All news

An EPD costs $24,000 and ten weeks. AI is changing the math.

One EPD costs $24,000 and 24 person-days, flat for seven years. AI-driven workflows cut inventory time 80 to 90% and mapped a 5,600-component portfolio in seconds.

Published

June 10, 2026

Publisher

Miguel Altamirano, Sluicebox

Category

Industry Analysis

Inside the price of one EPD, and what AI removes from the workflow.

An EPD costs $24,000 and ten weeks. AI is changing the math.

Five years ago, an Environmental Product Declaration was a sustainability project. Today it gates revenue.

The shift started with buyers. In 2025, 330 corporate buyers asked 70,000 suppliers for product carbon data, a record. EcoVadis's 2024 Procurement Barometer found over 70% of procurement teams now name sustainability among their top buying criteria. Building rating systems like BREEAM and LEED reward projects that include verified EPDs. Manufacturers with EPDs win specifications; those without lose them.

Regulation is intensifying. CSRD requires large EU companies to disclose Scope 3 emissions, including the products they buy and sell. ESPR phases Digital Product Passports into electronics from 2028. California SB 253 covers companies over $1B doing business in the state, with Scope 1+2 disclosure due August 2026 and Scope 3 in 2027.

Carbon disclosure timeline showing CSRD in effect now, California SB 253 in August 2026, SB 253 Scope 3 in 2027, and ESPR Digital Product Passport for electronics in 2028

Most suppliers have not kept up. When buyers asked for product-level carbon data in 2024, only 2% delivered. The 2% that did are already pulling ahead, qualified into specifications and prepared for the deadlines ahead.

The question is no longer whether to publish EPDs, but how to do so at a pace and cost the business can carry.

$24,000 per EPD. Flat for seven years.

One EPD costs $13,000 to $41,000 and takes about 24 person-days of expert work. That came from a 2017 survey of 15 EPD program operators. A 2024 study of Danish state subsidy invoice data found the median almost unchanged at $24,037. Seven years, no movement.

Those numbers cover external consultant time only. Internal hours for supplier chase, BOM cleanup, and review cycles are usually larger. A 20-product program runs into hundreds of thousands of dollars and a multi-year timeline.

The bottleneck is data, not methodology.

Most of the time is not spent modeling. It is spent chasing data. A 2025 peer-reviewed synthesis estimated that 70 to 80% of LCA time and cost sits in inventory data collection alone.

Horizontal bar showing 75% of LCA time goes to inventory data collection, only 25% to modeling and verification

In electronics, the data resists collection. Bills of materials arrive with missing weights, ambiguous manufacturer names, and mismatched part numbers. So even Fraunhofer IZM, the German research institute behind the LCAs for Framework Laptop and Fairphone, supplements supplier BOMs with physical teardown, X-ray imaging, and component-level microscopy.

Then comes the wait for suppliers. Only 36% respond when asked for climate data, and only 2% provide anything at the product level.

Bar chart showing 70,000 suppliers asked for carbon data, only 36% respond, only 2% provide product-level lifecycle data

The bottleneck is not methodology. It is the manual labor methodology has always required.

AI removes the manual labor, not the science.

Peer-reviewed work has shown that NLP-based BOM matching cuts LCA inventory time 80 to 90% versus manual processes. A 2025 whitepaper documents what that looks like at portfolio scale. Researchers from UT Austin, Western Digital, Vishay, and TTI co-authored it; TÜV SÜD independently verified the methodology. Western Digital had spent two years to collect 20% of its supplier data manually. With AI-driven LCAs, they mapped 100% of a 5,600-component portfolio in seconds, with 30% higher matching accuracy than manual workflows.

Western Digital case study: two years of manual outreach reaching 20% coverage versus seconds per component reaching 100% mapped across a 5,600-component portfolio with 30% higher matching accuracy

Three things shift inside the workflow.

Side-by-side workflow comparison: traditional six-step manual EPD process taking 10+ weeks versus a five-step AI-driven workflow taking 1-2 weeks, with methodology unchanged

BOM parsing becomes seconds, with confidence scoring per match. Supplier requests run in the background, sending, following up, and parsing replies into the model. ISO methodology (14040, 14044, 14067) applies uniformly across the portfolio, with output formatted for verifiers (TÜV, BRE, EPD International) without a reformatting pass.

Sluicebox is building this for electronics manufacturers.

EPDs become a live data layer.

When one EPD takes one or two weeks instead of ten, and costs 75% less than the industry baseline, engineering reads carbon at design time, procurement compares suppliers on carbon as a routine input, and sales answers specifier carbon questions in days.

Four downstream effects of a portfolio-wide EPD layer: supplier empowerment, revenue at the gate, regulatory headroom, and real ecodesign

Manufacturers publishing EPDs quickly qualify for revenue their peers cannot reach, meet CSRD and ESPR with room to spare, and spend less per declaration than the industry baseline. That is the new baseline forming.

Sources

More news

Explore the latest product and sustainability updates.

Discover Carbon Component Intelligence, built for the electronics industry.