Understanding Digital Advertising's Carbon Footprint in 2026

Every digital ad you run generates carbon emissions. Not metaphorically. Literally.

The servers running the auction, the networks moving the creative, the device rendering the ad: all of it draws electricity, and most of that electricity still comes from carbon-intensive sources. But in 2026 there is a second half to the story, one the industry spent years leaving out: the carbon of manufacturing that hardware in the first place. Counting both is what separates a comfortable estimate from an honest one.

Where ad emissions come from

There are two questions, not one. How much energy does an impression use? And how much carbon was embodied in building the hardware that serves it? GMSF v1.3, the current Ad Net Zero standard, answers both, across the three stages of an impression’s journey.

1. Inventory selection (the auction)

Before an ad is shown, it is bought. A single display impression can trigger hundreds to thousands of bid requests across a chain of intermediaries, each one running servers. The deeper the programmatic supply chain (read from a publisher’s ads.txt and app-ads.txt), the more infrastructure is activated, whether the bid wins or loses.

2. Creative delivery (the network)

The creative travels from origin to edge to device across CDNs, internet backbone, and mobile or fixed networks. Heavy files multiply the impact: a 5 MB video moves an order of magnitude more data than a 500 KB display unit.

3. Device usage (the screen)

The user’s device decodes and renders the ad and lights the screen for as long as it stays in view. On mobile this is a large share of the total, and it is where embodied carbon bites hardest: manufacturing a smartphone is carbon-intensive, and a slice of that footprint is attributable to every ad it displays.

The dimension older methods missed: embodied carbon

Spend-based shortcuts and use-phase-only models (such as GMSF v1.2) counted only the electricity. GMSF v1.3 adds the embodied phase: the manufacturing footprint of servers, network gear, and devices. It is often the single largest contributor, and leaving it out is the main reason earlier numbers looked reassuringly low. We cover that shift in detail in Carbon Intelligence adopts GMSF v1.3.

The numbers that matter in 2026

The working metric is gCO2PM: grams of CO2e per thousand impressions.

Under v1.3’s full life-cycle accounting, typical campaigns measure in the low hundreds of gCO2PM. The exact figure is driven by format (video costs more than display), device mix (mobile and connected TV carry heavy embodied loads), geography (a French nuclear grid is far cleaner than a coal-heavy one), and the depth of the programmatic chain.

For reference, the v1.3 performance bands we use run roughly: excellent below 195 gCO2PM, good up to 250, high up to 360, and critical above that. These sit about 5 to 6 times higher than equivalent v1.2 figures, not because campaigns changed, but because embodied carbon and full supply-chain depth are finally in the total.

Context: at scale, a large advertiser’s annual advertising footprint still runs to hundreds or thousands of tonnes of CO2e. The headline number is bigger under v1.3, but it is the one you can actually defend.

Why methodology matters

Not all carbon numbers are equal, and the gap between them is not a rounding error.

• A spend-based estimate multiplies your budget by a sector average. It can land near the right total by coincidence, but it cannot tell two campaigns apart, cannot guide optimization, and cannot be substantiated in an audit.
• A use-phase-only model (like GMSF v1.2) measures real activity, but counts only electricity, so it understates the true footprint.
• A full life-cycle assessment (GMSF v1.3) counts use and embodied emissions from operational data. It is bigger, truer, and audit-ready.

We changed our own conclusion when the methodology improved. Read why, and what it means for the “spend-based overstates by 450%” claim you may have seen before.

What is driving the need to measure

Three forces are making carbon measurement mandatory.

1. Regulation

CSRD (Europe): thousands of companies must report Scope 3 emissions, advertising included. The Omnibus I changes (December 2025) simplified some requirements but kept the disclosure and assurance obligations.

California: companies above 1 billion dollars in revenue must disclose Scope 3 emissions, which includes digital advertising.

SEC (USA): climate disclosure rules require material climate risks to be reported. For many advertisers, advertising emissions qualify.

2. Market pressure

Green-media requirements now appear in partner RFPs, procurement teams ask for carbon data, investors scrutinise Scope 3, and consumers increasingly favour sustainable brands.

3. Cost optimization

Companies that measure advertising carbon discover something useful: high-carbon placements often correlate with waste. Made-for-advertising sites, oversized creative, bloated supply chains, and poor targeting all tend to be high carbon and low value at the same time. Fix the carbon problem and you usually fix the performance problem too.

How to measure your footprint

Step 1: Use GMSF v1.3

Not spend-based proxies, and not use-phase-only models. v1.3 measures impressions served, creative weight and format, supply-chain depth, device split, geography, and the embodied carbon of the hardware involved.

Step 2: Collect campaign data

Gather impressions, creative specifications, geographic breakdown, device distribution, and supply-path detail from every platform.

Step 3: Model the full life cycle

Total footprint =
  Inventory selection (auction infrastructure) +
  Creative delivery (network transfer) +
  Device usage (rendering and screen)
  ... each counted for both USE and EMBODIED carbon

Step 4: Apply carbon intensity factors

CO2 = energy (kWh) x regional grid intensity (gCO2/kWh)

Regional factors matter enormously. Using Ember 2024 grid data, a kilowatt-hour ranges from roughly 30 gCO2 in Norway and 44 in France to about 344 in Germany and well above 600 in coal-heavy grids. The same impression can differ by an order of magnitude depending on where it is served.

A real-world example

European e-commerce company, annual advertising footprint

• Total spend: 3.2 million euros
• Impressions: 450 million across Google Ads, Meta, DV360, and Amazon

Measured properly under a full life-cycle model, the footprint is several times what a use-phase-only number would have shown, because embodied carbon and full supply-chain depth are now included. More importantly, the measurement is granular, so the team could see exactly where the carbon sat:

• Made-for-advertising domains: 27% of the footprint
• Oversized creative: 20%
• Long supply chains: 17%

Actions taken: blocked roughly 8,000 MFA domains, compressed the creative library (average file size down 52%), consolidated to four preferred SSPs, and switched on carbon-aware bidding.

Year-two results: emissions down 44%, spend down 6%, return on ad spend up 18%. The carbon optimization and the performance optimization turned out to be the same project.

The bottom line

Digital advertising’s carbon footprint is:

• Real: every impression generates measurable emissions, in use and in embodied hardware.
• Bigger than the industry admitted: full life-cycle accounting lands well above older use-phase numbers.
• Measurable: GMSF v1.3 provides an accurate, audit-ready methodology.
• Optimizable: reductions of 40% or more are achievable while maintaining or improving ROI.

The companies succeeding in 2026 treat carbon as a performance metric alongside CPA, ROAS, and CTR, and they measure it with a method they can defend.