GPU & AI system buyback

Sell GPU & AI Systems

Highest value

We buy data-center GPUs and AI systems in bulk — the highest-value category we trade. NVIDIA H100, A100, L40S, H200 and B200; AMD Instinct MI300X; Intel Gaudi. Whether you are decommissioning an AI cluster, liquidating cancelled-project stock, or clearing surplus from a refresh, we take the lot as-is and quote it against live global demand.

What we accept

Accelerator models we buy

From current-generation training accelerators to previous-gen inference cards. If it is a data-center GPU or AI accelerator, we likely trade it — in volume.

Model Form Factor Notes
NVIDIA H100 80GB SXM5 / PCIe Current-gen AI training accelerator; peak demand
NVIDIA A100 40/80GB SXM4 / PCIe Mainstream data-center GPU; strong resale market
NVIDIA L40S PCIe Inference and graphics workload tier
NVIDIA H200 SXM5 141GB HBM3e variant; new generation
NVIDIA B200 SXM Blackwell architecture; latest generation
NVIDIA V100 SXM2 / PCIe Previous-gen; still traded for legacy clusters
NVIDIA T4 / A30 / A40 PCIe Inference and virtualization tier
AMD Instinct MI300X / MI250 OAM AMD accelerator; MI300X in high demand
Intel Gaudi 2 / 3 OAM Intel AI accelerator; emerging resale channel

We also buy HGX and DGX integrated systems (HGX H100, HGX A100, DGX H100, DGX A100) and GPU server nodes from Supermicro, HPE and Dell — complete chassis with GPUs, NICs and motherboards. Send the full configuration and we quote the lot.

Condition grading

Five tiers, firm payouts

Every unit is graded on intake. The grade maps to a payout range so the offer is firm — not adjusted after inspection. Below is what each tier means and the typical payout as a percentage of current market value.

Grade Description Typical payout range
Grade A New or sealed in original packaging 70–90% of current market
Grade B Used, tested, full function confirmed 50–70% of current market
Grade C Used, minor faults (e.g. one fan down) 30–50% of current market
Cosmetic Functional but visible aesthetic wear 20–40% of current market
Defective For-parts / parts recovery 5–20% of current market
Valuation factors

What drives your GPU payout

Five factors determine the bulk offer. You do not need to grade the units yourself — send the model list and condition notes, and we do the rest.

Model & SKU

The specific GPU model and SKU — SXM versus PCIe, 40GB versus 80GB versus 141GB — sets the baseline. Current-generation H100 and H200 command the highest unit pricing.

Configuration & VRAM

Memory capacity directly drives pricing. An 80GB H100 is worth substantially more than a 40GB A100, and the 141GB H200 sits at the top of the current market.

Quantity

Larger lots get better unit pricing and faster pickup scheduling. A full rack of eight H100s moves faster through our resale channel than a loose pair.

Condition

Every unit is graded A through Defective on intake. Each tier maps to a payout range; we test and grade before the offer is signed so there are no post-inspection adjustments.

Market demand

Live global demand for AI accelerators drives premium pricing on current-generation GPUs. Previous-generation stock still moves but at a discount to the latest SKUs.

How selling works

From your list to paid, in three moves

01

Submit inventory

Send your GPU asset list with models, quantities and condition notes. Use the quote form or email a spreadsheet to [email protected].

02

Get valuation

We assess the lot against live global demand and return a firm bulk offer with line-item pricing by model and grade. No auction, no marketplace — one buyer takes title.

03

Ship & get paid

Accept the offer and we arrange pickup across the USA and Canada. Payment is released per the signed terms — typically wire transfer within 1–3 business days.

Get a bulk offer

Turn idle GPUs into capital.

Send your GPU and AI system asset list — models, quantities, condition. We come back with a firm bulk offer, then arrange pickup and payment. Bulk lots only.