Inside the AI data center, every part has two price tags.

One is what it costs to buy and run. The other is what it leaves behind when it is retired. Click any marker to see both.

CostWaste

$725B in planned 2026 capex at Google, Amazon, Microsoft and Meta

AI server e-waste figures follow de Vries-Gao , which counts server hardware only, not the cooling, power or building equipment around it. Custom chips such as Google's TPUs are not publicly reported and enter only the upper bound. Figures elsewhere come from different years and methods; read them as orders of magnitude.

The problem

The AI build-out has a back end.

Every server bought for AI becomes a depreciating asset and, eventually, a stream of circuit boards and batteries. How that hardware is retired decides how much value a company recovers and how much waste the world absorbs.
$725B

$725B

Planned 2026 capex at Google, Amazon, Microsoft and Meta, up 77% from about $410B a year earlier .

Data: 2026

Up to 224.8M kg

Up to 224.8M kg

Upper bound for AI server e-waste in 2030, about 0.13 to 0.22 Mt, based on how many AI servers chip-packaging capacity can produce. That is in the same range as the yearly e-waste of countries such as Denmark and Austria .

Data: 2026

22.3%

22.3%

Share of the world's 62 Mt of 2022 e-waste documented as properly collected and recycled .

Data: 2022

24 to 31%

24 to 31%

Lower lifetime cost of ownership per server when hardware gets a second life with downstream users, modeled for grids from Quebec to Virginia .

Data: 2020

Amazon and Meta made opposite calls on how long servers last

Hardware lifespan is an accounting choice worth billions. In January 2025 Amazon shortened the useful life of a subset of its servers and networking gear from six years to five, citing the pace of AI development, and expected about $700 million less operating income in 2025, on top of $920 million in accelerated depreciation for early retirements . The same month Meta lengthened its server life to 5.5 years and expected roughly $2.9 billion less depreciation in 2025 .

Public policy

E-waste policy strength

Select a country to inspect the regulation in force.
Loading map…
ComprehensiveEstablishedEmergingLimitedNo PolicyNo Data

No country selected

Click a country on the map to see the policies tracked there.

Circularity practice

Circularity recommendations

Strategies to cut e-waste across the hardware lifecycle.

Design

Design for Disassembly

AI priority

Specify screw-based fastening, labeled subassemblies and tool-light access so servers, GPUs and cooling modules can be separated cleanly at end of life.

Impact
High
Feasibility
Medium
AI relevance
High

Design

Modular and Upgradeable Architecture

AI priority

Decouple accelerators, memory and storage from the chassis so a node can be upgraded rather than replaced when a new AI hardware generation arrives.

Impact
High
Feasibility
Medium
AI relevance
High

Use

Reuse and Refurbishment

AI priority

Redeploy retired training hardware into inference, staging or edge workloads, extending useful life by two to four years before recycling.

Impact
High
Feasibility
High
AI relevance
High

End of life

Component Recovery

AI priority

Harvest DIMMs, SSDs, power supplies and interconnects from decommissioned nodes to feed spare-part pools and cut new procurement.

Impact
Medium
Feasibility
High
AI relevance
High

About this project

To what extent do current sustainability reporting frameworks and e-waste policies adequately address electronic waste generated by AI-driven data center infrastructure?

Why this exists

CircularAI Monitor is an interactive observatory tracking the e-waste and hardware lifecycle impact of AI infrastructure expansion — data centers, GPUs and servers. It brings together public policy, corporate disclosure and circularity practice in a single, source-linked view.

Who it's for

  • Policymakers and regulators

    Click a country to see its e-waste policies and what those policies actually say.

  • Sustainability and ESG professionals

    See what each company publicly discloses about e-waste, with a focus on data center–related e-waste.

  • Researchers and academic institutions

    Every figure is source-linked, with full references below.

What's in the dashboard

  • E-waste policy strength

    Select a country to inspect the regulation in force.

    Open full page
  • Company disclosures

    What monitored companies publish about e-waste — no ranking, no scoring.

    Open full page
  • Circularity recommendations

    Strategies to cut e-waste across the hardware lifecycle.

    Open full page
  • Frameworks & agreements

    Supranational law, treaties, voluntary frameworks, industry pacts and certification standards

    Open page
  • Regulatory base

    Requirements, authorities and documents behind every tracked policy

    Open page
  • Reports

    Source documents behind every published disclosure

    Open page

How the evidence was gathered

The observatory brings together public policy, corporate disclosure and circularity practice. Company entries show only what each company publicly reports, with a source on every metric; nothing is scored or inferred.

Frameworks and agreements
EU WEEE Directive · ESRS E5 · EU Taxonomy · EU Servers Ecodesign · EU EED · EU Waste Framework Directive · GRI 306 · ISSB / IFRS S1 & S2 · SASB · ISO 59000 · Basel Convention · Stockholm Convention · Bamako Convention · CNDCP · R2v3 · e-Stewards
Policies
WEEE Directive · ElektroG3 · AGEC Anti-Waste Law for a Circular Economy · Circular Economy Implementation Programme · RCRA · California SB 20 · Virginia CRRA · Provincial EPR Programs for Electronics · National Solid Waste Policy — Electronics Sectoral Agreement · REP Law 20.920 · Act on Promotion of Recycling of Small Waste Electrical Equipment · E-Waste (Management) Rules, 2022 · WEEE Fund and Circular Economy Promotion Law · National Television and Computer Recycling Scheme · Circular Economy Policy 2021-2031 · EPR Regulations for Electrical and Electronic Equipment · EPR Guidelines for the Electrical/Electronics Sector · Waste Management Law — Electronic Waste Chapter
Companies
Atlas Server Systems · Dunes Digital · Kanso Compute · Meridian Data Estates · Northwind Cloud · Silicon Forge · TerraByte Africa · VerdeCloud