In most Saudi IT departments, "colocation versus cloud" is argued like a matter of faith: the infrastructure veterans defend their racks, the newer engineers assume cloud won years ago, and the CFO just wants one number. All three are working from the wrong frame. Colocation and cloud are not generations of technology where one replaces the other — they are different procurement models for the same physics, and each wins under conditions you can actually write down.
This is the framework we use with clients. It ends with the five situations where colocation genuinely wins — because a comparison that always concludes "cloud" is marketing, not analysis.
What you are actually buying
Colocation: you rent space, power, cooling and network in someone else's data centre, and you own everything inside the rack — servers, storage, switches, their firmware, their failures and their refresh cycle. The facility keeps the lights on; the computing is your problem and your asset.
Cloud: you rent the computing itself as a service. The hardware, the facility, the redundancy and the 3 a.m. disk swap are the provider's problem; you consume capacity through a panel or an API — for example a Saudi-based cloud server provisioned in minutes rather than a purchase order that takes a quarter.
Everything else in the debate — cost, control, compliance, speed — is downstream of that ownership line.
The money: CAPEX vs OPEX, honestly
Colocation is a CAPEX-plus-OPEX model. Up front: servers, storage, network gear, installation, cabling and burn-in — hardware you will refresh roughly every four to six years, which means the purchase is not once but perpetual, in waves. Monthly: the colo fee (typically priced per rack or per kilowatt), circuits and IP transit, hardware support contracts, spares, and remote-hands fees when you need the facility's technicians.
Cloud is nearly pure OPEX: a monthly fee, sized to the service. No refresh cliff, no depreciation schedule, no capital tied up in metal — but also no asset on the books and a bill that never ends.
The honest arithmetic: at high, steady utilisation over a full hardware lifecycle, owned hardware in colocation can come out cheaper per unit of compute — that is precisely why large, predictable estates keep doing it. At low or variable utilisation, colocation means paying capital costs for idle capacity, and cloud wins without a contest. The break-even is not a universal constant; it depends on your utilisation curve, your cost of capital, and — the item that decides more cases than any spreadsheet — your staffing.
Staffing: the line everyone underestimates
A colocated estate needs people who can specify hardware, rack and cable it, patch firmware, run virtualisation, manage backups, respond to failures at any hour, and drive to the facility when remote hands are not enough. Genuine 24/7 coverage cannot be one heroic engineer; it is a rota, which means several qualified people or an outsourced operations contract. In Saudi Arabia's current market, senior infrastructure engineers are in high demand across the very projects building the Kingdom's data centres — recruiting and retaining them is a real cost and a real risk, whatever the salary line says.
With cloud, that labour is embedded in the fee. This is the quiet reason small and mid-sized Saudi companies land on managed cloud: not because racks are wrong, but because the team a rack deserves costs more than their entire infrastructure budget. If you cannot staff the rota, you do not have a colocation option — you have an outage schedule.
Compliance and control
Both models can satisfy Saudi data-residency expectations, since both keep data physically in-Kingdom. The differences are in texture:
- Physical control. Colocation gives you custody of the actual disks — relevant for a small set of regulated or contractual requirements (dedicated HSMs, chain-of-custody rules, hardware your regulator or client insists you own).
- Regulatory perimeter. Cloud providers serving the Saudi market fall under CST's cloud regulatory framework, and the NCA's Cloud Cybersecurity Controls split duties between provider and tenant — our NCA CCC guide maps that shared-responsibility line. In colocation, nearly the entire control stack above the concrete is yours to implement and evidence.
- Audit effort. Ironically, colocation often means more compliance work, not less: you inherit only the facility's certifications, and everything from OS hardening to backup policy is yours to prove. See our PDPL hosting guide for how residency obligations attach to configurations, not buildings.
