We cut your Azure bill by 20%, or we pay you A$50,000
Azure charges you for capacity you provisioned, not capacity you used: fixed VM sizes, App Service plans that bill around the clock, dedicated Synapse pools and Fabric capacity, and a hub and spoke network you pay to run in every region. We migrate, replatform and rearchitect your workloads onto Google Cloud, and we put a number on the result in writing.
The guarantee
If we hit 20%
Your annualised cloud run rate drops by at least 20% against your agreed Azure baseline.
If we miss 20%
Aviato pays A$50,000 to your business (AUD $50,000 cap). You keep the migrated platform and every design we produced.
For Azure estates spending A$50,000 or more per month. Eligibility confirmed in writing. Terms apply.
Where Azure spend goes, and what changes on Google Cloud
Structural differences, not a discount you lose at renewal.
Fixed VM sizes vs custom machine types
Azure Virtual Machines come in fixed D, E and F series sizes, so a service that needs 6 vCPU and 20 GB lands on the next size up and you pay for the gap. Compute Engine custom machine types let you set vCPU and memory to what the workload actually uses, across the whole fleet.
Discounts without a three-year commitment
On Azure, meaningful compute discounts come from Reserved Instances or a Savings Plan, which means committing one or three years ahead. Compute Engine applies sustained use discounts automatically on eligible machine types, up to 30%, with no commitment. Committed use discounts are there when you want them, not as the only route to a fair price.
AKS and App Service vs GKE Autopilot and Cloud Run
AKS, including AKS Automatic, bills you for the node VMs underneath your pods, so idle headroom is your cost. App Service plans bill for every provisioned instance whether it serves traffic or not. GKE Autopilot bills for the resources your pods request, and Cloud Run scales to zero between requests.
Synapse and Fabric capacity vs serverless BigQuery
Synapse dedicated SQL pools bill per DWU hour and Fabric capacities bill per capacity unit hour for as long as they are running. BigQuery separates storage from compute and scales slots up and down with demand, so a quiet night costs close to nothing. If Microsoft is already steering you from Synapse to Fabric, you are replatforming either way.
Regional VNets vs one global VPC
Azure VNets are regional, so multi-region estates end up with hub and spoke peering, per-GB peering charges on both sides of every link, and an Azure Firewall and gateway per hub. A Google Cloud VPC is global: one network spans every region on Google's private backbone, fronted by a single global load balancer IP.
Provisioned Cosmos DB vs pay per use databases
Cosmos DB provisioned throughput bills for the request units you reserve every hour, and autoscale never drops below a floor. Firestore bills per operation, and Spanner and AlloyDB give you relational scale without sizing for peak. We pick the target per workload during the assessment.
Azure services you stop paying for
The biggest savings come from Azure charges that have no equivalent on Google Cloud.
SQL Server licences
Azure SQL Managed Instance, SQL Database vCore and SQL Server on VMs. Removed where the database moves to Spanner or AlloyDB.
Windows Server licences
Removed for every application that moves from a Windows VM to modern .NET in Linux containers.
App Service plans
Always-on plan instances replaced by Cloud Run, which bills per request and scales to zero.
AKS node pools
Idle node headroom goes away on GKE Autopilot, where you pay for pod requests, not VMs.
Synapse dedicated SQL pools and Fabric capacity
Hourly warehouse capacity replaced by BigQuery, with storage and compute billed separately.
Cosmos DB provisioned throughput
Reserved request units replaced by Firestore pay per operation, or Spanner sized to real load.
Azure Firewall hubs and VNet peering
Per-region hub firewalls and per-GB peering charges replaced by one global VPC with Cloud NGFW policies.
Application Gateway and Front Door
Two layers of load balancing consolidated into one global external Application Load Balancer.
Azure Bastion and DDoS Network Protection
Identity-Aware Proxy provides admin access with no hourly host, and network-layer DDoS protection comes built into Cloud Load Balancing.
Modernise, don't just move
A lift and shift copies your Azure costs onto a new invoice. Most of the saving comes from these three modernisation paths.
SQL Server to Spanner
For systems that have outgrown a single database server, or that need to be always on.
- ✓Up to 99.999% availability SLA on multi-region configurations
- ✓Scales horizontally with no sharding and no read replica sprawl
- ✓No SQL Server licence, no patching, no maintenance windows
- ✓Online schema changes, and a PostgreSQL interface if your team prefers it
- ✓Sized in small processing unit increments, so you start small and grow
SQL Server to AlloyDB
For line of business applications that want PostgreSQL without giving up performance.
