Dear Cloudflare Impact team,
I host a small static website on Cloudflare Pages. I have been looking into the environmental impact of my site, and as a result I have ended up with a number of questions about you as my website host.
I have read the 2025 Impact Report, in the revision published on 28 January 2026, and the 2024 Emissions Inventory dated 15 July 2025. There were several things I found useful. The new Scope 3 reporting, the first climate-related financial risk disclosures, and the decision to report network and office electricity separately, and the detail given on twelfth-generation server efficiency. The questions below are a result of where I could not work the answers from the published material.
31,000 tonnes of historical emissions
The report describes the completion of your commitment to account for emissions from powering the network between launch and your first renewable-energy purchase in 2018. Approximately 31,000 tonnes of CO₂e were covered by five projects: Pacajai REDD+ in Brazil, Boone Forestry Improved Forest Management in Kentucky, BioLite Improved Stove Programme I and II in Uganda, and Clinton Landfill Gas Collection and Combustion in Illinois.
The commitment is worded as offsetting or removing. Looking at the five project types, they seem to be avoidance, protection or reduction rather than durable removal, although I know that improved forest management can produce removal credits as well as avoided emissions. How much of the 31,000 tonnes do you count as removal, and how much as avoided or reduced?
If possible, I would like to see the standards, methodologies and registries under which the credits were issued, along with serial numbers and confirmation that all of them have been retired. I would also like to see any available project-level quantities, since the split across the five projects is not clear. I would also like to know what work was done on additionality, permanence and double counting, particularly for the REDD+ and forestry elements.
Was the portfolio reviewed by anyone independent of Cloudflare and 3Degrees? The Emissions Inventory says that Shift Advantage’s review did not extend to your renewable-energy purchases, and the Impact Report says that no section other than the emissions disclosure was externally verified.
Renewable energy matching
The 2024 inventory puts location-based Scope 2 emissions at 62,782 tonnes CO₂e, of which 61,171 tonnes, or 97%, come from the network. Market-based Scope 2 is reported as zero, achieved by matching consumption with energy attribute certificates bought through 3Degrees under RE100’s technical and geographic criteria, with Verra-certified offsets used in certain island nations and for Scope 1.
I would like to know how that matching divides between long-term power-purchase agreements or direct renewable tariffs on one side, and unbundled certificates and offsets on the other. It makes a difference to how much additional clean generation a customer’s traffic can fairly be said to support.
Is there any intention to move towards hourly matching, or at least to something more granular than annual? I realise this is harder for a network spread across 125 countries than for a company running a few large campuses, and I would be interested in your own view of what is achievable.
The two energy figures
The GRI 302-1 disclosure gives total 2024 energy consumption as 182.89 GWh. The SASB disclosure, TC-SI-130a.1, gives 177.89 GWh. Both sit in the same appendix.
The energy intensity you report, 0.000110 MWh per dollar of revenue, is consistent with 182.89 GWh against 2024 revenue of $1,669.6 million; 177.89 GWh would come out at roughly 0.000107. I suspect the GRI figure is the intended one, but I would rather ask than assume. Which of the two figures should a reader rely on?
The footnote to the emissions table refers to 2023 verified offsets and renewable-energy purchases in the Emissions Inventory, and to 2024 in the Impact Report. I assume 2024 is right?
The reporting boundary
Energy data is recorded monthly for equipment under your operational control across more than 330 cities, but electricity used to run the data-centre facilities housing your points of presence is excluded where those facilities are not owned or controlled by Cloudflare, as is other equipment co-located in them. I understand this follows guidance from BSR and the Future of Internet Power, and that it is a defensible boundary under the operational control approach.
The effect, though, is that cooling, lighting and power conversion at the facilities where 97% of your electricity is consumed fall outside the inventory altogether. Under the GHG Protocol they would usually sit in Scope 3 Category 8, upstream leased assets, and the 2024 inventory covers Categories 1, 2, 4 and 5 only.
Do you collect Power Usage Effectiveness figures from your colocation partners, even though the electricity itself falls outside the boundary? If you do, will you publish them? Where facility-level disclosure is commercially impossible, regional ranges or a network-wide, energy-weighted average would still tell customers a great deal. Is there a timetable for adding Category 8, so that the excluded overheads appear somewhere in the accounts?
