Compute Exchange launches a secondary GPU Marketplace
Compute Exchange
Sign Up
RESERVED GPU RENTAL

LOCK GPU CAPACITY.
GUARANTEED ACCESS.
DEPLOY FASTER.

The institutional market for pre-owned, refurbished, and OEM-surplus NVIDIA data center GPUs — from H100 and H200 through B200, B300, GB200/GB300 racks, and Vera Rubin forward reservations. Identity-verified counterparties. Live quote returned within 24 hours.

14
GPU CLASSES
3
CONDITION TIERS
24h
QUOTE TURNAROUND
Request CapacityBrowse GPUs
Partner Network

AGGREGATED ACROSS
LEADING NEOCLOUDS

Compute Exchange aggregates reserved capacity from a verified network of leading AI-native cloud providers and hyperscalers. All partners undergo identity, capacity, SLA, and operational verification before quotes surface on the network.

You receive a normalized comparison across providers in a single quote response — rather than evaluating each neocloud's contract structure, billing model, and SLA terms in isolation. Compute Exchange stays neutral; we do not operate compute capacity ourselves.

IDENTITY VERIFIED PROVIDERS

Corporate registration, beneficial ownership, and operational track record verified before any provider lists capacity.

NORMALIZED QUOTE FORMAT

Apples-to-apples comparison across providers with different contract structures, term ladders, and SLA terms.

NEURAL AGGREATION

Compute Exchange does not operate compute capacity. We facilitate the introduction; the contract is between you and the provider.

How It Works

RESERVE IN THREE STEPS

01
SPECIFY

Tell us GPU model, quantity, region, term length, and SLA requirements. Multi-region or hybrid term structures supported.

02
MATCH

We aggregate live reserved-capacity quotes from verified neocloud partners in our network. You receive a shortlist with term options and lead times.

03
RESERVER

Select a quote and Compute Exchange facilitates the introduction. The provider contracts directly. Capacity provisions per the agreed schedule.

TERM AVAILABLE BY GPU

WHAT YOU CAN RESERVE

Click any GPU for full specifications, use cases, and reservation FAQ. Term availability reflects partner network depth in Q3 2026.

GPUGENERATIONTERM / Lenght / AvailabilityNotes
1MO3MO6MO12MO24MO36MO
VERA RUBINRUBINForward reservations; supply begins 2027
GB300 NVL72BLACKWELL ULTRARack-scale; early allocation, quoted per rack
GB200 NVL72GRACE BLACKWELLRack-scale NVL72; quoted per rack
B300BLACKWELL ULTRA288GB HBM3e; allocation-constrained, 6-month minimum
B200BLACKWELLAllocation-constrained; 3-month minimum commit
H200 SXMHOOPER141GB HBM3e; tighter supply, longer commits prioritized
H200 PCIeHOOPERH200 NVL; air-cooled PCIe hosts, wider provider base
H100 NVLHOOPER94GB HBM3; inference-tuned PCIe pair
H100 SXM5HOOPERDistributed-training workhorse
H100 PCIeHOOPERInference-optimized PCIe form factor
A100 80 GBAMPEREBest value for non-FP8 workloads
L40SADA LOVELACEFP8 acceleration, PCIe Gen4
A100 40 GBAMPERECheapest data-center reserved tier
V100 32GBBLACKWELLLegacy; 12-month max as supply winds down

Term availability reflects active partner-network supply for Q1 2026 and is subject to change. Lead times and term discounts depend on quantity, region, and SLA. Compute Exchange returns a live quote within 24 hours.

WHY RESERVE

RESERVED VS ON-DEMAND

GUARANTEED CAPACITY

Contractual guarantee that the reserved GPU count is available to you for the contracted window. Priority over on-demand traffic during shortage events.

PREDICTION SPEND

Lock unit economics for the full term. Budget and forecast confidently without exposure to on-demand rate volatility during peak demand cycles.

AGGREGATED NEOCLOUD PRICING

Compute Exchange aggregates rates across hyperscalers, AI-native clouds, and specialty providers. You see a normalized comparison rather than evaluating each provider in isolation.

VERIFIED PARTNER NETWORK

All providers in the network undergo identity, capacity, SLA, and operational verification before quotes surface. You know who is on the other side of every reservation.

DIRECT PROVIDER CONTRACT

Compute Exchange facilitates the introduction. Your contract is directly with the provider. We do not take ownership of capacity or escrow funds.

FLEXIBLE TERM LADDER

1, 3, 6, 12, 24, and 36-month standard. Custom terms (multi-year phased delivery, ramp schedules) available for large commitments.