- ✓Fully PostgreSQL compatible, with no licence cost
- ✓Built-in columnar engine runs reporting queries on live data, often removing a separate reporting database
- ✓99.99% availability SLA, with read pools that scale independently
- ✓Database Migration Service converts schemas and T-SQL code, with Gemini assisting on stored procedures
- ✓Vector search built in for AI features on your own data
Azure VMs to containers
For the fleet of VMs that each run one app at a fraction of their capacity.
- ✓Many apps packed onto shared capacity instead of one VM each
- ✓Cloud Run scales to zero; GKE Autopilot bills per pod request
- ✓.NET Framework apps upgraded to modern .NET on Linux, removing Windows licences
- ✓No OS patching cycle, and deployments measured in seconds
- ✓Apps that cannot move yet stay on Compute Engine with custom machine types
Not every workload should be modernised. The assessment shows which ones pay back, and the guarantee is measured on the scope you approve.
Azure to Google Cloud service map
Every common Azure service has a Google Cloud target. This is the starting point we refine against your estate.
| Azure today | Google Cloud target | What changes |
|---|---|---|
| Azure Virtual Machines | GKE Autopilot or Cloud Run where the app allows, Compute Engine where it does not | Containers first; custom machine types and sustained use discounts for what stays on VMs |
| Azure Kubernetes Service (AKS) | GKE Autopilot | Pay per pod request, no node pools to size or patch |
| App Service, Container Apps, Functions | Cloud Run and Cloud Run functions | Scale to zero, no always-on plan instances |
| Azure SQL Database and Managed Instance | Spanner, AlloyDB, or Cloud SQL for SQL Server | Remove SQL Server licences where the workload allows, keep them only where they earn their cost |
| Cosmos DB | Firestore, Spanner or Bigtable | Pay per use or scale-out relational, chosen per workload |
| Synapse Analytics and Microsoft Fabric | BigQuery | Serverless warehouse, no capacity to pause and resume |
| Logic Apps | Workflows and Application Integration | Pay per step executed, no standard plan to keep running |
| Azure Data Factory | Dataflow, Cloud Composer, BigQuery pipelines | Pipelines that sit next to the warehouse |
| Blob Storage and Data Lake Storage | Cloud Storage | One bucket model with Autoclass tiering |
| Azure OpenAI and AI Foundry | Vertex AI with Gemini and Claude | Frontier models beside your data in BigQuery |
| VNets, Azure Firewall, Front Door | Global VPC, Cloud NGFW, Cloud Load Balancing and Cloud Armor | One global network instead of a hub per region |
| Azure Bastion | Identity-Aware Proxy TCP forwarding | Secure admin access without a jump host to pay for |
| Azure Monitor and Log Analytics | Cloud Monitoring and Cloud Logging | Logs queryable in BigQuery |
| Microsoft Sentinel and Defender for Cloud | Google SecOps and Security Command Center | Detection and response backed by Google threat intelligence |
| Entra ID | Stays in place, federated to Google Cloud | Same sign-in, same groups, no identity migration |
Choosing a database for Azure SQL or Cosmos DB
Whether you are keeping SQL Server on Cloud SQL, modernising onto AlloyDB for PostgreSQL, scaling globally on Cloud Spanner, or mapping Cosmos DB APIs to Firestore or Bigtable, these tools help you pick the target.
What about our Microsoft stack?
Moving your infrastructure off Azure does not mean moving your business off Microsoft.
Entra ID and Microsoft 365 stay
We federate Google Cloud to your existing Entra ID tenant. Your people sign in the same way, your groups drive access, and Microsoft 365, Teams and Intune are untouched.
Licensing is modelled first
Microsoft's licensing terms favour Azure, and Azure Hybrid Benefit is real money. We price it into the business case up front: which licences move, which workloads run licence-included, and which are cheaper modernised onto Linux and PostgreSQL.
.NET runs well on Google Cloud
Modern .NET runs in Linux containers on Cloud Run and GKE. .NET Framework applications run on Windows Server on Compute Engine or Windows node pools in GKE while you decide whether to modernise them.
Your Azure commitments are planned around
We sequence migration waves against your reservation expiry dates and any MACC or Enterprise Agreement terms, so you are not paying for the same workload on two clouds.
We have moved a global platform between clouds with zero downtime
99.999%
Production SLA
10 Mos
vs 2.5 Yr Estimate
0 min
Cutover Downtime
10,000
Containers in 10s
Hapana, a global fitness and wellness platform, came to us from AWS rather than Azure, but the problem was the same one Azure customers bring us: rising infrastructure overhead, slow deployments and fragmented backend services. Previous estimates put the replatforming at 2.5 years.
