Meta Left RE100 While Building 7.5 GW of Gas Power—The Gap Between Annual Matching and 24/7 Clean Energy
Meta Left RE100 While Building 7.5 GW of Gas Power—The Gap Between Annual Matching and 24/7 Clean Energy
TL;DR: Meta confirmed that it is no longer a member of RE100, the corporate initiative for 100% renewable electricity, after roughly a decade. The exit comes as the company supports a major natural-gas buildout for AI data centers, including ten Louisiana plants totaling 7.5 gigawatts. Meta says it remains committed to matching data-center electricity use with 100% clean and renewable energy. Both statements can coexist under annual energy accounting: a company can buy enough renewable energy or certificates over a year while its facilities draw fossil-powered electricity during particular hours. The meaningful question is no longer only how many renewable megawatt-hours a company procures, but whether clean supply matches where and when the load runs.
“100% renewable” sounds like a physical description: every server powered by wind or solar every second. In corporate electricity accounting, it usually means something narrower.
A company can match its annual electricity consumption with renewable generation or energy attribute certificates even when the data center itself relies on a grid that burns natural gas at night. That system helped create enormous demand for renewable projects. It also becomes harder to interpret when AI data centers add round-the-clock loads faster than grids can build firm, carbon-free supply.
Meta’s departure from RE100 makes the accounting gap visible. It is not proof that the company stopped buying renewable energy. It is evidence that the old headline claim no longer tells the entire infrastructure story.
Table of Contents
- What Meta Confirmed
- How Annual Renewable Matching Works
- Why 7.5 Gigawatts Changes the Debate
- Annual Matching Versus Hourly Matching
- What a More Useful Disclosure Would Show
- The AI Power Constraint Is Now a Product Constraint
What Meta Confirmed
Meta told TechCrunch that it is no longer part of RE100 after about ten years, describing the separation as mutual. RE100, led by Climate Group, brings together more than 400 companies committed to using 100% renewable electricity. Its members span more than 150 markets and collectively consume more than 550 terawatt-hours per year, according to the initiative’s website.
The departure follows updated accountability guidance. RE100’s July 2026 document says its future progress table will publish a standardized percentage recognized under its technical criteria, a claim-quality metric, information about older projects, missing data, over-procurement, and non-credible claims. It also distinguishes project-specific procurement from retail contracts and unbundled energy attribute certificate purchases.
TechCrunch reported that Apple, Google, and Microsoft remain among RE100’s members. Meta did not tell the publication why it left. The timing, however, overlaps with a rapid buildout of gas generation around Meta’s data-center demand.
Meta’s position is that it remains committed to matching data-center electricity use “with 100% clean and renewable energy.” That is a procurement statement. It does not necessarily describe the generation physically serving a data center in every hour.
How Annual Renewable Matching Works
Electricity from different generators mixes on the grid. A buyer generally cannot direct one specific wind-farm electron to one specific server. Instead, companies use contracts and certificates to establish the environmental attributes associated with renewable generation.
In a simplified annual-matching example, a data center consumes 1 terawatt-hour during a year. The company procures 1 terawatt-hour of qualifying renewable electricity or certificates somewhere within the applicable accounting boundaries. Its annual ledger balances, even if the facility consumes heavily at 2 a.m. when local wind is low and solar output is zero.
This system is not meaningless. Long-term power-purchase agreements can finance new renewable plants, and certificates provide a traceable claim to renewable generation attributes. The problem is that equal annual totals can conceal a large hourly mismatch.
Imagine a facility drawing a constant 100 megawatts. Its annual consumption is:
100 MW × 8,760 hours = 876,000 MWh
Now imagine contracted solar assets generate 876,000 MWh during the year, concentrated in daylight. Annual matching reaches 100%. Yet the facility still needs 100 MW at midnight. Without storage, transmission, demand flexibility, or another clean firm source, the grid fills that gap with whatever generation is available. In a gas-heavy region, that can be natural gas.
The arithmetic is hypothetical, but the principle is exactly why electricity buyers are moving beyond annual claims.
Why 7.5 Gigawatts Changes the Debate
TechCrunch reported that Meta initially backed three large Louisiana gas plants for its Hyperion data center and later supported seven more. Together, the ten plants would provide 7.5 gigawatts.
Scale that number against ordinary data-center language. A 7.5 GW system running continuously would have a theoretical maximum annual output of:
7.5 GW × 8,760 hours = 65.7 TWh
Actual generation would be lower because plants do not run at full capacity every hour. The calculation is not a claim about planned output or Meta’s consumption; it simply shows the order of magnitude. RE100 says its entire member group uses more than 550 TWh annually. A theoretical 65.7 TWh would equal nearly 12% of that benchmark.
The comparison is intentionally rough, but it demonstrates why AI infrastructure is straining the language of corporate sustainability. These are not office-building loads that can be offset with a modest rooftop installation. They are power systems measured in gigawatts, with reliability requirements closer to industrial facilities.
Natural gas offers dispatchability. It can run at night, respond to load, and be built with technology utilities already understand. It also emits carbon dioxide and local air pollutants. Calling the surrounding portfolio “100% renewable” through annual matching does not change the physical emissions of the plant.
