Data Center Decommissioning: Process, ITAD, and Recycling

Data center decommissioning is the managed removal of servers, storage, network gear and supporting infrastructure from a room, cage or whole facility, ending with certified data destruction and a clean handover of the space. For a colocation tenant, the schedule is set less by engineering than by the contract: notice windows, access rules and restoration clauses determine what the exit costs.

Key takeaways

• The process has six stages: scoping, inventory, data sanitization, de-installation, ITAD (resale or recycling) and handover. Sanitization and chain of custody are the audit-critical steps.
• NIST SP 800-88 defines Clear, Purge and Destroy; IEEE 2883-2022 covers modern NVMe, flash and SMR media.
• Vendor-published ranges: $10-50 per drive for sanitization, $260-850 per server, $25,000-150,000 for many midsize projects, with 30-50% recovered through resale. These are marketing figures, not audited data.
• Colocation exits are governed by notice periods (commonly 30-60 days), auto-renewal and early termination fees that can equal all remaining monthly fees.
• Only 22.3% of the 62 million tonnes of e-waste generated in 2022 was documented as formally recycled (UN Global E-waste Monitor 2024).
• Tight markets change exit economics: with North American vacancy at a record-low 1.4% in H1 2026 (CBRE), the landlord can often re-let released capacity quickly, which is negotiating leverage for early release.

Why decommissioning is a separate project

Most teams plan deployments carefully and treat exits as an afterthought. The result is predictable: unlabeled racks, missing asset records, drives with unknown data classification, and a landlord deadline. Decommissioning also carries risks that deployment does not. A single unsanitized drive is a reportable data-breach candidate under most regimes, and equipment leaving a secure facility needs documented chain of custody.

Triggers vary: cloud migration, consolidation after M&A, hardware refresh, end of a colocation term, or moving a workload to a higher-density site. The process below applies to all of them, scaled by rack count.

The six-stage process

Stage	What happens	Typical output
1. Scoping	Define the boundary: racks, cross-connects, power circuits, cage build-out	Project plan, contract review, target dates
2. Inventory	Reconcile CMDB or DCIM records against a physical audit by serial and asset tag	Asset register with data classification
3. Sanitization	Wipe, purge or destroy storage media per policy	Per-serial erasure report
4. De-installation	Power down, unrack, label, pack and move under escort	Signed transfer manifests
5. ITAD	Test, resell, refurbish or recycle	Resale settlement, certificates of recycling
6. Handover	Remove cabling and cage materials, return space, close cross-connects and circuits	Landlord sign-off, final invoice

Inventory first

Physical audits routinely disagree with records. Reconciling before any wipe or removal prevents the two costly errors: decommissioning something still in production and losing track of a device that holds data. If you run DCIM software, export it first; our DCIM guide covers what a usable asset record contains.

Sequencing power and network

Shut down in dependency order: application traffic, servers, storage, then network. Cancel cross-connects only after the last dependent link is confirmed dark; see cross-connect pricing for how these are billed monthly and why forgotten ones keep charging.

Data sanitization: what the standards say

NIST SP 800-88 defines three levels. Clear uses logical overwriting for standard-sensitivity data. Purge uses methods such as cryptographic erase for SSDs that hold sensitive data. Destroy means shredding, disintegration or incineration, used for the highest sensitivity or failed media that cannot be verified.

Level	Method	Reusable afterwards	Typical use
Clear	Logical overwrite	Yes	Standard-sensitivity data, redeployment
Purge	Cryptographic erase, firmware secure erase	Yes	Sensitive data on SSD/NVMe
Destroy	Shred, disintegrate, incinerate	No	Classified data, failed drives

IEEE 2883-2022 supplements NIST 800-88 with implementation detail for NVMe, flash and shingled magnetic recording drives. Practical points:

• Demand a certificate per serial number, not a batch certificate.
• Decide on-site versus off-site destruction by data classification. On-site shredding removes transport risk but limits resale.
• Self-encrypting drives allow crypto-erase, which is faster, but verify the vendor implementation.

What it costs

Independent, audited cost benchmarks for decommissioning do not exist; the public figures come from ITAD providers and are best read as ranges to test quotes against.

