Data-centre work is rarely a “which model to buy” question. What actually blocks you is three things: how in-rack power and density are allocated, whether the optics and switches match, and how data is accounted for after old machines leave.

These three belong to different trades, usually different suppliers. We take only a few segments — equipment, optics, clearance, outsourcing and on-site support — and tell you directly which segment is outside our scope.

What this page does not cover

Boundaries first, to save later rounds.

  • We do not rent cabinets or resell power. We are not a data-centre operator; we do not provide cabinet space, and do not commit to kW quota or power redundancy.
  • We do not arrange licences, or sign property／operator contracts on your behalf. Such arrangements are handled directly between you and the licensed party.
  • We do not call ourselves any operator’s partner or official reseller. We have no agency relationship with Cisco, Juniper or Huawei, and do not represent any room side.
  • We do not paste other people’s client lists. Someone else’s customers are not our credentials.

Pressure on Hong Kong rooms now (public background)

The following is compiled from public information and industry observation to explain why room-engineering cadence has changed in recent years; it is not our client list, and does not represent any agency or partner relationship.

Cluster locations. Hong Kong data centres concentrate mainly around Tseung Kwan O, Kwai Chung and Tsuen Wan, forming several relatively tight clusters. Concentration has benefits (short fibre paths, easy cross-connect) and costs (power and cooling resources compete in the same area).

New supply and land. Data-centre land around Sha Ling was mentioned in public documents and reports around March 2026; such projects typically take years from proposal to completion and will not change existing cluster supply and demand in the short term.

Pressure from AI density. The clearest change in recent years is not the number of cabinets but power and interconnect density per cabinet. The same row that once held several 1U servers may now hold two high-power hosts plus a high-speed switch. The result: in-rack airflow must be redone, power allocation recalculated, and fibre and optics multiplied several times. Interconnect selection is on the AI／GPU high-speed networking primer; power and airflow checks before buying gear are on the rack power and cooling primer; on-prem GPU host checks are on the on-prem／private GPU server checklist.

The above is compiled from third-party public information and industry observation; it is not a client list and does not represent any agency or partner relationship. Market figures go out of date; actual supply and prices follow the quote you receive and the operator’s information.

Four things we can handle from your list

All four are quoted against the list: you give model, quantity, location and window; we reply with available／not available／when／how much. No brand story first.

1. Compatible optics

When a room expands, what runs out first is usually not the switch but the module. Our compatible optics are paired to your platform and reach; before dispatch we confirm three things: switch port spec, fibre type (multimode／single-mode), actual distance.

Warranty basis: compatible optics one year, not Lifetime. It is a compatible product, not an OEM module; this is stated on the quote, not glossed over.

Models and compatibility conditions are on the compatible optics page.

2. Used and refurbished Cisco／Juniper

For expansion, cold spare and identical-model replacement, used and refurbished gear usually suffice. We quote per unit by PID and serial, listing accessories (power, fans, modules, rack parts) item by item.

Warranty basis: used／refurbished generally 3–6 months. New gear follows OEM terms, generally one year, per the OEM.

Note on licence and support: Smart Net, Juniper Care and some software licences may not transfer; the support state of used gear must be checked unit by unit, not assumed to “probably transfer”. The judgement method is in the used procurement guide.

Selection and inspection detail are on the two procurement pages: used Cisco, used Juniper. If you want to validate setup in your own environment first, see the lab.

3. ITAD clearance and decommissioning

The other half of swapping gear is how the old machines leave. We handle: equipment inventory, valuation, dismantling and packing, storage-media data handling, disposition records.

There are three things we will not pretend to do: pre-issue a fixed destruction-certificate template, guarantee data is “definitely unrecoverable”, or commit to completing a valuation within a set number of hours. What we can do is write method, date, executor and disposition into a record, so you can account to finance, legal and property.

The transaction side is on the clearance and ITAD buyback page; how projects are scheduled (timeline, four-department responsibility, five documents) is in the room clearance and ITAD practical guide.

4. Network engineering outsourcing and remote-hand

The room is in Tseung Kwan O, the engineer is in Kwun Tong, and the gear waits for someone to plug cables on site — this is the most common outsourcing and remote-hand scenario.

We take work by scope: installation and setup, cutover windows, routing such as VPN／BGP, automation scripts, on-site support. Remote-hand means your people are on site and we guide them remotely; decision rights stay with you — we provide remote guidance, not on-site hands.

Scope and liability boundary are on the network engineering outsourcing and remote-hand page. We do not fix a price, because the same “rack it for me” can differ threefold in hours between rooms.

Prepare before quoting

With five items ready, the reply usually lands in one round; missing one means another round of questions.

  • Region and site: which district, which building, whether access badge and booking are needed; any freight-elevator time limits.
  • PID or list: equipment by PID plus quantity, not just series name; clearance by model, quantity, and whether storage media are included.
  • Purpose: expansion, replacement, cold spare, or whole-batch decommission. For the same batch, purpose changes the quote structure completely.
  • Window: workable dates and hours (day／night／holiday), and any maintenance freeze period that cannot be touched.
  • Storage media: quantity, type, whether already removed. This directly affects clearance procedure and records.

Send the list directly to [email protected], or through the contact and quote page; both reach the same team.

Which roles this work usually falls to

This is not a recruitment ad; it is the division we actually see in projects. Recognising which box you sit in makes it easier to judge who to approach.

  • Room engineering: racks, power allocation, airflow, cabling and labels. We take the equipment and cabling segment; power and rack space are handled by you and the operator.
  • Network: switching, routing, cutover windows, VPN and BGP. We can execute on site; decisions and change approval stay on your side.
  • Procurement: PID, quantity, lead time, warranty and terms. We deal with this segment directly; the quote lists item by item.
  • ITAD: inventory, valuation, storage-media handling, record archiving. We do execution and records; compliance requirements are set by you.
  • NOC: duty, alerts, incident handling. We do not sell 7×24 monitoring outsourcing, but can take over hardware replacement and on-site troubleshooting after an incident.

If you want to join us rather than find a supplier for a project, see careers.