Meta's gross emissions rose from 15.6 million to more than 21.7 million tonnes in a year. Its net emissions rose too, from 8.1 million to more than 11.8 million, even after offsets and removals. Headlines say Meta is "re-evaluating" its 2030 net-zero goal, but they don't tell you what to do if Meta, or any other vendor, runs part of your stack. The useful part is in the tables, and you can find the five numbers that matter in about 15 minutes.
This guide uses Meta's 2026 report as a worked example. One caveat first: every Meta figure below comes from Ars Technica's coverage of the report, not from the report itself. Check the primary document before you quote any of it in a procurement file.
The worked example: Meta, 2024 vs 2025
According to Ars Technica, Meta's report covers 2024 and 2025. The percentage changes below are calculated from the rounded figures in the article, so they can differ slightly from the percentages in the report.
| Metric | 2024 | 2025 | Change | |---|---|---|---| | Gross emissions (tonnes CO2e) | 15.6 million | 21.7 million+ | about +39% | | Net emissions (tonnes CO2e) | 8.1 million+ | 11.8 million+ | about +46% (article says 45%) | | Facility water consumption (gallons) | 825 million | 790 million | about -4% | | Water use in water-stressed areas (gallons) | 198 million | 246 million | about +24% | | Water tied to purchased electricity (gallons) | 19 billion | 26 billion+ | at least +37% |
The "+" marks mean the article says "more than." Treat the percentages as approximate.
Step 1: Find gross emissions and check the direction
Gross emissions are the total greenhouse gases a company causes before any offsets, removals or renewable-energy accounting. Start here, because it is harder to dress up than the other figures. Meta's rose from 15.6 million to more than 21.7 million tonnes.
The article says that is the biggest year-over-year jump among the four large US tech companies it compares. It lists Microsoft at 25 percent, Google at 18 percent and Amazon at 16 percent. It also says Meta likely still has one of the smallest total footprints of the group. A big percentage on a smaller base is still a strong signal, but it doesn't make Meta the biggest emitter.
Step 2: Compare net to gross, then look at the gap
Net emissions are what remains after adjustments. The article says Meta's approach includes building renewable energy projects to match emissions and using carbon removal. It doesn't spell out how Meta calculates the net figure, so read the methodology section in the report itself.
Here is the part headlines miss. Subtract the article's rounded numbers and the gap between gross and net grew from about 7.5 million tonnes to about 9.9 million tonnes. The adjustments got bigger, yet net still rose from 8.1 million to 11.8 million. Growth outran the offsetting, which is why both figures climbed. That is an inference from rounded figures, not a number Meta states. See related: can you be fired for union activity in the uk? explained for additional background.
What can the gap hide? In general terms, renewable matching usually means buying clean power on paper over a year, which doesn't guarantee clean power at the hour your workload runs. Removals and credits vary in quality. A company that leads with net or "100% renewable" and buries gross is telling you which number it prefers.
Step 3: Find the capital goods line in Scope 3
Scope 3 covers emissions outside a company's own operations, including its supply chain. "Capital goods" is the category for big purchased equipment. For a data center operator, that means servers, networking gear and, in many accounting setups, construction.
The article says Meta attributes much of its increase to capital goods such as servers. Those emissions are mostly baked in when the hardware is made, before it draws a watt. So a company can buy renewable electricity and still see emissions climb if it buys hardware faster. That is why AI buildouts hit the gross number so hard: more servers per year means more embodied carbon per year.
This matters to you as a buyer because efficiency claims about running a workload don't cover this line. Ask whether the vendor reports capital goods separately and whether it changed how it calculates them.
Step 4: Read three water numbers, not one
Meta's report separates facility water, water in stressed areas and water from purchased electricity. Facility consumption fell from 825 million to 790 million gallons. Stressed-area use rose from 198 million to 246 million gallons.
Then there is the big one. Water consumed by the power plants supplying Meta's electricity went from 19 billion to more than 26 billion gallons. Dividing the rounded figures, that is roughly 33 times the facility figure in 2025, up from about 23 times in 2024. The article's reading is that the larger water footprint comes from the power supply, not from cooling on site.
It also quotes Data Center Dynamics saying "Meta is the only hyperscaler to report indirect water use, a growing concern as data centers shift to closed-loop cooling systems that, despite using less water at source, require more electricity to run." That counts in Meta's favor on transparency, and it is a warning for you. If a vendor reports only on-site water, you are seeing the smaller number.
Step 5: Check whether the targets moved
The article quotes Meta's report saying it is re-evaluating "timelines, sequencing, and interim milestones," including the 2030 net-zero date and near-term intensity targets. It adds that Amazon, Google and Microsoft have kept their original goals so far. Ars Technica doesn't say whether they will follow Meta.
Look for the same signals in any report: a date quietly replaced by "long-term ambition," a target switching from absolute to intensity (emissions per unit of something), or a base year that changed. Intensity numbers can improve while total emissions rise, so never accept them alone.
The five-line checklist
Print this and keep it next to the report:
- Gross emissions, this year vs last year: up or down, and by what percent?
- Net emissions and the gap to gross: what closes it, and is it matching, removals or credits?
- Scope 3 capital goods: is it broken out, and does it explain the change?
- Water: facility, stressed-area and purchased-electricity figures, all three present?
- Targets: did any date, milestone or metric type change since the last report?
Which numbers should affect a purchase
Give weight to the gross trend, the methodology notes, and whether the vendor discloses indirect water and capital goods. Those are hard to dress up, and they show what the company owns up to. Give little weight to a headline "carbon neutral" label, a single intensity figure, or a future target with no interim milestones.
For most buyers, climate data should break ties and shape contract questions, not override price, security or reliability. It deserves more weight if you have your own reporting duties or customers who ask about your supply chain.
A vendor-questions template
Send this to your cloud or SaaS account rep:
We are reviewing our suppliers' climate disclosures. From your latest sustainability report, please confirm: (1) your gross emissions for the last two years, (2) your net emissions and what adjustments separate them, (3) your Scope 3 capital goods figure, (4) your direct, stressed-area and purchased-electricity water use, and (5) any changes to your targets or timelines since your last report. If any of these are not published, please tell us when they will be.
What this checklist cannot do
It cannot verify a company's data. You are reading what the vendor chose to report, and an audit or assurance statement, if one exists, is the only check on it. It also cannot fairly compare reports that use different methods, boundaries or base years, so two vendors' numbers may not line up even when the labels match. And the Meta figures here are second-hand from one article, so confirm them in the primary report.
Your next action
Pull the latest sustainability report from your main cloud or SaaS vendor today. Fill in the five checklist lines and the 2024 vs 2025 table with its numbers, then send the questions template for anything missing. Expect gaps and slow answers. A vendor that can't supply gross emissions or water figures has told you something already.



