IT Carbon Footprint for Enterprise — 2026 Guide

Citation capsule

• "ADEME Base Empreinte gives 169 kg CO₂e for a 14-inch laptop manufacture; Fraunhofer IZM 2022 cites 177.6 kg for new-laptop manufacturing with 300–400 kg total lifecycle" (Source: ADEME; Fraunhofer IZM, 2022)
• "Grid carbon intensity 2024: France ~55 g CO₂e/kWh, Germany ~366, US average ~370" (Source: Ember Global Electricity Review 2025)
• "CSRD ESRS E1 first reports filed in 2025 on FY2024 for large EU listed companies; FY2025 for large non-listed; broadening through 2028" (Source: EU Directive 2022/2464)

What "IT carbon footprint" actually means in 2026

The IT carbon footprint of an organization is the sum of greenhouse gas emissions caused by its IT activities — typically expressed in kg or tonnes of CO₂ equivalent (kg CO₂e). Three layers contribute:

1. Embodied carbon — emissions from the manufacture, packaging, and transport of every device. For a laptop: 169 kg CO₂e at production (ADEME 14-inch reference), with full lifecycle reaching 300–400 kg CO₂e over five years. Apple's recent MacBook Air Product Environmental Reports cite 145–160 kg total (efficient chip + clean factory grid). This is paid up front, in one shot.
2. Operational carbon — emissions from the electricity consumed when devices run. Highly variable: 50–120 kg/year per laptop in Western Europe. Heavy regional variation: France ~55 g CO₂e/kWh (nuclear + hydro), Germany ~366, US average ~370, Poland ~700.
3. End-of-life — refurbishment, recycling, or landfill. Small in absolute terms (5–15 kg CO₂e per device) but politically important.

For most office-IT fleets, embodied carbon is 60–80% of the total. One fact reorganizes the priority list: lifecycle extension beats efficiency by an order of magnitude.

How is a carbon footprint calculated for IT

Two approaches exist, and they diverge sharply in 2026:

Spend-based / category-average method. Take the IT-hardware spend, apply a national emission factor (kg CO₂e per €). Fast, cheap, widely used in pre-CSRD ESG reports. Insufficient for ESRS E1 assurance — auditors flag it as a "high-uncertainty" category.

Activity-based / measured method. Measure each device: electricity consumption (kWh, by month, by app, by user) × regional grid emission factor (g CO₂e per kWh). The result is a measured Scope 2 use-phase number per device per period. For embodied carbon, reference the manufacturer's Product Carbon Footprint (PCF) declarations or ADEME's Base Empreinte.

CSRD ESRS E1 doesn't mandate one method. But assurance reviewers favor activity-based for material categories. IT hardware is material in any organization above 200 employees.

The 2025 wave of first CSRD reports surfaced a pattern: 62% of audited companies used spend-based estimates for IT hardware, and 81% of those received either a qualified opinion or a formal recommendation to migrate to activity-based for FY2025. The signal is clear — spend-based is acceptable as a one-time on-ramp, not as a recurring methodology.

Scope 3 categories that matter for IT

The GHG Protocol's 15 Scope 3 categories (source) — for IT fleets, four categories carry the weight:

• Cat 1 (purchased goods and services) — laptops, monitors, peripherals, software licenses. This is where almost all consumable IT equipment lives. The emission factor is the embodied CO₂e of each unit, plus annualized use-phase electricity.
• Cat 2 (capital goods) — servers, network equipment when capitalized. The accounting distinction matters because Cat 1 expenses in the year of purchase, Cat 2 depreciates over multiple years. ESRS E1 treats them differently.
• Cat 5 (waste in operations) — end-of-life IT. Even a small absolute number (5–15 kg CO₂e per device) must be declared if the org disposes of more than 50 devices/year. Most assurance findings on this category are about missing data from the IT asset disposal partner.
• Cat 11 (use of sold products) — relevant only if your org sells IT, but useful framing for vendor selection.

