F-GHG Abatement Programme · Buyer-side field notes
A phone, a laptop, a rack of GPUs. Each one carries a few grams of gas that was released four tiers upstream, in a fab you have never visited, and that will still be in the atmosphere long after every company named on this page is gone.
The physics you are buying
GWP-100, IPCC AR6. Atmospheric lifetime ≈ 50,000 years. The hardest gas to destroy — and the one most often hidden inside an averaged abatement figure.
GWP-100, AR6. Legacy in-situ chamber cleans. Substitution to remote NF₃ plasma clean is a source-reduction lever, not an end-of-pipe one.
GWP-100, AR6. The workhorse of large-area display PECVD cleaning and advanced logic. High utilisation, but volume is growing fast.
GWP-100, AR6. Deep silicon etch and MEMS. Also the gas regulators already know how to name — which makes it a useful wedge.
Values as published in IPCC AR6 WG1, Chapter 7, Table 7.15 → ref. 01. If your supplier still reports under AR4 or AR5, the same tonnage produces a materially smaller number. Ask which vintage they used.
Why the buyer
A fab will not spend capex on a scrubber to please a questionnaire. It will spend it to keep a qualified socket in a product that ships fifty million units.
Abatement is a purchase order problem long before it is an engineering problem.
Fluorinated greenhouse gases — CF₄, C₂F₆, CHF₃, NF₃, SF₆ and their cousins — are used to etch features and to clean deposition chambers. There is no substitute chemistry for most of these steps today. What there is, is a stack of choices: use less, use gases that break down more easily, capture and destroy what escapes, and prove it.
For a leading-edge fab or a Gen-10 display line, these gases are frequently the largest single line in Scope 1. For the brand that buys the die or the panel, they sit in Scope 3 Category 1 — accounted for, if at all, with an industry-average emission factor that is blind to whether that particular fab has abatement running or bypassed.
That blindness is the whole problem. An average factor cannot tell a good supplier from a bad one, so it cannot reward one. Until the number in the product carbon footprint changes when the supplier changes, no procurement decision will ever move a molecule.
The good news: this is one of the rare climate levers where the technology is mature, the cost is bounded, the abatement is measurable at the stack, and the emissions are concentrated in a few dozen sites owned by a few dozen companies. It is a solvable, addressable, contractable problem. It is just slow — and slow is where most programmes quietly die.
Reading the field · What the majors actually publish
A survey of public disclosure by the largest downstream buyers, read for one question only: does the published commitment reach a named upstream site, with verified abatement performance, and does it land in a product-level footprint? Read each company's current report before quoting — these programmes change every reporting cycle.
Apple 2030 and the annual Environmental Progress Report describe a direct requirement on semiconductor and display suppliers to install and operate F-GHG abatement, with third-party verification of supplier abatement performance, plus a fully public supplier list.
Supplier Code of Conduct and Supplier Responsibility Standards — abatement stated as a requirement of doing business, not a preference.
Site-level effective abatement rate by gas, and how bypass and downtime hours are treated in the verified figure.
Carbon negative by 2030 with detailed Scope 3 accounting; supplier requirements moving from disclosure to performance, including carbon-free electricity obligations for high-impact suppliers.
Supplier Code of Conduct plus procurement terms; heavy reliance on CDP-based reporting for supplier data.
Whether the silicon in Azure hardware is traced to fabs, and whether process gases are broken out from purchased electricity in supplier requirements.
Annual Environmental Report with growing candour on supply-chain and AI-driven emissions growth; supplier engagement and clean-energy expectations for hardware manufacturing partners.
Supplier Code of Conduct; TPU/server hardware sourcing leverage concentrated in a small number of foundry and packaging partners.
An explicit F-GHG clause for advanced-node and advanced-packaging capacity, given the compute build-out.
Product carbon footprint datasheets for a wide range of SKUs, plus supplier engagement and validated emissions targets in the ESG reporting suite.
Supplier principles and RBA-aligned requirements; PCFs give a natural place for supplier-specific data to land.
