on progress toward that goal.
According to the reports, Meta left the group after it was no longer able to meet the RE100's technical criteria for renewable energy. Meta has increasingly been investing in non-renewable energy sources including gas power. However, the company Meta claimed that it is still committed to clean energy even though it is leaving the RE100.
It's likely that Meta is being driven by conflicting priorities.
"This doesn't mean Meta has stopped investing in renewables," said Boris Kolev, global head of technology at JA Worldwide. "It means reliability, speed to capacity and AI competitiveness now appear to be more important than remaining within an external renewable-energy framework. A divorce can be mutual and still be driven by incompatible priorities."
The AI power crunch behind the exit
In 2016, when Meta (then Facebook) joined RE100, the modern generative AI boom didn't exist. The AI buildout in recent years has vastly surpassed what the current renewable energy pipeline can provide.
The International Energy Agency (IEA) forecast that U.S. electricity consumption from data centers is projected to rise from roughly 185 terawatt-hours in 2024 to about 425 terawatt-hours by 2030. The IEA also expects that natural gas and coal together are projected to meet more than 40% of the additional electricity demand from data centers globally through 2030.
Meta is aligned with the IEA's forecast and is building out natural gas-powered facilities in at least two U.S. states. In Ohio, Meta is building out its Prometheus 1 gigawatt data center campus that will benefit from a 200-megawatt natural gas plant, in addition to pulling from nuclear power.
Entergy Louisiana is constructing 10 gas plants in Louisiana that are projected to provide approximately 7.5 gigawatts of power to Meta's Hyperion data center campus.
The power strain traces back to how AI workloads differ from ordinary cloud computing.
AI workloads broke the assumptions that data centers were built on, according to Srinivas Chippagiri, a senior member of technical staff in cloud infrastructure at Salesforce. The public cloud was sized for general purpose, mostly-idle-tolerant workloads that spread demand across time and regions.
" AI training and inference are the opposite: power-dense, sustained, and concentrated in specific facilities close to the accelerators and the data.," he said.
"Still Committed" — The certificate gap
Even though Meta has left the RE100 and is actively building out gas power plants, it is still claiming that it is committed to renewable energy sources.
Meta's claim is based on its use of Energy Attribute Certificates [EACs], which follow the same market-based logic as carbon offset trading. The idea is that a company buys a credit for renewable energy that's produced somewhere to offset the non-renewable energy it is using. According to Meta, for every megawatt-hour of electricity its operations use, the company buys and retires one EAC.
It's a claim that has a few skeptics.
"Meta claiming 100% clean energy through certificates is an accounting gimmick, not actual change," said Matthew Roling, who teaches carbon accounting and climate finance at Northwestern's Kellogg School of Management.
The point of any carbon offset rests on a key concept called additionality, he said. The idea is that buying an EAC means that somewhere clean energy is being built, but this isn't always accurate.
For example, Meta could buy a cheap certificate from a wind farm that's been running for 15 years in another state and claim that its coal-powered data center is clean, although nothing new has been built and the grid has not changed, Roling said.
Meta leaving RE100 while still claiming to be 100% clean is greenwashing.
Matthew RolingExecutive Director and Assistant Clinical Professor, Northwestern Kellogg School of Management
"Buying a certificate from an old wind farm doesn't put one new watt of clean power anywhere, it's like buying a receipt for a tree someone else already planted and calling it your forest," he said.
Research published in 2022 backs up Roling's assertion by finding that companies were claiming a 31% cut in their footprint on paper, when the reality was closer to 10%.
"Meta leaving RE100 while still claiming to be 100% clean is greenwashing," he said.
Jeremy Roberts, senior director of research and content at Info-Tech Research Group is also skeptical about the use of certificates. In particular, Meta is doing its offset purchases based on its annual energy consumption, which might not be entirely accurate, he said.
"Hourly matching is a tougher test and would require Meta to balance its consumption with renewable generation on an hourly basis, not just when it's convenient," Roberts said. "It's easier to do that over a whole year, given things like seasonal trends in generation and demand."
Not just Meta — An industry-wide reckoning
The AI buildout is not limited to Meta. Every hyperscaler faces the same pressure from AI and its immense power requirements, including Apple, Google and Microsoft, which still remain among RE100 members.
Meta is not alone in using fossil fuels to power data centers, Roberts said.
"Google has done it and Amazon has been actively seeking out locations near gas plants to meet power needs," he said.
Roling expects the climate commitments from Big Tech vendors to continue to fall apart in the face of AI pressure and AI has destroyed that brand halo.
"Fifty-five percent of US data center power already comes from gas and coal, and that share is climbing," Roling said. "Now, they can't build data centers fast enough, so they've all totally walked back, watered down, or abandoned their climate commitments. Watch Apple, Google, and Microsoft next, they're on the same path Meta was two years ago."
What this means for CIOs' own commitments
Meta's exit from RE100 and the overall shift in power requirement that AI demands is shifting how CIOs are considering ESG commitments.
Companies are making calculated decisions to scale as quickly and aggressively as possible and environmental concerns are not top-of-mind, according to Roberts.
"Will they attract increased scrutiny? Almost certainly. Will that change their ultimate direction? Probably not," he said. "If gas plants are going to help these companies participate in what they perceive as a generational technology shift that could unlock trillions of dollars in value, I anticipate they will build gas plants."
That leaves the burden on individual buyers. A vendor's headline sustainability claim and its operational reality can be two different things.
"Most CIOs and IT leaders ask their SaaS or AI vendor 'are you using clean energy,' get a 'yes,' and stop there," Roling said.
Roling recommends that CIOs ask three things to dig deeper on the issues:
First, ask for power usage effectiveness [PUE], a simple efficiency score for the specific data center running your workload. The industry average is 1.58 and the best facilities run under 1.2. It's a bad sign if a vendor can't or won't provide this number to a potential customer.
Second, ask for the power mix by region, not the company average.
Third, ask if their clean energy claim is annual or hourly. Annual reporting is a shell game, hourly means something real, Roling said.
Simply taking vendor claims at face value isn't enough either. The paperwork behind a claim also matters.
Documentation should explain exactly how the company arrives at its emissions numbers and if it's shifting all power generation to renewables and are practicing hourly matching, Roberts said.
"In short, where is the electricity that is actually powering the datacenters coming from? What are the company's total emissions?" he said.
Call to Action for CIOs
For CIOs there are a few specific action items that can be taken as power demand exceeds renewable supply:
Reassess sustainability targets. AI is pushing demand past what renewables can supply, so CIOs should reassess sustainability targets, including net-zero emissions and clean energy goals for the current reality.
Integrate energy into capacity planning. AI workloads are concentrated and power-dense, so energy availability belongs in capacity planning, not only sustainability reporting.
Align procurement and sustainability teams. Energy sourcing terms belong in the contract conversation itself, not as a follow-up compliance question.
"CIOs have the power of the purse and they need to start using it to drive better outcomes," Roling said.
Sean Michael Kerner is an IT consultant, technology enthusiast and tinkerer, and has been known to spend his spare time immersed in the study of the Klingon language and satellite pictures of Area 51. He has pulled Token Ring, configured NetWare and has been known to compile his own Linux kernel. He consults to industry and media organizations on technology issues.
Dig Deeper on Sustainability and ESG strategy and leadership
