
CO₂e · META 2025
11.8Mt
EMISSIONS YOY
+45%
GLOBAL DC POWER
415 TWh
DC DEMAND BY 2030
~2×
Meta has effectively shelved its 2030 net-zero emissions target, blaming the energy demands of AI infrastructure for making the timeline unworkable. The company's annual sustainability report reveals a 45% year-on-year increase in net carbon-equivalent emissions, reaching 11.8 million tonnes — a figure itself calculated using accounting methods that climate advocates dispute. Executives Rachel Peterson and Blair Swedeen framed the retreat as a matter of complexity rather than abandonment, writing that 'the reality has been far more complex' than the steady-growth, falling-cost trajectory assumed in 2020.
Extraction Map
The retreat is structurally predictable. When Meta set its value-chain net-zero goal in 2020, data centres were growing but AI had not yet triggered the current hyperscale buildout. Today, 403 of the most powerful US data centres alone account for roughly 2% of national electricity use and produce emissions comparable to Denmark's. Globally, data centres consumed about 415 terawatt hours in 2024 — approximately 1.5% of world electricity supply — and the International Energy Agency projects that figure could nearly double by 2030. Meta is not adapting to bad luck; it is feeding the demand that broke its own promise. The accounting deserves scrutiny. Net-zero pledges allow companies to count purchased carbon credits against gross emissions. The 11.8 million tonne figure already incorporates these offsets, and climate advocates have long questioned whether the credited removals are real, permanent, or additional. When the gross number rises 45% in a single year, the entire offset framework looks decorative — a ledger trick that buys time rather than cuts pollution. Meta is not alone. Google, which pledged carbon-free energy by 2030, says its emissions are still rising. General Motors backed the Trump administration's weakening of fuel economy rules, retreating from its 2035 zero-emission vehicle target. Walmart acknowledged it would miss its first emissions goal. Coca-Cola and Pepsi have quietly shifted their baseline years and softened their ambitions. The pattern is consistent: pledges made during a period of climate consensus are being abandoned under political cover from the second Trump administration. The political economy is worth naming plainly. These companies made climate pledges when doing so carried reputational upside and negligible cost. Now that honouring those pledges would require real capital expenditure and operational constraints — building renewable supply at the pace of data centre demand, or slowing that demand — the pledges evaporate. The costs of emissions remain externalised onto the public, while the revenues from AI infrastructure accrue to shareholders. This is a textbook extraction cycle: private benefit, socialised harm. The Nature Sustainability study published this week adds empirical weight. With approximately 1,200 hyperscale data centres currently online and hundreds more planned, the energy trajectory is not a projection — it is a construction schedule. If corporate climate targets flex every time growth accelerates, those targets are not constraints. They are marketing. What remains is a governance vacuum. Voluntary pledges have been tested and found disposable. Without binding regulatory mechanisms — carbon pricing, emissions caps, mandatory renewable procurement at the rate of new load — the AI buildout will proceed on the cheapest available electrons, which in most grids still means fossil generation at the margin. The 2030 goal is not deferred. It is dead. What replaces it will determine whether AI becomes the most energy-intensive industry in history with no emissions accountability, or whether the infrastructure boom is channelled through enforceable decarbonisation requirements.
Worth reading?
SKIP
Score
20
/100
Extractive. AI infrastructure profits accrue to Meta shareholders; emissions costs are socialised onto the global public. Voluntary climate pledges function as reputational shields with no enforcement mechanism, discarded when they conflict with growth.
What Is GRIN?
GRIN stands for Generativity, Resilience, Innovation efficiency, and Novelty openness. These four dimensions capture the essential characteristics of any memetic intelligence — whether a nation, a corporation, a religion, or a movement.
About GRINGenerativity
1/5
Resilience
1/5
Efficiency
1/5
Novelty
1/5
Extractive
4/5
Inverted
Does this create new capability or just redistribute existing value?
Zero new capability for anyone but Meta. AI buildout generates private returns while destroying the shared climate commons. No generative spillover to public or planet.
Verified Facts (10)
Meta's net carbon-equivalent emissions rose 45 percent in 2025 to 11.8 million tonnes
Meta set a goal of net-zero emissions across its entire value chain by 2030 in 2020
403 of the most powerful data centres in the United States accounted for about 2 percent of the country's electricity use
Globally, data centres used about 415 terawatt hours of electricity in 2024, about 1.5 percent of supply
Data centre electricity use could nearly double by 2030, according to the International Energy Agency
About 1,200 hyperscale data centres are online, and hundreds more are planned
General Motors set a goal of selling only zero-emission vehicles by 2035
Walmart acknowledged in 2024 that it would miss its first emissions target
Google pledged in 2020 to run on carbon-free energy by 2030
Coca-Cola and Pepsi have changed the baseline years for their emissions targets
More latest news
Read all