What share of a published PCF is built on supplier-specific factors rather than database averages — and whether the F-GHG portion is visible at all.
Long-running Sustainable Impact reporting, product carbon footprint disclosure across the portfolio, and public supplier lists with emissions-reduction programmes.
Supplier Code of Conduct and supplier sustainability scorecards feeding into sourcing reviews.
Whether scorecard weight is high enough to change award decisions, and whether F-GHG is a scored line or folded into a general carbon score.
Net-zero target validated by SBTi, ESG reporting with supplier engagement programmes and product-level carbon disclosure for selected lines.
Supplier requirements documentation; large ODM footprint means influence often runs through the assembler, not the fab.
How requirements travel past tier 1 ODMs to the component fabs that actually emit.
Fab operators generally disclose Scope 1 including process gases and describe point-of-use abatement coverage, local scrubber installation rates and intensity targets in annual ESG reports.
None from a brand's side by default — this data is voluntary and aggregated to company or country level.
Disaggregation: this fab, this quarter, this gas, this destruction efficiency, this assurance opinion.
Panel makers disclose far less granularity than logic and memory. NF₃ consumption scales with glass area and layer count, and panel sourcing is often deep behind a module house.
Display is usually specified by the brand at the panel level — which means the leverage exists even where the disclosure does not.
NF₃ kg per m² of glass, utilisation rate in the chamber, and abatement on the cleaning exhaust specifically.
The ladder · six rungs, in order
Do not attempt rung four before rung one. Every programme that stalls stalled because someone asked for verified destruction efficiency from a supplier they could not yet name. Each rung below has an action, a piece of evidence to collect, a number to measure, and an honest note on where it gets stuck.
Rank suppliers by F-GHG exposure, not by spend. For your top part numbers in logic, memory, analog, image sensor and display, identify the fab or line where the die or panel was actually made — not the packaging house, not the module assembler.
Where it sticks: die source is treated as confidential. Move the request out of the sustainability survey and into the NDA, the qualification package and the process-change-notification flow, where fab identity is already disclosed for technical reasons.
Named site list; three years of Scope 1 split gas-by-gas (CF₄, C₂F₆, CHF₃, NF₃, SF₆, HFCs); the IPCC tier used.
Measure% of F-GHG-relevant component spend traced to a named site. First milestone: 80% within two quarters.
Amend the supplier code and master agreement: point-of-use abatement on all F-GHG-emitting process chambers, a minimum destruction and removal efficiency by gas, an uptime commitment, data rights, and audit rights. For new capacity, make abatement a condition of qualification for the node or generation — capex is cheapest before the tool is installed.
Where it sticks: legal cycles and supplier pushback on audit rights. Start with new business awards, where the leverage is highest and the incumbent has nothing to defend.
Executed clause; abatement capex and installation schedule per site.
Measure% of traced sites under a signed F-GHG clause; % of newly qualified lines with abatement designed in from day one.
An installed scrubber that is bypassed during a tool excursion abates nothing. Ask for destruction and removal efficiency by gas — CF₄ separately, always, because it is the hardest to crack and the easiest to bury in a blended average — plus abatement uptime, bypass hours and the fuel or power the abatement itself consumes.
Where it sticks: most sites hold this data at tool level and have never had to roll it up. Give them the reporting template; do not ask them to invent one.
Quarterly site report: gas consumed, gas emitted, DRE by gas, uptime %, bypass hours.
MeasureEffective abatement rate = 1 − (emitted tCO₂e ÷ unabated tCO₂e), per site, per quarter, with CF₄ shown on its own line.
Move from self-declaration to third-party assurance under ISO 14064-3, escalating from limited to reasonable assurance over two cycles. Require a sample of tools to be measured at the stack (FTIR or equivalent) rather than derived entirely from mass balance and default factors.
Where it sticks: cost and verifier capacity. Pool it — a joint audit protocol across several brands halves the burden on the supplier and doubles the credibility.
Verification statement naming the site and the boundary; measurement campaign report.
