SSP, Strategic Supply Partners Live inventory
The desk
GPU procurementNodes and clusters, landed to your rack ● Live inventoryVerified lines, indicative pricing
By platform
B300 and GB300Current-generation Blackwell H200 and H100New stock and fleet rotation All platformsThe full catalogue
Rest of the stack
Memory, servers and storageThe kit around the accelerators Networking and opticsTransceivers, switching, cabling
Capacity without capex
GPU-as-a-serviceReserved, committed spend, on demand Bare-metal leasingDedicated clusters on a term
Decide first
Cloud vs bare metalWhere each one actually wins Reserved vs on demandThe break-even, calculated Cluster cost calculatorThe whole bill, not the GPU line
Space and power
Colocation and powerSingle racks to 100 MW Above 40 kW per rackWhere the shortlist collapses
Where
United States Europe Asia Pacific Middle East and Africa All regions →
Supply side
Have capacity to fill?Tell us what you hold
Analysis
InsightsWhat things cost, with the numbers Sizing guidesMemory, nodes, power, cost ComparisonsBuy, lease or rent
Start here
How an engagement worksThe first week, held to dates Find your fitThree questions, the right desk Glossary
Talk to the desk
● Live inventory
Buy GPUs
GPU procurementB300 and GB300 H200 and H100All platforms Memory, servers and storageNetworking and optics
Rent Clusters
GPU-as-a-serviceBare-metal leasing Cloud vs bare metalCost calculator
Data Centre Capacity
Colocation and power United StatesEurope Asia PacificMiddle East and Africa All regionsHave capacity to fill?
Insights
InsightsSizing guides ComparisonsHow it worksGlossary
Talk to the desk →
GPU procurement · GPUaaS · Colocation

The buyer's desk for AI infrastructure.

SSP sources GPUs, GPU-as-a-service agreements and data centre capacity for the party that will actually run the workload. We sit on your side of the table, and we connect you directly to the supplier. Nobody sits between you and them.

Tell us what you need See live inventory

Buyer-side by design. We hold no inventory and take no position in the deal. When a line is wrong for you, we say so.

HWB300 HGX · channel allocationHWGB300 NVL72 · build slotsHWH100 HGX · from stockHWH200 HGX · fleet rotationSVCGPUaaS · reserved & committed spendDCColocation · up to 100 MWHWRTX 6000 · workstation & dense computeHWB300 HGX · channel allocationHWGB300 NVL72 · build slotsHWH100 HGX · from stockHWH200 HGX · fleet rotationSVCGPUaaS · reserved & committed spendDCColocation · up to 100 MWHWRTX 6000 · workstation & dense compute

We connect the people buying AI infrastructure directly to the people who supply it: OEMs, ODMs, distributors, GPU clouds and data centres. GPUs, GPU-as-a-service, colocation. One desk, one file, no layers in between.

The landscape

Where the desk sits.

Every requirement crosses this map once. OEMs, ODMs, distributors, GPU clouds and data centres on one side, the parties running workloads on the other, and one desk in between that holds no inventory and takes no position.

OEMs & ODMsfactory channelDistributorsallocationGPU clouds& neocloudsData centresspace & powerLogisticsfreight & IORAI teamstraining & inferenceEnterprise& HPCSovereign& government SUPPLY FLOWS IN · IT LEAVES AS ONE FILE · YOU CONTRACT WITH THE SUPPLIER DIRECTLY
Access policy

We turn work away.

A desk that never declines anything is selling something. Ours is not. Here is who gets a file opened, and who does not.

Open to

The party running the workload

Whoever's workload occupies the capacity and whose balance sheet stands behind the term, or an advisor they name in writing. That is who we act for, exclusively.

Structure

One layer, never two

You contract directly with the distributor, the cloud or the facility. We are not in the chain and we do not add one. An allocation shopped through four parties gets declined, by us and by the supplier.

Compliance

KYC before any introduction

Run first, not after the commercial terms. It is the reason suppliers quote our buyers real numbers instead of screening quotes.

Declined

Positions assembled to flip

Brokers and resellers without a nameable end user do not get a file. If the requirement is wrong for the desk, we say so in the first reply rather than take the meeting.

Three desks

Pick the one you are living in.
Or bring us the whole problem.