Annual Matching Versus Hourly Matching
Hourly, or 24/7, matching asks a stricter question: for each hour in a specific region, was enough carbon-free electricity available to match the buyer’s consumption?
| Accounting approach | Core question | What it can hide |
|---|---|---|
| Annual matching | Did renewable procurement equal yearly use? | Hourly and local fossil dependence |
| Hourly matching | Did clean supply match use in each hour? | Less; still depends on credible location and data |
| Physical islanded supply | Did dedicated assets directly serve the facility? | Backup and lifecycle impacts may remain |
Hourly matching encourages a different portfolio. Solar alone cannot cover midnight. Buyers need combinations of wind, batteries, geothermal, nuclear, hydro, long-duration storage, flexible demand, or clean generation in regions connected by adequate transmission.
It is also harder and often more expensive. That is the point. A stricter measure exposes the hours when decarbonization is difficult instead of allowing abundant midday solar to erase fossil-heavy nighttime consumption on an annual spreadsheet.
TechCrunch noted that Microsoft is pursuing hourly matching and cited Google’s work pairing renewables with batteries. These approaches are not perfect, but they move the claim closer to the physical reality of a continuously operating data center.
What a More Useful Disclosure Would Show
The public does not need a single binary badge—renewable or not renewable. It needs a layered scorecard.
First, disclose annual electricity consumption and renewable procurement, separated by market. That prevents excess procurement in one place from obscuring fossil-intensive use somewhere else.
Second, disclose the share matched hourly with carbon-free energy. A company might report 100% annual renewable matching and 65% hourly carbon-free matching. Those numbers tell a more honest story together than either does alone.
Third, separate new project-specific contracts from unbundled certificate purchases. RE100’s updated guidance moves in this direction by publishing procurement types and certificate coverage.
Fourth, report the emissions and local pollution associated with dedicated gas generation rather than netting it away in a renewable claim. Communities near power plants experience physical emissions, not annual accounting.
Finally, disclose load growth. A company can add renewable procurement and still increase total fossil generation if data-center demand grows faster. Absolute consumption, marginal supply, and matched percentage all matter.
These disclosures would not require companies to abandon annual procurement. They would prevent annual procurement from being mistaken for 24/7 physical decarbonization.
The AI Power Constraint Is Now a Product Constraint
AI companies once treated electricity as an input purchased after the product roadmap was set. Gigawatt-scale campuses reverse that relationship. The availability of power, transmission, cooling, and permits determines where and how quickly the product can scale.
Gas is attractive because it converts fuel infrastructure into dependable electricity on a timetable that can be easier to control than a regional transmission build. But a gas-dependent roadmap also creates fuel-price exposure, carbon risk, local permitting conflict, and a credibility problem for climate commitments.
Meta’s RE100 exit should therefore be read less as a membership story and more as a signal that AI’s infrastructure reality is colliding with sustainability frameworks designed in an era of smaller loads. RE100’s tighter accountability guidance is one response. Hourly matching is another. Neither creates clean firm power by itself.
The practical challenge is to build enough generation, storage, and transmission that 24/7 clean claims become physically achievable rather than merely better audited. Until then, the honest description of many AI data centers will be two statements at once: the company procures renewable energy equal to annual use, and the grid serving the facility still burns fossil fuel in critical hours.
That may be uncomfortable, but it is more useful than arguing over whether one “100%” label is technically permitted.
Frequently Asked Questions
Did Meta stop buying renewable energy?
No such conclusion follows from leaving RE100. Meta says it remains committed to matching data-center electricity use with 100% clean and renewable energy, and TechCrunch reports continued renewable procurement.
What is RE100?
It is a Climate Group-led initiative for major companies committed to 100% renewable electricity. Its website reports more than 400 members across more than 150 markets.
How can a company use gas and still claim 100% renewable energy?
Under annual matching, it can procure renewable electricity or certificates equal to its yearly consumption even when local power during some hours comes from fossil generation.
How much gas generation is connected to Meta’s Louisiana plans?
TechCrunch reports ten plants associated with the Hyperion project totaling 7.5 GW.
Is hourly matching the same as direct renewable power?
No. It is a stricter accounting method that matches consumption and carbon-free supply by hour and location. Physical delivery, grid mix, storage, and backup resources still matter.
Key Takeaways
- Meta confirmed it has left RE100 after about a decade, while saying it remains committed to 100% clean and renewable energy matching.
- Ten Louisiana gas plants associated with Meta’s Hyperion data-center project would total 7.5 GW, according to TechCrunch.
- Annual matching can reach 100% on a yearly ledger while fossil generation still serves the load in specific hours.
- A hypothetical 7.5 GW running continuously equals 65.7 TWh per year; this is a scale illustration, not a generation forecast.
- More useful disclosure would combine annual procurement, hourly carbon-free matching, procurement type, local emissions, and absolute load growth.
- AI’s electricity demand is turning energy availability into a product and deployment constraint, not a back-office procurement choice.
How this was written: AI assisted with research, arithmetic, and structure. Every source, calculation, and final claim was checked and edited by a human operator.
Comments ()