A glowing network-security illustration maps cloud data centers and connected servers across the United Kingdom.
Choosing a UK cloud region is really a question about which outage you need to survive. A London-only design with several zones can ride out a zone failure. It can't ride out the loss of the whole region. AWS, Azure and Google Cloud all offer London capacity, but they differ on zone counts, on whether a second UK region exists, and on how transfer charges and residency rules apply to a recovery design.

A caveat first. A JournalArta comparison published on 19 November 2026 notes that it could not find a like-for-like hourly Linux VM rate in the pricing views it reviewed. So there is no compute-cost winner here. What follows covers resilience, transfer pricing and data-access risk, which are the parts the published evidence supports.

A glowing network-security illustration maps cloud data centers and connected servers across the United Kingdom. Zones and regions: what each provider actually offers​

AWS London (eu-west-2). AWS announced on 19 August 2026 that it had added a fourth Availability Zone, eu-west-2d. The stated purpose is more capacity and fault isolation. Standard London pricing applies. AWS's Availability Zones documentation lists four UK AZ IDs for the region, euw2-az1 to euw2-az4.

AWS's own documentation also warns that this may not be uniform. It says that as zones grow, expansion can become constrained. AWS may then restrict new resources in a constrained zone, so different accounts can see different numbers of zones in a Region. Before you rely on a fourth zone, check the zones your account can actually use. The AWS documentation shows the CLI command aws ec2 describe-availability-zones, with a zone-type filter and a region argument. For London, the region is eu-west-2.

Azure UK South and UK West. Microsoft's region list shows UK South (uksouth) in London with three availability zones, paired with UK West. UK West (ukwest) is in Cardiff and shows no availability zones. That makes Azure the only one of the three with a documented second UK region paired to London.

Microsoft is blunt about what a pair does not give you. Its documentation says deploying into a paired region doesn't automatically provide high availability, disaster recovery or failover. You have to build and test your own plan. It also says not to treat Microsoft-managed failover as your main DR approach. For geo-redundant storage, that failover happens only in catastrophic situations and after repeated failed recovery attempts. Microsoft adds that even where a region has zones, some services may not support them there.

Google Cloud London (europe-west2). The JournalArta piece lists three zones, and Google's Compute Engine documentation says machine-type availability varies by zone. Google also recommends spreading workloads across multiple zones and multiple regions. Its documentation gives an example of backing up a region with instances in another region.

ProviderRegionZonesSecond UK region documented?
AWSLondon (eu-west-2)Four, announced 19 Aug 2026Not in the evidence reviewed
AzureUK South (uksouth)ThreeYes, UK West (Cardiff, no zones)
Google CloudLondon (europe-west2)ThreeNot in the evidence reviewed

Takeaway: More zones help with local faults. Only a second region addresses loss of London. If the policy says "stay in the UK," the second UK region is the deciding feature, and the evidence reviewed names only Azure's.

Where the transfer money goes​

Azure's bandwidth page gives the figures behind the article's arithmetic.

  • Internet egress: The first 100 GB a month is free. Between regions within Europe, Azure lists $0.02 per GB. The same page's FAQ says outbound data between two regions is charged at the normal rate, while inbound is free.
  • Premium network egress: Europe-sourced traffic over Microsoft's premium global network is $0.087 per GB for the next 10 TB a month. The transit-ISP routing-preference table lists $0.08 per GB.
  • Worked examples: At $0.02 per GB, 1,000 GB replicated from London to Cardiff is about $20. That is before storage and standby compute. 1,000 GB of public egress with the full free allowance gives 900 GB at $0.087, or $78.30.

Third-party trackers corroborate the $0.02 inter-region rate for Europe. InfoQ also noted that Microsoft still charges for inter-region transfers. That same coverage says Azure dropped charges between availability zones in 2024. For a multi-zone London design, that makes Azure's zone-to-zone traffic cheaper than on providers that still charge for it.

AWS says customers get 100 GB a month of free internet egress, aggregated across services and Regions. That allowance is for internet traffic, not replication between regions.

The Google figure of $0.12 per GiB for Cloud Storage egress to UK internet destinations comes from the JournalArta article. I could not independently confirm it. Third-party trackers quote a similar $0.12 first-tier figure for Google, but they are secondary sources. Treat it as a number to confirm against Google's current SKU for your route.

Do not rank providers on these figures. Azure's rate is per decimal GB and Google's is per binary GiB. They cover different services, allowances and routes. Ask every vendor to quote the same byte volume, source service and destination.

Currency and quoting​

Microsoft says its prices are calculated in US dollars and converted using London closing spot rates set at month end. Actual prices vary by agreement, purchase date and currency. Google directs non-USD customers to SKU prices in their billing currency. Record the currency and quote date with each rate so finance can reconcile it to invoices.

Residency is more than server location​

The JournalArta piece makes several useful points here.

  • Google storage locations: Google's EU multi-region does not include London. A London–Frankfurt dual-region includes Germany. If object data at rest must stay in the UK, use the London regional bucket.
  • Provider position statements: Microsoft says it may replicate data within the selected geography, with service-specific exceptions. AWS describes a commitment to customers' chosen Regions, subject to legal requirements and service terms. These are broad statements, not guarantees for every backup, log or support workflow.
  • Remote access: Under the UK ICO's test, access by a separate overseas legal entity can be a restricted transfer even when the servers are in the UK. The ICO's example is a UK business giving its US parent remote support access. Where a transfer is restricted, you need an adequacy regulation, appropriate safeguards or an exception. Relying on safeguards requires a transfer risk assessment. This is not legal advice.

Recovery exercises deserve the same scrutiny as production. Teams exporting diagnostics or routing telemetry through tools outside the production design can create a new access path.

A practical checklist before you sign​

  1. Set maximum acceptable downtime and data loss first.
  2. Decide whether you need to survive a zone failure or a London region failure.
  3. If London region loss must be survivable inside the UK, require every bidder to price a second UK site.
  4. For Azure, confirm the database, replication feature and VM capacity in Cardiff. UK West has no zones, and Microsoft warns that service support varies.
  5. For AWS, confirm which zones your account sees and whether your instance types are available in them.
  6. For Google, check machine type and feature availability in each London zone.
  7. Run a timed restore covering identity, secrets, DNS and application dependencies.
  8. Model storage and replication costs for normal running, and compute and outbound traffic for failover.
  9. Ask each provider to name the legal entity behind every support and operational access path.
  10. Award to the lowest total-cost bid that meets your recovery and residency requirements.

Bottom line​

On the evidence reviewed, Azure has the most direct documented route to a two-region UK design, and AWS has the most in-region zones. Neither fact settles the price. A cheap egress rate means little if the recovery site lacks the service you need. A fourth zone means little if the requirement is surviving a regional loss. The comparison has to be done on your own workload, with matched quotes and a tested restore.

 

References

  1. AWS London vs Azure UK South vs Google Cloud Costs and Residency - JournalArta JournalArta 2026-11-19T07:43:49+00:00
  2. Regions and zones | Compute Engine | Google Cloud Documentation docs.cloud.google.com
  3. Azure Eliminates Inter-Availability Zone Data Transfer Fees to Encourage Deployment Best Practices - InfoQ infoq.com