Strategy / single vs multi-cloud costs
Multi-Cloud Cost Management in 2026: when it makes sense and what it costs
Multi-cloud is trendy and often expensive. The one overhead you can price from list rates is cross-cloud egress: the sending cloud bills every byte, at $0.09/GB from AWS us-east-1, $0.087/GB from Azure and $0.12/GiB from GCP Premium Tier (first-tier rates, checked 9 Nov 2026). This page covers where the rest of the overhead comes from, the use cases that justify it, and the FOCUS specification that makes multi-cloud cost reporting tractable.
Six categories of multi-cloud overhead
Where the multi-cloud premium comes from
| Category | Cost impact | Why it shows up |
|---|---|---|
| Duplicate tooling | A second set of licences or instances | Monitoring, security, identity, CI/CD often need cloud-specific configurations or whole separate instances. |
| Team expertise | Broader hiring and training | Engineers need depth across multiple clouds. Recruiting Azure + AWS + GCP expertise is harder than depth in one. |
| Cross-cloud data transfer | Sender pays internet egress: AWS $0.09/GB, Azure $0.087/GB, GCP $0.12/GiB | First-tier list rates (AWS us-east-1 and Azure first 10 TB, GCP Premium Tier first TiB), checked 9 Nov 2026. Replication, cross-cloud API calls and federated queries all pay it on every byte that leaves the sending cloud. |
| Fragmented commitments | Each cloud's commitment sized separately | Savings Plans, reservations and CUDs only discount usage on the cloud they were bought on, so each estate needs its own forecast and its own unused-commitment buffer. |
| Different billing models | Tooling and analyst time | FOCUS specification helps, but tagging, allocation rules, and chargeback still vary by provider. |
| Licensing gotchas | Variable | Some software licences (Oracle, SAP, Microsoft) are priced or permitted differently by cloud. Check the licence terms before placing a workload. |
Six legitimate multi-cloud cases
When multi-cloud is the right answer
Regulatory data residency
Data must stay in specific jurisdictions where one provider has stronger presence
Example: EU customer data on Azure (broad EU region coverage) plus US SaaS workloads on AWS
Best-of-breed services
Specific service genuinely outperforms competitors
Example: BigQuery on GCP for analytics, AWS Lambda for event processing, Microsoft Entra ID for identity
Disaster recovery
Cloud-level outage protection for tier-0 services
Example: Active-active across two clouds for payment processing or critical SaaS
Acquisition integration
Inherited cloud accounts that are not worth migrating immediately
Example: Acquired company on GCP plus parent on AWS, migrated on a planned timeline
Vendor lock-in mitigation (genuine)
Real strategic risk, not theoretical
Example: Enterprise with regulatory or geopolitical concerns about single-vendor concentration
Cost arbitrage at scale
Specific workloads materially cheaper on a specific cloud
Example: GPU training workloads where one provider has lower list rates or available capacity (see the GPU table on /compute)
Reference architecture / FOCUS as substrate
FOCUS is the substrate that makes multi-cloud cost reporting work in production without per-provider transformation pipelines.
AWS and Azure publish FOCUS-formatted cost exports, Google Cloud offers a FOCUS view over its BigQuery billing export, and Oracle Cloud publishes FOCUS reports. Loading them into a single warehouse (BigQuery, Athena, Snowflake) lets a single query run across providers with consistent column names and units.
Multi-cloud cost tools
Tools that span providers
FOCUS specification (open standard)
Free
A common schema for billing data from AWS, Azure, Google Cloud and Oracle Cloud. The substrate for any multi-cloud cost analysis.
Vantage
Flat monthly tiers by tracked spend: Starter free (to $2,500/mo), Pro $30/mo, Business $200/mo, Enterprise custom
Multi-cloud cost reporting, anomaly detection, native FOCUS support, Kubernetes allocation.
CloudHealth (Broadcom)
Not published: quote
Multi-cloud governance, policy enforcement, chargeback, established at large enterprises.
Apptio Cloudability (IBM)
Not published: quote
Multi-cloud TBM, business mapping, executive reporting.
Spot (Flexera)
Not published: quote (spot.io now redirects to flexera.com)
Automated Spot optimisation across providers. Strong for Kubernetes.
Finout
Yearly fee set by cloud spend tier, quote only; no per-seat or percent-of-bill fee
Multi-cloud cost allocation with strong shared-cost (Kubernetes, networking) handling.
Holori
See Holori's pricing page
Provider-agnostic cost dashboards with focus on smaller teams.
Worked example / cross-cloud transfer
What moving 1,000 GB a month between clouds costs at list rates
Inbound data is free on all three providers; the cloud the data leaves bills it as internet egress. Multiply your own monthly volume by the sending cloud's rate. Free allowances and volume tiers above the first tier are not applied here.
| Sending cloud (list rate tier) | Rate | 1,000 GB a month (GCP: GiB) |
|---|---|---|
| AWS (us-east-1, first 10 TB/mo) | $0.09 per GB | $90 |
| Azure (Microsoft network routing, first 10 TB/mo) | $0.087 per GB | $87 |
| GCP (Premium Tier, first TiB/mo) | $0.12 per GiB | $120 |
Sources: AWS Data Transfer price list (us-east-1), Azure Retail Prices API (Bandwidth, East US 2), Google Cloud network pricing page. Checked 9 November 2026. AWS and Azure price per GB, Google per GiB; 1,000 GiB is under 1 TiB, so the GCP row stays in its first tier.
Common questions
FAQ
How much does multi-cloud actually add to my bill?+
There is no published industry percentage, and it depends on how tightly the clouds are coupled. The parts you can price are list rates: every byte one cloud sends to another is billed as internet egress by the sender (AWS $0.09/GB in us-east-1, Azure $0.087/GB, GCP Premium Tier $0.12/GiB at first-tier rates, checked 9 Nov 2026), so 1,000 GB a month replicated from AWS costs about $90 before any free allowance. The rest (a second tooling stack, broader hiring, commitments sized per cloud) depends on your own contracts and team.
When does multi-cloud actually save money?+
Rarely on cost alone. The legitimate cost-saving cases are narrow: GPU training where one provider has lower list rates or available capacity, analytics where one managed service suits the workload far better, or capacity a single provider cannot supply. For typical web and SaaS workloads, a single cloud avoids cross-cloud egress and duplicate tooling.
Should I run active-active across two clouds for resilience?+
Only for tier-zero workloads where cloud-level failure is unacceptable. Examples: payment processing, public-facing SaaS where downtime triggers contractual penalties, or regulated services. For everything else, multi-region within a single cloud covers most failure modes with far less operational complexity.
What is the FOCUS specification and why does it matter for multi-cloud?+
FOCUS is the FinOps Open Cost and Usage Specification, ratified at version 1.4 in June 2026 (1.3 in December 2025). It gives billing data from AWS, Azure, Google Cloud and Oracle Cloud common column names and units, so a single SQL query can run across providers. Without FOCUS, multi-cloud cost analysis requires per-provider transformations and constant maintenance.
Does multi-cloud weaken vendor negotiation leverage?+
Sometimes. The theoretical multi-cloud argument is that you can move workloads to negotiate harder. In practice, sales teams know if you are bluffing. Real leverage comes from genuine alternatives: actual workloads running on a competitor cloud at scale, or a credible architecture plan to migrate. Splitting spend also splits the commitment you can offer each provider.
Continue reading