NVIDIA HGX node
Desk 01 · Hardware

GPU procurement

  • H100 to GB300, RTX 6000 and 5090
  • Direct OEM & ODM channel access
  • New, refurbished and cancelled orders
  • Priced landed, not ex-works
  • Freight, duties and import handled
Open the desk →
GPU cloud hall
Desk 02 · Service

GPU-as-a-service

  • Reserved, committed spend, bare metal
  • Terms sized to the workload
  • Agreements read line by line
  • An exit from the hyperscaler bill
Open the desk →
Data centre exterior
Desk 03 · Facilities

Colocation & data centres

  • Retail and wholesale, up to 100 MW
  • Global coverage, matched to requirement
  • Air, rear-door and liquid cooling
  • Space, power and cooling as one file
Open the desk →
Live inventory

On the desk this week.

Verified with the holder before it is listed. Prices are indicative, ex-works USD, and move with the market.

B300
NVIDIA B300 HGXUS · 7 days · from $563,000 / node
GB300
GB300 NVL72 buildUS · 25 days · from $5.9M / build
H100
NVIDIA H100 HGXUS · 7 days · from $240,000 / node
H200
NVIDIA H200 HGXEU · 10 days · from $395,000 / node
RTX 6000
RTX 6000 BlackwellEU · 15 days · from $13,000 / GPU
A100
NVIDIA A100 HGXUS · 7 days · from $125,000 / node
Open the full live inventory →

See a line that fits? Send node count and target region to the desk. Acknowledged within one hour.

Find your fit

Four kinds of buyer walk in here.

Neoclouds & GPU resellers

Big node counts, resell or partition, take-or-pay.

AI training & inference

H200 / B300 nodes, financing, an escape from the cloud bill.

Enterprise AI & HPC

Modest compute plus the whole stack around it.

Sovereign & government AI

Data residency, named facilities, compliance first.
Who pays us

Short version: not you.

Paid from the supply side.

Winning business carries a cost of sale that exists whether or not we are involved. That is where our compensation comes from, under disclosed arrangements.

No inventory, no position.

We earn nothing extra if you take one line over another. The only deal we push is the one that still makes sense in year three.

You own the relationship.

You contract directly with the supplier and keep that relationship outright. We stay on the file after signature.

Method

The same four steps, every requirement.

01

Qualify

What the workload actually needs, what it is worth, and the compliance file that lets suppliers quote it properly.

02

Match

The requirement goes to the suppliers whose position fits it, not whoever is loudest that week.

03

Negotiate

Landed numbers, not headline rates. You see the shortlist and the reasoning behind every name.

04

Deliver

Through freight, import and commissioning. The engagement ends when the workload runs.

Coverage

Capacity is a local problem
with a global market.

Global compute network
North AmericaEurope NordicsMiddle East Asia PacificAustralia Latin America

Power price, permitting and latency all move by region, so the right answer is rarely the market nearest you. Every country page carries the local picture.

Before you reach out

The questions people actually ask.

Researching rather than buying yet? The GPU Buyer's Playbook is the five-page version of this site, free. Or take the thirty-second router to the right desk.

Who pays you, and how does that not distort the advice?
The supply side does, under disclosed arrangements, because winning business carries a cost of sale that exists with or without us. What keeps it straight is arithmetic: we hold no inventory and no position, we earn nothing extra whichever line you take, and one bent number would cost us every deal after it.
Why not just go to the supplier directly?
You can, and you will get their number. You will not get the market's number. Price moves with scale, term, prepayment, power cost and how badly a supplier needs an anchor that quarter. We see those positions daily, so your requirement lands where it fits.
Do you work with brokers or resellers?
The desk acts for the end user or an advisor that party names in writing. An intermediary who can name and verify the ultimate offtaker gets a straight conversation; a position being assembled to flip gets a polite no.
Is my requirement too small?
Ask anyway. The hardware desk runs from single nodes to clusters. Service and facility engagements need enough scale to work on both sides, and if yours does not, we say so in the first reply.
How fast can you move?
Acknowledged within one hour, first sourcing pass within one business day.
Direct line

Tell us what is on the table.

Early, mid-negotiation, or just weighing options. Node count, workload and target region is enough to start.

See live inventory

Goes to the desk, not a marketing list. No obligation. KYC runs before any introduction, and you contract directly with the supplier.

✓

Sent. We are on it.

Your enquiry has landed with the desk. Acknowledged within one hour.