Water
The report treats water as immaterial and discusses consumption at your leased offices. But the boundary that excludes facility electricity also excludes the cooling water used at the facilities housing your equipment, and that water is consumed on your behalf.
Do you collect Water Usage Effectiveness or other cooling data from your data-centre partners? A qualitative account, or some indication of how much of the network sits in water-stressed regions, would be useful.
Efficiency and total consumption
The Gen 12 figures show 84% more requests per second per kilowatt than Gen 11, and over 2,700 legacy servers replaced during 2025 by 1,661 newer ones, using nearly 40% less hardware while carrying more traffic.
Requests per second per kilowatt is a ratio, though, and a network can improve on it while its total consumption climbs. Yours has gone from 63.18 GWh in 2022 to 182.89 GWh in 2024. Would you consider publishing an absolute measure alongside the ratio, such as requests served or data transferred per unit of electricity, together with total network consumption? That would let customers tell efficiency gains apart from growth.
The report also credits the smaller fleet with reduced embodied carbon, transport and electronic waste. Was any of that quantified? The 40% is a count of units, and the embodied emissions of building 1,661 servers and retiring 2,700 are not obviously captured by the average-spend method used for Category 2 capital goods. If a fuller calculation exists, it would be worth publishing. How many of the retired servers were recycled?
Assurance
What accompanies the emissions data is Shift Advantage’s stakeholder letter, which states that it is not the assurance report, is not a substitute for one, and is provided solely for Cloudflare’s benefit. It contains no opinion or conclusion. Would you publish the ISO 14064-3 assurance report itself, or a summary of it including the opinion and any qualifications?
Information on independence would be welcome. Shift Advantage describes itself as having provided consulting and advisory services to support the calculation of your 2024 emissions; it then verified that calculation; and it was separately engaged for the climate scenario analysis behind the new TCFD disclosures. As the programme matures, there is a case for a verifier with no advisory relationship, and I would be interested to know whether you have considered this.
Targets
GRI 302-4 says that Cloudflare has applied to join the Science Based Targets initiative and has begun developing carbon reduction targets. Three other places in the same report (the sustainability commitments page, GRI 305-5 and the TCFD metrics section) say that Cloudflare has committed to setting near-term targets with the SBTi, which is also how the 2023 ESG Index put it.
Could you clarify where things actually stand with the SBTi, and whether there is a date by which targets will be submitted for validation? I would like to know whether there will be absolute reductions across Scopes 1, 2 and 3, and how growth in the network is handled within them.
Independent certification
You report against GRI, SASB and TCFD, participate in the UN Global Compact and the Global Network Initiative, undergo independent GNI assessments and obtain limited assurance for your emissions. The GNI assessment is more or less the kind of independent scrutiny I have in mind, applied to human rights rather than to environmental performance.
Have you considered an equivalent whole-company assessment covering environment, transparency, governance, workers and communities, whether B Corp certification or something comparable? If you have looked at it and decided against, I would find it useful to know what form of independent accountability you think better suits a company of your kind.
Why I am asking
What prompted this email is that your disclosure is good enough for the remaining gaps to be visible.
Furthermore, the International Energy Agency expects global data-centre electricity consumption to roughly double, from 485 TWh in 2025 to around 950 TWh in 2030, or about 3% of global electricity demand (Key Questions on Energy and AI, 2026). In the United States, Lawrence Berkeley National Laboratory found for the Department of Energy that data centres used approximately 176 TWh, 4.4% of national electricity, in 2023, and projected between 6.7% and 12% by 2028 (2024 United States Data Center Energy Usage Report).
Water and the gaps in reporting are garnering the same attention. Writing in AGU Advances, Privette, Barros and Cai argue that without standardised facility-level reporting, regulators, water managers and communities cannot properly assess data centres’ cooling demand, the water consumed in generating their electricity, or their supply-chain impacts. Here in the UK, an oral question in the House of Lords on 8 July 2025 drew the response that the Environment Agency is working to improve its understanding of the sector’s water and resilience needs, and the Minister confirmed that the AI Energy Council’s sustainability working group is considering whether new metrics should sit alongside PUE. My questions about PUE and Water Usage Effectiveness are much the same ones the Government has started to ask.
I intend to publish this letter on my website and to encourage other site owners to put similar questions to their own providers. It seems to me that customers ought to check environmental claims rather than take them on trust.
I would be grateful for a response, or for a pointer to anything I have missed.
Kind regards,
Amy Kitcher