Quarterly · Global · Public sources only
A standing watch on life-cycle assessment across the chip industry — what the largest manufacturers actually publish, what stays inside the fab fence, and which numbers we are asking for next. Built entirely from public disclosures, with every source linked.
Corporate climate reporting in this industry is genuinely good by global standards. Every leading manufacturer publishes annual Scope 1 and Scope 2 emissions, most have them independently assured, and several carry validated science-based targets. That is not the gap.
The gap is product. Life-cycle assessment asks a different question from corporate accounting: not how much did this company emit last year, but what does one wafer, one die, one packaged part actually cost the planet, from silica sand to end of life. Almost nobody answers that in public — not per node, not per process, not with a stated functional unit. Customers designing a phone, a car, or a data centre are therefore modelling the single most energy-intensive step of their supply chain from generic databases and educated guesses.
That matters more each quarter. Leading-edge nodes consume more energy per wafer than the nodes they replace, advanced packaging adds steps rather than removing them, and AI demand is scaling capacity faster than grids decarbonise. Corporate totals can improve through renewable procurement while the footprint embodied in each individual part climbs. Only life-cycle data separates those two stories.
We are not asking anyone to emit less this quarter. We are asking them to publish the number per wafer.
Selected by manufacturing scale, technological leadership, and influence over what the rest of the chain is asked to disclose. Reviewed and re-scoped each quarter.
6 companies · sets the energy intensity of every advanced node
7 companies · highest wafer volumes, deepest process-gas use
8 companies · the embodied carbon nobody counts
5 companies · the demand signal that unlocks supplier data
Read this as prevailing practice within each tier, not as a grade for any one company. Individual leaders sit ahead of their tier; laggards sit behind it. Company-level scoring begins in Issue 02.
| Disclosure item | Tier A Foundry |
Tier B Memory / IDM |
Tier C Equipment |
Tier D Fabless |
|---|---|---|---|---|
| 01Scope 1 and Scope 2, market-based | ● | ● | ● | ● |
| 02Scope 3 itemised across all 15 categories | ◐ | ◐ | ◐ | ◐ |
| 03Independent assurance of GHG figures | ● | ● | ● | ◐ |
| 04Fluorinated gases by species, with abatement efficiency | ◐ | ◐ | ○ | ○ |
| 05Electricity volume and renewable instrument type, by region | ◐ | ◐ | ◐ | ◐ |
| 06Water withdrawal, discharge quality and recycling rate | ● | ● | ◐ | ○ |
| 07Product carbon footprint of finished goods | ○ | ○ | ◐ | ● |
| 08Footprint per wafer, per die, or per technology node | ○ | ○ | ○ | ○ |
| 09Full ISO 14040/14044 study, publicly downloadable | ○ | ○ | ○ | ◐ |
| 10Machine-readable dataset rather than PDF only | ○ | ○ | ○ | ◐ |
Any credible semiconductor LCA has to get these six right. Each one is a place where a reasonable-looking methodological choice can move the published result by a wide margin.
Fabs never stop. Lithography, plasma etch, vacuum systems and cleanroom air handling run continuously, and for most manufacturers grid electricity is the largest single line in the inventory. The quality of the renewable claim behind it — long-term power purchase agreements versus unbundled certificates — changes the reported number more than almost any process improvement.
Perfluorocarbons, NF₃ and SF₆ used in etch and chamber cleaning carry global warming potentials in the thousands to tens of thousands over a hundred-year horizon. What matters is not consumption volume alone but destruction and removal efficiency of the abatement fitted to each tool class — which is rarely broken out.
A large 300 mm fab draws water on the scale of a small city, and purification, recycling and wastewater treatment each carry their own energy and chemical burden. Water is well reported in volume terms; the embodied energy of treating it usually is not.
Polysilicon and finished wafers, photoresists and specialty chemicals, rare gases such as neon and helium, substrates and packaging metals. Primary supplier data is thin, so most inventories fall back on generic databases — introducing uncertainty precisely where geography and energy mix matter most.
An advanced lithography or deposition tool is a heavy, materially complex object with a long service life and a large power draw. Fab capital equipment is one of the most consistently under-reported categories in the sector, and it is growing as capacity expands.
For logic destined for data centres, lifetime electricity can dwarf manufacturing; for a low-power mobile part it usually does not. Because the comparison depends entirely on assumed lifetime, duty cycle and grid intensity, the assumptions must be published alongside the result — or the result means nothing.
Ordered by how much each one would improve the comparability of semiconductor life-cycle data across the whole industry, not by how easy it is to produce.
With the functional unit, system boundary, reference year and reference fab stated. A single defensible figure per node beats a decade of aggregate corporate totals.
Consumption alone cannot be converted into an emission. Publish assumed and measured destruction and removal efficiency, and say which GWP set was applied.
Contract type, vintage and geography of every renewable claim. A certificate purchased in another market does not decarbonise the wafer.
Stop folding tools, construction and logistics into a single "other" line during the largest fab build-out in the industry's history.
How fab-wide energy, gas and water are assigned to a specific product, mask layer or customer. Without this, no two published footprints are comparable.
Per site, per year, per indicator, in a structured file. Every analyst currently retypes the same tables by hand, and every retyping introduces error.
Embodied carbon of manufacture, lifetime power and gas draw in operation, serviceability and end-of-life recovery. This closes the largest open category in every fab's inventory.
A shared functional unit and boundary, developed through industry bodies and major customers together — instead of twenty-six private reporting templates that cannot be added up.
Everything on this page is assembled from documents any reader can open. We do not interview, we do not receive privileged data, and we do not model a value where a company has already published one. Where a figure is unavailable, we record the absence rather than substituting an estimate — the gaps are the finding.
Each quarter we re-read the current sustainability, ESG and annual reports of the watchlist, log what changed, and re-run the matrix in § 03. Corrections are published in the open: if a company shows us that a mark is wrong, we change it in the next issue and say so.
Next quarter, Issue 02 focuses on: capital goods under Scope 3 category 2 as fab construction accelerates; fluorinated-gas abatement disclosure at tool level; and the first wave of European sustainability reports filed under ESRS E1 by semiconductor firms, to see whether regulated reporting delivers product-level granularity that voluntary reporting has not.
Updated quarterly · corrections published in the openLinks resolve to each publisher's official portal, where the current edition of the cited document is hosted. We name the document rather than deep-linking a file, so the reference stays valid as each new annual edition replaces the last.