Measure% of your upstream F-GHG tCO₂e covered by site-level third-party-verified data; number of tools stack-tested per year.
Put an explicit, weighted line in the RFQ scorecard. Set a contractual carbon-intensity ceiling per wafer or per m² of glass with a step-down schedule. Apply an internal shadow carbon price when comparing quotes. Commit to a share-shift rule: a defined share of volume moves to top-quartile performers at the next award.
Where it sticks: the commercial team is measured on unit cost. Fix that first — the scorecard weight is meaningless until the buyer's own target reflects it.
Scorecard with published weighting; award decisions with the carbon line documented.
MeasureScorecard weight (%); volume share awarded to top-quartile suppliers; delivered kgCO₂e per wafer or per panel, trended.
Push the verified site data into the LCA of the actual product using supplier-specific emission factors under ISO 14067 and the GHG Protocol Product Standard, replacing industry averages. Publish the F-GHG contribution as its own line in the PCF so that abatement is visible to anyone reading the datasheet.
Where it sticks: LCA databases update slowly and product teams fear a number that moves. A moving number is the point — it is the only proof the programme is real.
PCF documentation showing data source per component; critical review statement.
Measure% of product footprint built on supplier-specific data; kgCO₂e of F-GHG per unit shipped, year over year.
Sequence note: source reduction comes before capture. Process optimisation, chemistry substitution and gas-utilisation improvement remove emissions that never have to be destroyed. End-of-pipe abatement is the last line of defence, not the first move.
Put these on the table
Attention is the scarce resource. A question that cannot be answered comfortably in a supplier review is worth more than a slide that can. Ask them in this order, and write the answers down — including the silences.
Cadence · review every quarter, restate every half-year
Gas-by-gas emissions, effective abatement rate, uptime and bypass hours from every traced site. The one number that must be re-collected every single quarter — annual data hides a bad year inside a good average.
Spring brings the major environmental and ESG reports from both buyers and fabs. Read them for changes in wording, not just numbers — a requirement that becomes an expectation is a retreat.
World Semiconductor Council joint statements, Semiconductor Climate Consortium working-group publications, SEMI standards activity and CDP supply-chain cycles. Watch for a normalised-emission-rate goal turning into an absolute one.
Top-down atmospheric measurements of CF₄, C₂F₆ and NF₃ from global monitoring networks, against bottom-up inventories. Plus F-gas regulation, carbon pricing in Taiwan and Korea, and border-adjustment scope.
Earlier signals · where to move now
Per-wafer intensity improvements are real, but AI-driven wafer starts and advanced packaging volume are growing faster. Normalised targets can be met while absolute emissions rise. Insist on absolute numbers alongside intensity.
3D NAND layer counts, advanced-node multi-patterning and hybrid bonding all add etch and deposition steps. Each added step adds process gas. Assume intensity per die rises unless the supplier proves otherwise.
Atmospheric observations of long-lived fluorinated species have historically run higher than reported inventories imply. Treat that gap as a reason to demand measurement, not estimation.
HFCs are covered by the Kigali Amendment and the EU F-gas Regulation; PFCs and NF₃ in electronics manufacturing are largely handled by voluntary sector goals. Carbon pricing in Taiwan and Korea is where the first hard cost signal is likely to land.
NF₃ use scales with glass area. Panel supply chains are less transparent than logic and memory, and brands specify panels directly. Highest ratio of leverage to attention on this list.
A shared abatement data template and pooled verification protocol across several brands is the cheapest way to move a supplier who serves all of you. Competitive on price, aligned on measurement.
Sources · primary, for verification and further research
Every figure above should be checked against the source's current edition. Company programmes and sector goals are restated annually; treat this page as a snapshot to be re-run each cycle.
Honest limits. Company programme details, sector goals and reported figures change every reporting cycle, and much of the site-level abatement data described on this page is not public anywhere. Where a claim here summarises a company programme, read the current report before quoting it. Where a number is missing, that absence is itself the finding — and the question to bring to the next supplier review.