HQ NORCROSS, GA · 470-284-5400
Critical Technology Services
Rows of dark server racks threaded with orange and green network cabling in a data center.
GPU BUYBACK

Sell your data center GPUs

CTS buys used and surplus GPUs and GPU servers from businesses, including NVIDIA HGX and DGX systems. Send your inventory list and we will follow up with a buyback quote.

NVIDIA HGX · DGX · H100 · A100 · L40S · NIST SP 800-88 · Certificate per serial
GPU BUYBACK

Who buys used data center GPUs?

Critical Technology Services (CTS) buys used and surplus enterprise GPUs and GPU servers from businesses, including NVIDIA HGX and DGX systems and H100, A100 and L40S hardware, as part of its IT asset disposition (ITAD) and buyback services. Every drive in the servers we collect is sanitized to NIST SP 800-88 or physically destroyed under a documented chain of custody, with a certificate of destruction per serial number. Send an inventory list and we will review it and follow up with a quote.

GET A GPU BUYBACK QUOTE

Tell us what GPUs are coming out.

Send the models, quantities and locations. Our team will review the list and follow up with a buyback quote.

  • Model, memory, SXM or PCIe, quantity, condition and location.
  • Certificate of destruction per serial number for every drive.
  • Proceeds paid, or credited against your decommission or next build.

Prefer to talk now? 470-284-5400 · sales@critical-technology.com

REQUEST A GPU BUYBACK QUOTE

Company size (optional)

How can we help? (optional)

Reply within one business day · R2v3 · ISO 9001 / 14001 / 45001

We won't share your info. See our Privacy Policy.

Need immediate support or RFQ dispatch?

WHAT WE BUY

Data center GPUs and GPU servers we buy.

NVIDIA HGX and DGX systems

Including A100 and H100 platforms.

H100 and A100 GPU servers

Including H100 SXM5 and A100 80GB configurations.

L40S inference nodes

Including 8-GPU L40S servers.

GPU and rack servers

Including Supermicro 4U GPU servers, Dell PowerEdge and HPE ProLiant.

Memory and optics

The DDR4 and DDR5 ECC server memory and optical transceivers around the cluster.

Enterprise networking

Including switches from Arista, Cisco and Juniper.

Have H200, L40 or L4 hardware, loose PCIe cards, or a model not listed here? Send the list and ask us. What we can offer depends on model, condition, completeness and current market demand.

WHO WE BUY FROM

Who we buy from.

  • Data centers and colocation facilities

    Retiring racks, surrendering a cage, or consolidating sites.

  • Enterprises refreshing GPU clusters

    Moving to a newer GPU generation and retiring the previous one.

  • Teams exiting to the cloud

    On-premises infrastructure left stranded by a cloud migration.

  • Systems integrators, VARs and MSPs

    Retired hardware from the refreshes they sell to their clients.

HOW IT WORKS

From inventory list to settlement.

  1. 01 · SEND YOUR INVENTORY LIST

    Models, quantities and locations. A spreadsheet or an export from your asset system is fine.

  2. 02 · VALUATION BEFORE PICKUP

    We review the list against the secondary market and follow up with a buyback quote before anything moves.

  3. 03 · DECOMMISSION AND COLLECT

    Our data center team de-racks the cluster. Assets are serialized and logged into your client portal under an unbroken chain of custody.

  4. 04 · DESTROY THE DATA

    Every drive is sanitized to NIST SP 800-88 or physically shredded, with a certificate of destruction per serial number.

  5. 05 · TEST, GRADE AND REMARKET

    Hardware is tested and graded, then remarketed. Equipment with no resale value is recycled to R2v3.

  6. 06 · SETTLEMENT

    Line-item recovery proceeds are paid by wire or check, or credited against project logistics and services.

WHAT TO SEND

What to send for a GPU buyback quote.

Model
For example H100, A100 or L40S, and the system if it is an HGX or DGX.
Memory
For example A100 80GB.
Form factor
SXM or PCIe.
Standalone or in servers
Loose cards, or installed in servers (give the server make and model).
Quantity
Per model and configuration.
Condition
Working, tested, or known faults, and whether the units are complete.
Location and timing
Where the hardware sits and when it needs to leave.
Send your GPU list
WHY CTS

Why sell your GPUs to CTS.

01

Certified operator

CTS holds R2v3, ISO 14001, ISO 9001 and ISO 45001, and the Norcross facility holds NAID AAA certification.

02

In-house, not a broker

We run the wipe and destruction ourselves, with a certificate of destruction per serial number.

03

Built for dense GPU hardware

Our technicians are trained in handling high-density AI infrastructure: DGX and HGX systems, liquid-cooled nodes, InfiniBand switches and optical transceivers.

04

Net against your next build

Replacing the cluster? We quote the new build and the buyback on the systems it replaces together, so you see the net cost of the upgrade.

Retiring the whole room? See data center decommissioning. Buying the next generation? See AI infrastructure. Certifications are listed on our certifications page.

FAQ

GPU buyback questions we get a lot.

01

Do you buy individual GPUs or full servers?

We buy AI and GPU servers, including NVIDIA HGX and DGX systems, along with the memory and optics around them. If you have loose GPU cards, send the list with model, memory and quantity and ask us.
02

Which GPU models do you buy?

Including NVIDIA HGX and DGX systems, H100 SXM5 and A100 80GB servers, and L40S inference nodes. For other models, such as H200, L40 or L4, send your list and ask us. What we can offer depends on model, condition, completeness and current market demand.
03

Do you buy SXM and PCIe GPUs?

We buy SXM-based HGX and DGX systems, including H100 SXM5. For PCIe cards, tell us the model, memory and quantity, and whether they are loose or installed in servers, and ask us.
04

What information do you need for a quote?

The GPU model and memory, SXM or PCIe, whether the GPUs are loose or installed in servers (with the server make and model), the quantity, the condition, and where the hardware is located. A spreadsheet or asset export is fine.
05

What happens to the data on the servers?

Every storage device is tracked by serial number and sanitized to NIST SP 800-88 or physically shredded, and a certificate of destruction is issued per serial number before any remarketing.
06

Can you decommission the GPU cluster on site?

Yes. Our data center team de-racks and decommissions dense AI and GPU clusters, including NVIDIA DGX and HGX A100 and H100 systems, liquid-cooled nodes, InfiniBand switches and optical transceivers, under a documented chain of custody. Ask us about your site and timeline.
07

How are we paid?

Recovered proceeds are paid by wire or check, or credited against project logistics and services, with a line-item settlement per asset.
READY TO START?

Send your GPU list. We will follow up with a quote.