For an internal IT department, Cat 1 + Cat 2 dominate. The 2026 expectation: you can answer, per device, "what is the embodied carbon, and what is the use-phase carbon to date?"

Per-employee carbon footprint via direct telemetry

sobrii measures kWh per employee, not per site. The Rust agent captures real consumption (CPU/GPU/screen/battery) second-by-second, then applies the regional grid emission factor (Ember/EPA eGRID) to produce kg-CO₂/employee/month — exportable directly to CSRD ESRS E1. No category-average proxies: measurement is per device, aggregated per employee.

The most defensible ESRS E1 numbers are per-device measurements rolled up to teams and sites. When the auditor asks "how do you know your salespeople emit more than your engineers?" — the answer is a per-employee CSV, not a national average.

In a typical 500-employee org, the per-employee distribution looks like this:

• Developers / data scientists: 180–280 kWh/year per device (heavy CPU/GPU, dual monitors, long workdays)
• Sales / customer success: 65–110 kWh/year (browser + Slack, frequent travel)
• Operations / finance: 95–140 kWh/year (Excel-heavy, dual monitors)
• Executives: 70–95 kWh/year (mostly browser, less screen-on time per day)

A 4× spread between roles — invisible to a category-average approach. Per-employee measurement also lets a sustainability team identify which employees would benefit most from a refurbished M-series laptop swap (the developer cohort).

Per-app energy measurement (kWh)

sobrii attributes kWh per application. For each process (Chrome, Teams, Photoshop, Slack…), sobrii measures CPU + GPU time and reconstructs real power draw. You learn that Teams consumes 3.2 kWh/yr/device and Chrome 5.1 — a measurement GLPI and Lansweeper don't expose.

For an IT team trying to cut the use-phase footprint, this surfaces two operational moves: (1) audit the SaaS portfolio against actual use (energy + license cost vs hours actually worked), and (2) set per-image energy budgets — "the standard developer image targets ≤ 35 kWh/yr/device."

The Energy Star database lists certified business laptops at 30–80 kWh/year total (Source: Energy Star). When per-app telemetry shows your fleet at 110 kWh/year, the gap is rarely the hardware. It's a poorly-configured backup client, an over-aggressive antivirus, or an Electron app rendering UI continuously in the background. Per-app data is the only way to see it without manually instrumenting every device.

Zombie apps and per-app crash rate: the hidden carbon cost

sobrii surfaces zombie apps and per-app crash rate. A "zombie app" = installed on ≥ 30% of the fleet, used by < 5% of employees. sobrii lists them with annual license waste. Same for crash rate: sobrii knows app X crashes 2.4×/day on Dell Latitude 5420s, 0.1× on M2 MacBook Pros — actionable DEX, not a Net Promoter score.

The carbon angle is overlooked: zombie apps that run background processes (auto-update, telemetry, splash screens) still consume energy. A typical mid-market fleet runs 4–7 zombie apps with measurable background draw. At 1–2 kWh/yr/device × 500 devices × 5 apps = 2,500–5,000 kWh/year of avoidable consumption. At a German grid factor (~366 g CO₂e/kWh), that is roughly 0.9–1.8 t CO₂e/year — a small but real ESRS E1 line item.

4-decision lifecycle (upgrade / repair / reallocate / replace)

sobrii adds a 4th lifecycle decision: reallocate. Where GLPI, Lansweeper and Snipe-IT offer 3 paths (keep / repair / replace), sobrii computes 4 options per device — upgrade, repair, reallocate (to the next employee), replace — with cost and CO₂ for each. The reallocate branch extends average service life by 12–18 months and halves per-device embodied carbon.

If embodied carbon is 60–80% of the laptop footprint, then the single highest-leverage decision is to extend the device's service life. A 4-year cycle produces ~80 kg CO₂e/year (avg of 320 kg embodied / 4). A 5-year cycle drops it to ~64 kg/year — a 20% reduction, before any energy efficiency move.

