Cloud Costs in the UAE: Finding Charges Beyond Compute

A Dubai distributor moves its customer portal to the cloud and budgets for the application servers. Two months later, the server bill is close to plan, but the total is higher. The portal sends traffic through a gateway, stores copies of uploaded files and retains logs from a troubleshooting exercise. The recovery environment also has its own charges. The finance team needs to connect each line of the invoice to a configuration or activity somebody can explain.

Start with the detailed bill and trace the workload behind it. A useful review shows what the business pays to run, observe, protect and recover the service. It also distinguishes costs that rise with activity from commitments that will continue even if a workload is retired. Use that evidence to decide which charges are expected, which need investigation and which can be reduced without weakening the service.

Reconcile the bill before looking for savings

Collect the invoice, the detailed usage export and the contract or pricing schedule for the billing period. Include accounts paid through a reseller or another department. Record the billing currency, the rate used in the AED budget, credits, discounts and applicable tax treatment as confirmed by finance. Identify what the detailed export covers. A report that excludes support or a marketplace subscription can be useful for engineering while still falling short of the total paid by the company.

Tie each material charge to the account, subscription or project, service, meter, unit and period. Note the region where the provider reports it, if available. Distinguish charges for current use from commitments, refunds or adjustments from an earlier period. Separate the timing of a payment from the cost attributed to a workload. An annual subscription paid this month should appear in the cash forecast, even if the organization spreads its cost across the year for management reporting.

A UAE Ministry of Finance carousel highlights federal digital procurement. For a private business reviewing a cloud bill, the relevant discipline is to follow the commercial record from the order and contracted terms to the invoice. Ask finance to verify the amount paid and engineering to verify the usage. The two records answer different questions.

Follow data through the network

Draw the route for a common operation such as downloading a catalogue or uploading proof of delivery. Identify the components that process traffic, where it crosses a network boundary and where a copy goes next. A private application server might make an outbound call through a gateway. A large file might also pass through a load balancer, be copied to another environment or be downloaded by an external user. Each step can have a different meter depending on the provider and configuration.

AWS's virtual private cloud documentation describes NAT gateway charges based on availability time and the amount of data processed. That distinction helps explain why an otherwise quiet environment can still have a charge and why the charge may rise with traffic. Check separate data transfer terms as well as the gateway meter, and use the actual region and contract when calculating a change. AWS NAT gateway pricing documentation

Ask whether a billed path is necessary before changing it. A data transfer between parts of an application may be important to recovery or isolation. Moving both components into one failure boundary can lower a charge while making the service less resilient. If a private endpoint is proposed, compare its fixed and variable charges against the current route. Check that the endpoint supports the service and fits the access design. Test the actual route and inspect the following usage export before recognizing a saving.

Count storage activity and recovery separately

An application can incur storage charges even if the amount of original customer data changes slowly. Check how many copies exist across snapshots, replicated buckets, test environments and exports. Confirm which system creates them and how long each is retained. Identify data the business must preserve and data it can safely remove. A snapshot with no clear owner may still be needed for recovery or for a defined retention period.

The Amazon S3 pricing page separates categories such as storage, requests, retrieval and data transfer. Use the relevant service terms when considering a different storage tier. A lower price for storing a copy can come with different terms for retrieval or minimum storage duration. For a UAE business, compare the charge for preserving the records with the cost and time required to read them when they are needed. Test a representative restore before changing how critical files are kept. Amazon S3 pricing categories

Price recovery for an ordinary month and for a real event. The first includes retained copies and standby capacity. The second can include temporary infrastructure, data reads, transfers and extra support while the team validates the restore. Ask which resources need to be provisioned at the recovery destination and which are kept ready. Keep the expected cost of an exercise visible even if it happens only a few times a year.

Check what the business is paying to observe

Logs can grow faster than customer traffic. A release may add a field to every request or turn on debugging. A service can send similar data to two tools because different teams configured them. Find the environment, table or source behind the charge. Compare the start of a spike with release, incident and configuration records.

Microsoft's Azure Monitor cost guidance discusses data ingestion and retention as separate cost drivers. Look at both before changing the setup. A noisy source can raise ingestion even if the retention policy stays the same. A longer policy can increase the stored volume without any new traffic. Ask security and operations which records they rely on and the period they need them. Change the scope or detail where the owners can validate that incident investigation and required retention are preserved. Microsoft Learn on Azure Monitor cost and usage

Compare production and test. A test environment might process duplicate feeds or retain logs as long as production even though its purpose is short-lived. Confirm the intended behavior before applying different settings. Also check who pays for central monitoring, identity and security tools. Their bills can sit in a shared account while the application budget assumes they are included elsewhere.

Allocate shared costs and recalculate the unit

FinOps Foundation guidance describes allocation using account or project information and methods for distributing shared costs, including fixed or proportional splits. Choose a driver the finance team can explain, such as bytes processed for a shared data service or active seats for a common subscription. If a driver is unavailable, label the split as provisional and keep unassigned spending visible. Check that all allocations add back to the source invoice. FinOps Foundation allocation guidance

Suppose the hypothetical distributor spends AED 12,000 on compute, AED 5,000 on networking, AED 4,000 on storage and backups, AED 3,000 on monitoring and AED 2,000 on allocated support. The monthly total is AED 26,000 before tax and internal labour. Non-compute charges account for AED 14,000, about 54% of the total. At 20,000 correctly completed orders, the cloud bill is AED 1.30 per order. The example uses illustrative amounts, not published rates.

Now suppose the next month still completes 20,000 orders. A change in the route adds AED 2,000 in network charges and debug data adds AED 1,000 in monitoring. The bill rises to AED 29,000, or AED 1.45 per order. A drop in the server price would not explain that movement. The team can trace the higher cost to the new path and the additional data, then decide whether either change is worth keeping. Include failed attempts in the spending and count only completed orders in the denominator so the metric reflects the cost of producing usable work.

Use the same method when volume changes. If the AED 26,000 bill stayed flat while completed orders dropped to 16,000, the unit cost would rise to about AED 1.63. No billing meter needs to increase for that to happen. Keep order volume and the level of service beside the unit metric so the business can distinguish lower demand from a technical cost problem. Also note changes in order size or file volume if an order in one period uses substantially more infrastructure than an order in another.

Turn the first review into a monthly control

Assign an owner to each material meter and agree what change requires investigation. A network charge can move with customer traffic, an invoice can increase after a credit expires and storage can rise because backups were retained longer than planned. Compare usage and price separately. Look for when a change first appeared and what deployment, contract event or business activity coincided with it.

Before acting, record what the proposed change would affect. Shortening log retention might be reversible as a setting, but records already deleted may be unavailable for an investigation. Removing a backup copy may reduce both cost and recovery options. Test alternatives with the owner and set a review date. Validate the result against a complete billing period and confirm that the workload still passes its operational checks.

After an investigation, save the query or export used to identify the charge. A second reviewer should be able to run it for a different period and trace the same meter to the invoice.

Review commitments alongside consumption. Ask when a support contract renews, whether a marketplace subscription is billed separately and which charges continue after a resource is removed. Give each owner a date to confirm the service is still needed. In a reseller arrangement, request enough detail to match the supplier's invoice to provider usage and identify separately priced support. If the detail is unavailable, record that gap in the next contract review.

The distributor can use the first review to build a recurring bill that separates baseline charges, volume changes and one-off events. Its next decision might be to correct an unnecessary route, reduce debug volume or simply budget for the cost of recovery. The evidence should show why each charge exists and what the business would give up if it changed the service to remove it.