Latency and connectivity
Inside the Kingdom, this is close to a tie: a workload in a Riyadh colocation facility and one in a Riyadh cloud region both reach Saudi users in single-digit-to-low milliseconds. Colocation gives you one genuine edge — you choose the facility and its carriers, which matters for telecom-heavy architectures and private interconnects. Cloud gives you the opposite edge: presence in multiple cities (and abroad) without building anything. If you need Jeddah and Dammam footprints next quarter, that is a click in a cloud panel and a construction project in colocation.
The five cases where colocation genuinely wins
- You own significant, young hardware. If last year's capital purchase has four years of useful life, migrating to cloud means paying twice. Sweat the asset; revisit at refresh time.
- High, steady, predictable utilisation at scale. Render farms, core databases pinned at high load, telecom workloads — the elasticity premium of cloud buys you nothing if the load never varies.
- Special hardware. GPUs you own, HSMs, licensing dongles, appliances, or anything a regulator or enterprise client requires on owned equipment.
- Licensing economics. Some enterprise licenses are dramatically cheaper on controlled physical cores — a calculation the Broadcom era has made both more painful and more important, as we unpack in our VMware exit options guide.
- Deep control requirements. When your security model requires knowing exactly which humans can physically touch the disks, custody beats contracts.
If none of these five describes you, the burden of proof flips: cloud is the default, and colocation needs a specific justification.
The pattern most Saudi companies actually land on
Not either/or — a hybrid with a small core: owned or colocated hardware for the one or two workloads that genuinely qualify under the five cases, and managed cloud for everything else (web, email, files, apps, dev/test, DR). The mistake worth avoiding is the accidental hybrid — a rack kept alive out of habit, hosting workloads that would cost less and fail less on a SAR 199/month plan.
| Factor | Colocation | Cloud |
|---|---|---|
| Up-front capital | High (hardware + setup) | None |
| Monthly cost shape | Facility fee + support contracts | Single predictable fee (managed) or usage-based |
| Staffing needed | 24/7-capable infra team | Provider's problem |
| Scaling speed | Weeks to months (procure, rack) | Minutes |
| Physical control | Full | Contractual |
| Compliance burden | Mostly yours | Shared with provider |
| Strongest fit | Steady heavy loads, special hardware | Variable loads, small teams, speed |
Two ways we can help — from both sides
Because our parent company builds server rooms and data-centre infrastructure for clients across the Kingdom, we sit on both sides of this decision without needing to force it. If your five-case analysis points at owned infrastructure, the server infrastructure team designs and builds it properly. If it points at cloud, Skyline Cloud's Saudi data-centre-hosted services get you there without a single purchase order — start with the free 14-day trial, no credit card, and benchmark a real workload before deciding anything.
Frequently asked questions
Is colocation cheaper than cloud in Saudi Arabia?
At high, steady utilisation over a full hardware lifecycle, and with a team already in place — it can be. For variable or modest workloads, once staffing and hardware refresh are counted honestly, cloud is usually cheaper. The staffing line decides more cases than the hardware line.
Does colocation satisfy PDPL data residency?
A Saudi facility keeps data in-Kingdom, which addresses the residency dimension — but PDPL obligations (security, rights, breach notification) still apply to you as the operator of everything in the rack.
Can I move from colocation to cloud gradually?
Yes, and that is the sensible route: migrate workload by workload at hardware refresh points, keep the qualifying core colocated, and connect the two. Test the cloud side with a trial account before committing the first real workload.
What happens to my hardware if I exit a colocation contract?
That is governed by your contract's exit terms — notice periods, decommissioning assistance, and equipment removal windows. Negotiate them at signing, not at exit; our data-centre provider checklist covers the clauses.
Decide on paper, not on faith
Run the five cases against your actual estate, price the staffing rota honestly, and let the workloads sort themselves into the right model. And since one side of the comparison can be tested for free in an afternoon — open a Skyline Cloud trial account (14 days, no credit card), deploy something real, and give the cloud column of your spreadsheet actual numbers instead of assumptions.
Comments
0 total · 0 threads