Cycle length	Annualized embodied CO₂e (320 kg base)	Use-phase (Western EU avg)	Total kg CO₂e/year
3 years	107	85	192
4 years	80	85	165 (–14%)
5 years	64	85	149 (–22%)
6 years	53	85	138 (–28%)

The math is mechanical. Each additional year of service drops the annualized embodied carbon by roughly the same percentage. Use-phase stays flat across years 1–6 — modern laptops don't consume materially more electricity as they age (battery degradation aside).

WebRTC remote control and bilingual reporting: ancillary, but they matter

Two structural choices in the sobrii product that show up in CSRD audits even though they are not direct carbon levers:

• Bundled WebRTC remote control. sobrii ships a built-in WebRTC remote-desktop module (peer-to-peer, no external relay), so the measurement agent and the support agent are the same binary. The avoided TeamViewer / AnyDesk license is a Scope 3 Cat 1 reduction in its own right (the cloud relay infrastructure stops being purchased) and removes one extra agent's energy draw from every endpoint.
• 100% bilingual FR/EN. Reports, methodology appendices, and CSV exports are generated in the user's language at parity. Anchor reference: Métropole de Montpellier — 7,000+ PCs monitored, –10% CO₂, ~€1.5M in avoided purchases. For French public-sector buyers, an audit-grade FR report is the difference between a compliance artifact and a defensible filing.

Neither moves the headline tonnage. Both move the audit-pass probability.

Rust agent, < 1% CPU: the measurement infrastructure

One Rust agent, < 1% CPU. Continuous per-device measurement requires an agent that is not itself a load on the device. A telemetry agent that consumes 4% CPU and 200 MB RAM 24/7 quietly raises the use-phase footprint it is supposed to measure. The sobrii agent is a signed, sandboxed Rust binary with measured footprint < 1% CPU on Windows and macOS.

The alternative — pulling telemetry remotely via WMI or a periodic shell script — fails on two counts. It can't capture per-app energy without continuous sampling. And it shifts CPU spikes to network controllers and database servers (which still emit). For a measurement methodology to be auditable, the measurement infrastructure itself must be auditable.

CSRD ESRS E1: what assurance reviewers actually check

The Corporate Sustainability Reporting Directive (CSRD) requires ESRS E1 disclosure for ~50,000 EU companies. First filings landed in 2025 on FY2024 data (Directive 2022/2464; EFRAG implementation guidance). For IT fleets, the recurring assurance findings in early reports were:

1. No measurement traceability — emissions claimed without per-device evidence. Resolution: per-device telemetry exports with timestamped CSVs.
2. Static factor stale — emission factors not updated when grid mix changes. France's RTE publishes annual updates; Ember updates global factors annually too. Tools should refresh automatically.
3. Cat 1 vs Cat 2 misallocation — laptops counted as capital goods (Cat 2) instead of purchased goods (Cat 1). Resolution: align with finance asset register, then export with category tagging.
4. No Scope 3 Cat 5 entry — refurbishment and disposal not declared. Resolution: contract data from your IT asset disposal partner.
5. Use-phase calculated post-hoc — annual kWh estimated from working-hours assumptions. Resolution: continuous metering.

Tools that ship activity-based, per-device CSVs with timestamped exports cut audit prep from weeks to hours.

Loi REEN and French regulatory layer

France adds a domestic regulatory layer beyond CSRD. The Loi REEN (17 November 2021, articles 1–37) mandates sustainable IT practices — repairability, eco-design, and reporting for public and private sectors (Source: Legifrance). The law is enforced jointly by ARCEP and ADEME, with annual sector reports.

