40 to 60 Times Higher: The Full-Infrastructure AI E-Waste Estimate

In September 2026 the environmental non-profit Basel Action Network published an analysis with a different boundary from everyone else. Count the whole data-centre stack - servers and accelerators, power distribution, cooling, backup power and storage, switches and cabling - and then add the consumer computers, phones and telecommunications equipment pushed into early retirement by new compute requirements. On that boundary, 395 to 617 million tonnes of AI-driven electronic equipment would be retired between 2025 and 2050. That is 40 to 60 times higher than previous estimates. The model did not change. The inventory did.

Change the boundary and the conclusion changes

1. Why the Number Jumped: The Boundary Moved

Most earlier estimates counted servers and accelerators. BAN's analysis counts the full infrastructure: power distribution, cooling, backup power and energy storage, networking such as switches and cabling, plus the consumer computers, phones and telecom equipment displaced by new compute requirements. The conversion rule BAN publishes is simple: every gigawatt of data-centre capacity corresponds to roughly 70,000 metric tonnes of e-waste. Move the denominator from servers to the entire infrastructure and add a consumer replacement wave to the numerator, and the result rises by an order of magnitude. That methodological difference is itself the most instructive part of the report.

# 1. Why the number jumped: the boundary moved, not the model
BOUNDARY_OLD = ["AI servers", "accelerators"]
BOUNDARY_NEW = [
    "AI servers and accelerators",
    "power distribution equipment",
    "cooling equipment",
    "backup power and energy storage",
    "networking (switches, cabling)",
    "consumer computers and phones displaced by new compute requirements",
    "telecommunications equipment with the same displacement effect",
]

def estimate(boundary, gw_capacity, tonnes_per_gw=70_000):
    # BAN's rule of thumb: every gigawatt of data-centre capacity corresponds
    # to roughly 70,000 metric tonnes of e-waste. The headline changed because
    # the inventory changed.
    return gw_capacity * tonnes_per_gw

2. The Headline Figures, Kept Attached to Their Assumptions

As published: between 2025 and 2050, AI-driven equipment retirement totals 395 million to 617 million metric tonnes. Packed into standard shipping containers and laid end to end, that is roughly 23 million containers, enough to circle the Earth about six times. At annual resolution it works out to roughly 8.6 million to 13.1 million metric tonnes retired per year. The input assumptions matter as much as the output: 70,000 tonnes per gigawatt; capacity figures citing a McKinsey projection of up to 219 GW worldwide by 2030; and 4,871 new data centres in the US construction pipeline. Quote environmental projections with their boundary attached, or the number becomes a slogan.

// 2. The headline figures, exactly as published
const banFullInfrastructure = {
  published: "2026-09-15",
  organisation: "Basel Action Network (BAN)",
  scope: "first-of-its-kind full-infrastructure analysis",
  projection: {
    window: "2025-2050",
    retired_tonnes: "395,000,000 - 617,000,000",
    versus_previous_estimates: "40x - 60x higher",
    containers: "about 23 million standard shipping containers",
    laid_end_to_end: "would circle the Earth roughly six times",
    annual_rate: "8.6m - 13.1m metric tonnes retired per year",
  },
  inputs: {
    tonnes_per_gw: 70_000,
    capacity_2030_cited: "up to 219 GW (McKinsey projection)",
    us_datacenters_in_pipeline: 4871,
  },
};

Retirement volume is set by the replacement interval

3. The Replacement Clock Is a Procurement Fact

The most engineering-flavoured section of the report is the physical inventory. Racks of computer servers weighing 1,360 kilograms each. Switches of up to 30 kilograms. Hundreds of kilograms of copper cable connecting each rack. And the projection that such equipment will not be designed for repair and reuse, needing complete replacement every two to five years. Translated into procurement language: the replacement interval you write into a specification today determines the magnitude of retirement volumes for a decade, before a single workload runs. When equipment is not repairable, replacement is disposal with extra steps.

# 3. The replacement clock is the real driver - and it is a procurement fact
ASSET_CYCLE = {
    "server_rack_weight_kg": 1360,
    "switch_weight_kg": "up to 30",
    "copper_cable_per_rack_kg": "hundreds of kilograms",
    "designed_for_repair_and_reuse": False,
    "replacement_interval_years": (2, 5),
}

def embodied_waste(fleet, horizon_years):
    low, high = ASSET_CYCLE["replacement_interval_years"]
    # A 2-5 year replacement interval means the hardware you specify today
    # determines waste volumes for the next decade, before any workload runs.
    return {
        "replacements_per_rack": horizon_years / high,
        "approx_replacements_at_fast_cycle": horizon_years / low,
        "note": "if equipment is not designed for repair, replacement equals disposal",
    }

4. Five Things a Specification Can Actually Change

No single team can bend the global capacity curve, but five requirements fit in a procurement document. Longevity: ask for a published service life and a spare-parts commitment. Repairability: score bidders on field-replaceable units and documentation. Reuse path: name the downstream owner before the first shipment arrives. Accounting: report e-waste per gigawatt the way you report PUE. Destination: ask where retired equipment physically goes, and under whose jurisdiction. The report's argument is not that AI should stop; it is that the retired-hardware ledger currently has no owner, and a specification is the cheapest place to assign one.

# 4. What a team can actually change in a spec sheet
PROCUREMENT_LEVERS = {
    "longevity": "require a published service life and a spare-parts commitment",
    "repairability": "score bidders on field-replaceable units and documentation",
    "reuse_path": "name the downstream owner before the first shipment arrives",
    "accounting": "report e-waste per GW the same way you report PUE",
    "siting": "ask where retired equipment physically goes, and in whose jurisdiction",
}

def score(bid):
    return sum(1 for k, requirement in PROCUREMENT_LEVERS.items()
               if requirement in bid["commitments"])

# The report's point is not that AI should stop. It is that the retired-hardware
# ledger has no owner today, and a specification is the cheapest place to fix that.

A specification is the cheapest place to assign ownership of retired hardware

5. A One-Page Hardware Retirement Plan

A template you can use immediately: asset class (inference rack, 1,360 kg per rack); designed service life (five years) against actual replacement interval (2.5 years); a named downstream owner, combining a reuse partner and a certified recycler with a take-back clause in the contract; materials of concern (lead, mercury, cadmium, PFAS); the metric (metric tonnes retired per gigawatt per year, benchmarked against BAN's 70,000 tonnes per gigawatt); an explicit list of what the plan does not cover, such as consumer displacement and telecom equipment; and a review date tied to the next capacity expansion. Turning an invisible cost into a line item is the most direct thing this report offers.

{
  "hardware_retirement_plan": {
    "asset_class": "inference rack, 1360 kg per rack",
    "designed_service_life_years": 5,
    "actual_replacement_interval_years": 2.5,
    "downstream_owner": "named reuse partner + certified recycler, with a take-back clause in the contract",
    "materials_of_concern": ["lead", "mercury", "cadmium", "PFAS"],
    "reporting": {
      "metric": "metric tonnes retired per GW of deployed capacity per year",
      "baseline_reference": "BAN projects 70,000 t per GW across the full infrastructure boundary",
      "cadence": "quarterly, published internally"
    },
    "exclusions_not_covered": ["consumer device displacement", "telecom equipment"],
    "review_due": "at the next capacity expansion review"
  }
}