The constraint is not space. It is power.
AI has changed what a data centre requirement looks like. Nobody is asking for cabinets any more; they are asking for megawatts, at densities most of the existing estate was never built to carry, on dates that do not move. We broker the capacity that can actually take it.
We broker capacity for the hardware we also sell.
Most colocation brokers have never racked a GPU node. Most GPU resellers have never negotiated a power contract. The gap between those two facts is where AI deployments go to die.
EOGSB does both. When we ask an operator whether a 10U air-cooled 8× B200 node at 14 kW is going to work in their hall, we already know what the answer needs to be, because we are the ones shipping the node. We know what a 5.2 MW cluster does to a facility's cooling envelope, because we have designed the cluster.
That is not a marketing line. It is the reason a client can give us one requirement and get back one delivery, instead of managing a hardware vendor, a freight forwarder and a colocation broker who are each optimising for something different.
Marketing Services Agreements
We hold MSAs with operators across our markets. In practice, that means we tend to see new allocation and space releases early — and frequently get first refusal on replacement capacity before it is generally marketed.
Cooling literacy
Air, rear-door heat exchanger, direct-to-chip liquid, in-row CDU. We qualify facilities on what they can actually carry, not what their brochure says.
Tier III or nothing
We do not shortlist Tier II facilities for production AI workloads. Concurrent maintainability is not a nice-to-have when a training run has been going for three weeks.
Five numbers, and we can start.
Send us these and we will come back with real options and real terms — not a calendar invite for an intro call.
Total IT load
In megawatts, and stated as IT load — not total facility power. If you are not sure which one your number is, that is fine. We will help you establish it. It is the difference that breaks budgets.
Density per rack
kW per cabinet, and whether the design is air-cooled or liquid. A 10U air-cooled 8-GPU node and a 4U direct-to-chip node are two entirely different facility conversations.
Market and jurisdiction
Where it must sit, and why. Latency, data residency, export control, tax, incentive regimes and grid carbon intensity all put real constraints on the shortlist.
In-service date
The date the first node has to be earning. Everything reverse-engineers from this: allocation, freight, customs, fit-out, cross-connect and burn-in.
Term and growth
Initial term, expansion rights, and how much headroom you want reserved. Contiguous growth space is the single most valuable thing to negotiate for at signature, and the hardest to get later.
We issue the RFI
Under NDA, to a qualified shortlist, with the power basis, usable kW, cross-connect model and remote-hands terms forced into the open before anyone talks price.
Data centre brokerage, answered.
Availability, pricing and terms are facility-specific and move constantly. Nothing on this page is an offer or a guarantee of capacity.
What does a data centre broker actually do?
We sit between the party that needs space and power and the facilities that have it. We take a requirement — kW per rack, total load, density, cooling, jurisdiction, in-service date — and we run it against the operators we hold relationships with, then come back with real options: usable kW, cooling architecture, cross-connect and bandwidth model, commercial terms, and the honest gotchas. We negotiate, and we stay in the deal through fit-out and installation.
Who pays EOGSB — the client or the facility?
The commercial model varies by facility and by deal structure. Where EOGSB is engaged under a Marketing Services Agreement with the operator, the operator carries the fee. Where the client engages us directly for advisory or as principal on a full supply-and-host delivery, the arrangement is set out in writing before we start. Either way, you will know exactly how we are paid before you commit to anything.
Do you only broker for high-density GPU deployments?
That is our specialism and it is most of what we do — 40 kW to 130 kW+ per rack, air and direct-to-chip liquid. We also handle conventional enterprise colocation, but the reason clients come to us is that most brokers cannot speak fluently about CDUs, rear-door heat exchangers, ORV3 busbars and what a 5.2 MW IT load actually means for a facility.
Is 5.2 MW IT load or total facility power?
That question is exactly why you want a broker who has been through it. The difference is roughly 25–30% of the number, and it is the single most common source of blown budgets in GPU colocation. We force it to be answered in writing in every RFI we issue, before anyone quotes.
Which markets do you cover?
The United States, Malaysia and Indonesia, with wider ASEAN coverage through partner facilities. We hold Marketing Services Agreements with operators in these markets.
Send us the requirement. We come back with options.
Node count, GPU class, kW per rack, target market, in-service date. That is enough for us to open the conversation with the right suppliers and the right facilities. Under NDA from first contact.