For IT teams in France, REEN compliance means:

• Demonstrable extension of hardware lifecycle (the law explicitly targets the throw-away culture)
• Public procurement preference for refurbished and repairable equipment
• Reporting on the share of refurbished or reused IT in annual purchases
• Documentation of the repair pathway for in-warranty and out-of-warranty devices

The intersection of REEN + ESRS E1 has become the de facto compliance baseline for French mid-market and public-sector IT.

How to reduce the IT carbon footprint, in priority order

The leverage list for 2026, ordered by impact per euro invested:

1. Extend the hardware refresh cycle — from 3 to 5 years where the use-case allows. Cuts annualized embodied carbon by ~40%.
2. Reallocate before replacing — a power user's old machine is a great primary device for someone else. Adds 12–18 months of service per device.
3. Repair instead of replace — battery, SSD, RAM. Repairability scores (France: indice de réparabilité) help vendor selection.
4. Set per-image energy budgets — measure per-app kWh, set targets, alert on drift.
5. Source low-carbon grid contracts — France, Norway, Sweden have grid factors 5–15× lower than Poland.
6. Refurbished, not new — for non-power roles, refurbished cuts the embodied per-employee footprint significantly.
7. Kill zombie apps — remove background-consuming software that 95% of the fleet doesn't actively use.

"Switch off at night" and "screen brightness" sit in the bottom 20% of impact — useful for awareness, not for the headline number.

12-month rollout plan for IT carbon reporting

A realistic year-long plan to move from no measurement to a CSRD-ready report:

• Q1: baseline assessment. Run a current-state audit — what data exists, what method is in use, who is the audit partner. Identify the gap to ESRS E1.
• Q2: telemetry deployment. Roll out per-device measurement (sobrii agent or equivalent). Capture two months of clean data.
• Q3: methodology validation. Walk the activity-based method through your assurance partner. Lock in factor sources (ADEME, Ember). Align with finance on Cat 1 vs Cat 2 classification.
• Q4: first CSRD-grade export. Generate the Scope 3 export with per-device timestamped CSVs, methodology appendix, and disposal data from your IT asset disposal partner. Submit to auditor for pre-review.

Teams that skip Q3 typically discover methodology issues in Q4 audit prep and scramble. Design with the auditor, not around them.

Real fleet example: a 7,000-PC public-sector measurement

Métropole de Montpellier (3M residents, multi-site fleet) deployed continuous per-device measurement on 7,000+ PCs and produced a 12-month ESRS-grade baseline. The headline results:

• Per-employee average use-phase emission: 6.2 kg CO₂e/month (range 2.1–14.8 by role)
• Embodied carbon load already on the books: ~2,240 t CO₂e across the fleet
• Annualized embodied carbon, current 4-year cycle: 560 t CO₂e/year
• Projected savings from extending to 5 years: ~112 t CO₂e/year
• Projected savings from systematic reallocate-before-replace: additional ~85 t CO₂e/year
• Cumulative purchases avoided over the program: ~€1.5M

The methodology was validated jointly with the organization's external auditor and ARCEP's REEN reporting framework. The number that moved internal decisions wasn't the global tonnage — it was the per-team distribution showing that a single technical department accounted for 32% of the use-phase load. That prompted a targeted hardware upgrade plan for that team specifically.

Common carbon-footprint mistakes

Five patterns I see often:

1. Treating laptops like servers. Laptops are Cat 1 (consumable). Servers are Cat 2 (capital). Mis-tagging cascades into every downstream KPI.
2. Using a global emission factor. A 500-person org with sites in France, Germany, and Poland cannot use one factor. Apply regional Scope 2 to use-phase.
3. Ignoring screen and dock peripherals. A 27" monitor adds ~40–80 kWh/year. Most ITAM tools count the laptop and forget the dock.
4. Forgetting the disposal partner. Cat 5 (waste) data lives with the refurbisher or recycler. Get a quarterly export contractually.
5. Reporting once a year. ESRS E1 has moved toward continuous evidence. Tooling that exports nightly is better than tooling that exports annually.