BROW BY GPU

RESERVED BY MODEL

RESERVED VERA RUBIN
Rubin — 2027 supply
TERM RANGE
12MO - 36MO
View Details
RESERVED GB300 NVL72
Blackwell Ultra
TERM RANGE
12MO - 36MO
View Details
RESERVED GB200 NVL72
Grace Blackwell
TERM RANGE
6MO - 36MO
View Details
RESERVED B300
Blackwell Ultra
TERM RANGE
6MO - 36MO
View Details
RESERVED B200
Blackwell
TERM RANGE
3MO - 36MO
View Details
RESERVED H200 SXM
Hopper
TERM RANGE
1MO - 36MO
View Details
RESERVED H200 PCIe
Hopper
TERM RANGE
1MO - 36MO
View Details
RESERVED H100 NVL
Hopper
TERM RANGE
1MO - 36MO
View Details
RESERVED H100 SXM5
Hopper
TERM RANGE
1MO - 36MO
View Details
RESERVED H100 PCIe
Hopper
TERM RANGE
1MO - 36MO
View Details
RESERVED A100 80GB
Ampere
TERM RANGE
1MO - 36MO
View Details
RESERVED L40S
Ada Lovelace
TERM RANGE
1MO - 36MO
View Details
RESERVED A100 40GB
NVLink 3.0 / NVSwitch (600 GB/s)
TERM RANGE
1MO - 36MO
View Details
RESERVED V100 32GB
NVLink 2.0 (300 GB/s)
TERM RANGE
1MO - 36MO
View Details
RESERVED MI325X
CDNA 3
TERM RANGE
1MO - 36MO
View Details
RESERVED MI350X
CDNA 4
TERM RANGE
3MO - 36MO
View Details
RESERVED MI355X
CDNA 4
TERM RANGE
1MO - 36MO
View Details

Indicative ranges based on aggregated supplier quotes and broker market data, Q3 2026. Final pricing depends on quantity, region, condition, accessory bundling, and warranty terms. Compute Exchange does not publish active counterparty listings.

Frequently Asked Questions

RESERVED GPU
RENTAL, EXPLAINED

What is reserved GPU rental?+
Reserved GPU rental is a contracted commitment to GPU compute capacity for a fixed term — typically 1 to 36 months — that locks in a specific number of GPUs for your use across the contracted window. The provider guarantees capacity is available to you with priority over on-demand traffic during shortage events. In exchange, you commit to the spend regardless of actual utilization.
What term lengths can I commit to?+
Compute Exchange aggregates reservations across 1, 3, 6, 12, 24, and 36-month terms. Shorter commits suit pilot programs and dev environments; 12-month and longer commits unlock the deepest discounts and the strongest capacity guarantees. Custom terms (multi-year phased delivery, ramp schedules) available for large commitments.
Who are Compute Exchange's partner neoclouds?+
Compute Exchange aggregates reserved capacity across a verified network of leading AI-native cloud providers and hyperscalers. All providers undergo identity, capacity, SLA, and operational verification before quotes surface. Specific provider names are disclosed during the quote process — Compute Exchange does not publish individual provider rates or rosters publicly.
When should I choose reserved over on-demand?+
Reserved is the right choice when your workload has predictable load curves over the term — production inference clusters, multi-week training runs, dev environments with steady utilization. On-demand stays better for spiky or experimental workloads where utilization could fall well below the breakeven against the reserved commit. Reserved also matters when capacity guarantees are needed during shortage events.
How long does reserved capacity take to provision?+
Most reservations of available models (H100 PCIe, A100 80GB, L40S) provision within 1 to 2 weeks of contract signing. Tighter-supply parts (H200, B200, B300) and rack-scale systems (GB200/GB300 NVL72) may take 4 to 8 weeks or provision on phased delivery schedules; Vera Rubin reservations are forward commitments against 2027 supply. Compute Exchange surfaces current lead times in the quote response and can structure phased delivery for large commits.
Can I commit across multiple GPU models or regions?+
Yes. Multi-GPU reservations spanning different models (e.g. H100 SXM5 for training plus L40S for inference) and multi-region deployments (US, EU, APAC) are common across the partner network. Compute Exchange consolidates the quote into a single normalized comparison so you can compare full-fleet economics across providers.

LOCK YOUR GPU CAPACITY

Submit a reservation request and Compute Exchange returns a live quote across the verified neocloud partner network within 24 hours.

RESERV CAPACITY
DISCLAIMER

Reserved GPU rental quotes are aggregated from verified third-party providers. Compute Exchange facilitates introductions and surfaces term availability but does not operate compute capacity, take ownership of provider contracts, or guarantee SLA delivery. All reservation terms, including SLA, billing, and remedies, are negotiated directly between buyer and provider.