Item	Vendor-published range	Source type
Drive sanitization	$10-50 per drive	ITAD provider blogs
Full server decommissioning	$260-850 per unit	ITAD provider blogs
Domestic logistics	$100-300 per pallet	ITAD provider blogs
Midsize project	$25,000-150,000	ITAD provider blogs
Value recovery on resale	20-50% of original hardware value	ITAD provider blogs

Cost drivers are rack count, hardware age (older gear resells for less and may cost money to recycle), destruction versus wipe, distance, and out-of-hours access. Resale value falls quickly for hardware more than four or five years old, so timing the exit against refresh matters.

ITAD: resale, refurbishment and recycling

IT asset disposition converts retired equipment into either cash or documented recycling. A competent provider will test and grade equipment, remarket usable units, and route the rest to certified recyclers, settling proceeds against fees.

Evaluate providers on:

1. Certifications. Ask for current e-waste recycling and data-destruction certifications and verify them with the issuer.
2. Chain of custody. GPS-tracked transport, sealed containers and signed manifests.
3. Downstream transparency. Named recyclers and country of final processing.
4. Settlement terms. How resale proceeds are split, and over what period.

The wider context is sobering. The UN Global E-waste Monitor 2024 reports 62 million tonnes of e-waste generated in 2022, with 22.3% documented as formally collected and recycled, and projects 82 million tonnes by 2030. Data center gear is a small share but is high in recoverable metals, and enterprise buyers increasingly ask for recycling certificates for ESG reporting.

The colocation contract: where exits get expensive

For tenants, the contract sets the timeline and the price of leaving. Points reported in legal and advisory guidance:

• Notice. Commonly 30-60 days of written notice before the end of the term. Missing it can auto-renew the contract for a period equal to the initial term.
• Early termination. Fees typically equal all remaining monthly commitments. On a 3-year contract at $300,000 per month with 18 months left, that is $5.4 million.
• Decommissioning plan. Some contracts require a detailed plan 30 days ahead for tenants with 10 or more cabinets.
• Restoration. Tenants often must remove racks, cabling and cage materials and return the space in original condition.
• Access and escorts. Removal needs scheduled windows and asset-tag notification for anything leaving the building.
• Holdover. Staying past the term usually carries a penalty rate.

Read these clauses at signing, not at exit; our colocation contract terms guide covers the negotiation side, including step-down rights that let you shrink capacity in stages.

Market conditions change your leverage

CBRE recorded record-low vacancy of 1.4% in North American primary markets in H1 2026 and wholesale asking rents above $196 per kW per month in H2 2025 (see the Colocation Price Index). When capacity is scarce, a landlord can re-let released megawatts fast and may agree to early release or a buy-out below the full remaining commitment. In markets with higher vacancy, expect to pay closer to the contractual maximum.

Moving instead of closing

Many exits are relocations. The decommissioning of the old site and the deployment at the new one overlap, and the cost of running both is the main budget risk. Plan a parallel-run window, and compare destinations before committing: see colocation vs cloud TCO and browse candidate sites in the data center catalog. For a shortlist of service companies, see our services directory, which lists what companies state on their own sites and is not a ranking.

Common mistakes

• Wiping before reconciling inventory against production dependencies.
• Accepting batch-level destruction certificates.
• Missing the auto-renewal notice date.
• Leaving cross-connects and power circuits active after the move.
• Ignoring restoration obligations until the final week.
• Choosing an ITAD vendor on resale promise alone, without downstream transparency.

What to do

1. Pull the contract now and calendar the notice deadline, restoration terms and any decommissioning-plan requirement.
2. Run a physical audit and reconcile it with your asset records at least 60 days before the exit.
3. Classify data by sensitivity and choose Clear, Purge or Destroy per class under NIST 800-88.
4. Get at least three itemized ITAD quotes, and benchmark them against the ranges above.
5. Require per-serial certificates and named downstream recyclers.
6. Cancel cross-connects and circuits only after traffic confirmation, and get landlord sign-off in writing.
7. If you are moving, request quotes for the destination capacity through our quote form.