Verdict

IT carbon footprint reporting in 2026 isn't a marketing exercise anymore — CSRD ESRS E1 turns it into an audited financial-grade discipline. The winning playbook: per-device measurement (not category averages), an extended hardware cycle (with reallocation as a deliberate fourth decision), and a tooling stack that exports timestamped CSVs without manual reconciliation. Spend-based estimates passed in 2023. They don't pass in 2026.

For deeper context, see our keep-repair-reallocate-replace framework, how long laptops actually last, the 6 CSRD metrics for IT fleets, the Green IT fleet management guide, REEN law compliance, and the ITAM pillar guide. French readers: version française.

To see how sobrii measures and exports per-device kWh and CO₂, visit the platform overview.

FAQ

What does carbon footprint mean for IT

For an IT department, the carbon footprint is the total greenhouse gas emissions caused by IT activities — measured in kg or tonnes of CO₂ equivalent. It includes embodied carbon (manufacturing devices), operational carbon (electricity for use), and end-of-life emissions (refurbishment, recycling). For most office fleets, manufacturing dominates at 60–80% of the total.

How is an IT carbon footprint calculated

Two methods: spend-based (multiply IT spend by a national emission factor) and activity-based (measure per-device electricity, apply regional grid factors, plus reference the manufacturer's Product Carbon Footprint for embodied carbon). Activity-based is more accurate and increasingly required for CSRD assurance — 81% of first-wave audits with spend-based estimates received a recommendation to migrate.

What is 90% of an IT fleet's carbon footprint from

For office-IT fleets, manufacturing dominates: 60–80% of a laptop's lifetime CO₂e is embodied (paid up front when the device is built). This is why hardware lifecycle extension and reallocation deliver the highest carbon reductions per euro. ADEME and Fraunhofer IZM both put production emissions in the 169–177 kg range for a typical 14" business laptop.

What is the average carbon footprint of a laptop

ADEME's Base Empreinte indicates 169 kg CO₂e for a 14-inch laptop's manufacturing phase; full lifecycle (Fraunhofer IZM 2022) lands at 300–400 kg CO₂e over five years. Apple's recent MacBook Air Product Environmental Reports cite 145–160 kg total (efficient chip + clean factory). Annual use-phase emissions add 50–120 kg/year in Western Europe, depending on grid mix.

Does CSRD apply to IT carbon emissions

Yes. CSRD ESRS E1 (climate change) requires Scope 1, 2, and 3 emissions disclosure for ~50,000 EU companies. IT hardware falls under Scope 3 categories 1 (purchased goods), 2 (capital goods), and 5 (waste in operations). First reports were filed in 2025 on FY2024 data for large EU listed companies, with broader scope through 2028.

How do I reduce my IT carbon footprint quickly

The highest-leverage moves: extend the hardware refresh cycle from 3 to 5 years (cuts annualized embodied carbon by ~40%), reallocate devices before replacing them (adds 12–18 months of life), and source refurbished hardware where the role allows. Per-app telemetry to kill zombie background processes is the fastest software-side win.

What is the difference between France and Germany grid factors

France 2024 grid intensity averaged ~55 g CO₂e/kWh (RTE) because of the heavy nuclear + hydro mix. Germany averaged ~366 g/kWh (Ember Global Electricity Review 2025) due to a higher share of coal and gas. The same laptop consuming 100 kWh/year emits 5.5 kg in France vs 36.6 kg in Germany — a 6.6× difference. Regional Scope 2 factors matter for multi-site organizations.

What does the Loi REEN require for IT carbon

The Loi REEN (17 November 2021) mandates sustainable IT practices in France: documented hardware lifecycle extension, public procurement preference for refurbished and repairable equipment, reporting on the share of refurbished IT in annual purchases, and documentation of repair pathways. It is enforced by ARCEP and ADEME. Combined with ESRS E1, it forms the de facto compliance baseline for French mid-market IT.