Sell your AI and data-centre GPUs — we pay top of the market.
Data-centre GPUs and AI systems are the highest-value hardware we buy. Maxicom buys used and surplus NVIDIA and AMD accelerators across Canada — whole DGX and HGX nodes, GPU servers, or loose cards — and pays you in CAD against your PO. Send your list, we destroy any data on board and arrange collection as part of the deal.
Accelerated compute holds its value — so we pay for it.
Of everything in a retired estate, data-centre GPUs and AI systems keep the most value on the secondary market — which is exactly why they deserve a proper offer rather than being lumped into a scrap or shredder job. Whether you're pivoting, consolidating, coming off a training run, or clearing a lease, Maxicom quotes each accelerator on its model, generation and condition and pays you in CAD. Here's what we buy.
NVIDIA data-centre GPUs
The core of the AI secondary market — current and previous generation, in cards or SXM modules.
NVIDIA DGX & HGX systems
Whole AI nodes and baseboards — we buy complete systems, not just the cards inside.
AMD Instinct
AMD's data-centre accelerators, whole nodes or loose.
Professional & workstation GPUs
Pro visualization and inference cards from workstations and render nodes.
GPU servers & nodes
Populated GPU servers — the whole box, cooling and interconnect included.
What drives the offer
AI hardware is priced far more precisely than commodity IT — the details matter, so put them on the list.
One card or a whole cluster.
- Whole clusters & nodes — training clusters being retired or consolidated, complete DGX/HGX systems.
- Loose accelerators — pulled cards, spares, warranty-return stock, matched sets.
- GPU servers — populated nodes with cooling and interconnect intact.
- Working or faulty — tested and healthy, or faulty units — tell us the condition and we price it.
Four steps, and you get paid.
Send your list
Models, memory, quantities and condition — a spreadsheet or photos work.
Get a written quote
A firm offer in CAD against your PO, priced to the live AI secondary market.
We collect
Collection arranged GTA-wide and nationwide, on a documented chain of custody.
You get paid
Any onboard data wiped, certificate issued, settled in CAD.
The teams retiring accelerated compute.
AI, research & HPC teams
AI companies scaling or changing platforms, university and research HPC groups refreshing clusters, and labs clearing last-generation accelerators.
Data centres, cloud & the trade
Colocation and cloud operators cycling fleets, plus resellers, ITADs and MSPs moving GPU stock and client estates — all on NDA, with discretion as standard.
Wiped, documented, and kept quiet.
AI nodes often carry NVMe storage and sensitive workloads. Any onboard data is destroyed before resale, and GPU deals are handled on NDA as standard.
How we price it — and how to get the best number.
Send an accurate list and you get an accurate price. What moves the offer on your accelerators is specific: the exact model and generation (H100 vs A100 vs L40S), memory (HBM) capacity, SXM vs PCIe, quantity and matched sets, warranty status, and condition. Quantity and matched lots help, current secondary-market demand sets the level, and any data-bearing drives are destroyed to standard as part of the deal rather than charged on top. The more precise your list — models, specs, quantities, condition — the sharper and faster the written CAD number comes back, with no automated instant-price guesswork.
AI & GPU buyback — the short answers.
How much is a used H100 or A100 worth?
Do you buy whole DGX / HGX systems?
Do you buy loose or pulled GPUs?
Do you buy faulty accelerators?
Is data on the AI nodes destroyed?
Do you buy from resellers and integrators?
Find out what your GPUs are worth.
Send your list — models, memory, quantities and condition. We'll come back with a written CAD offer, on NDA, with collection and certified data destruction arranged as part of the deal.