Aviato built a Terraform landing zone on Google Cloud, moved the backend microservices to Cloud Run, and kept member billing in sync with physical door controllers worldwide throughout. Production cutover landed in 10 months with zero minutes of customer downtime. The same playbook, landing zone and cutover approach is what we run for Azure estates.
How an Azure migration runs
Three phases, using Google's own Azure migration tooling.
Funded discovery and baseline
Read-only access to Azure Cost Management and Azure Monitor. We map dependencies, model licensing, agree your normalised 12-month baseline in writing and calculate the Google Cloud run rate. Aviato funds the assessment, so it costs you nothing.
Landing zone, identity and pilot
A Terraform Google Cloud foundation with project hierarchy, global VPC, Entra ID federation and security guardrails. We connect it to Azure with Cross-Cloud Interconnect or VPN and migrate a pilot workload with no production impact.
Production cutover
Migrate to Virtual Machines for Azure VMs, Database Migration Service for SQL Server, Storage Transfer Service for Blob Storage and BigQuery Migration Service to translate Synapse SQL. Continuous sync, pre-cutover validation and staged traffic shifting.
What the guarantee means in dollars
The minimum annual saving we contract to, at 20% of your agreed baseline. Set in AUD; other currencies shown for guidance.
Azure spend A$50,000 / mo
A$120,000
minimum saving per year
Azure spend A$100,000 / mo
A$240,000
minimum saving per year
Azure spend A$250,000 / mo
A$600,000
minimum saving per year
Miss the 20% and we pay you A$50,000 (AUD $50,000 cap).
We fund the assessment and guarantee the result
Aviato funds the upfront assessment: we review your Azure environment, map dependencies, model licensing and cost your target platform at no cost to you. If you go ahead and the migrated platform misses a 20% saving, we pay you A$50,000 (AUD $50,000 cap).
Guarantee terms and eligibility ▼
Eligibility: Commercial enterprises with a verified Azure infrastructure spend of at least AUD $50,000 per month over the preceding 3 billing cycles.
Baseline and assessment: The 20% reduction is measured against the previous 12 months of Azure spend, normalised for seasonality and agreed in writing before work starts. The client provides read-only access to Azure Cost Management and Azure Monitor, and allows Aviato to connect its assessment tooling.
Implementation scope: Aviato performs the migration to Google Cloud, including replatforming and rearchitecting. The client reviews and approves our designs and recommendations, and the guarantee applies to the approved scope. If the migrated platform does not deliver a minimum 20% annualised run-rate saving against the agreed baseline, measured 30 days after migration go-live, Aviato pays AUD $50,000 directly to the client entity.
Exclusions and limit: The guarantee does not cover growth in existing workloads, new workloads added after the baseline is agreed, or scope where recommendations were declined. Aviato pays at most one guarantee payout per client in any rolling 12-month period across all guarantee-backed engagements.
Assessment funding: Aviato funds the upfront assessment for eligible organisations. Funding for the migration itself is sought from Google Cloud where the workload qualifies and is subject to Google's approval.
Written approval required: Aviato confirms eligibility in writing before the guarantee attaches to an engagement. See the full terms and eligibility.
Questions we get asked
Is Google Cloud really cheaper than Azure?
For most estates we assess, yes, because you pay for capacity used rather than capacity provisioned: custom machine types instead of fixed VM sizes, automatic sustained use discounts, serverless BigQuery and Cloud Run instead of always-on capacity, and one global VPC instead of a hub per region.
Do we have to leave Microsoft 365 or Entra ID?
No. We federate Google Cloud directly to your Microsoft Entra ID tenant, so identity, single sign-on, Microsoft 365, and Intune device management stay exactly as they are while your infrastructure runs on Google Cloud.
What happens to our Windows Server and SQL Server licences?
We model licensing first during the funded assessment. Eligible licences can move, some workloads run licence-included on Google Cloud, and SQL Server databases can be modernised onto AlloyDB or Cloud SQL for PostgreSQL where economics favour it.
We have Azure reservations and a MACC. Is now the wrong time?
Usually not. We sequence migration waves around reservation expiry dates and commitment terms so you do not pay twice, giving you a verified alternative well ahead of your Microsoft renewal.
Will there be downtime during the cutover?
We plan every cutover for zero customer downtime, using continuous data replication, parallel validation before switchover and staged traffic shifting.
Talk to an architect who has done this before.
Bring your current setup and the outcome you need. You will get a view on the approach, the risks and roughly what it costs.
Straight to a senior GCP architect. No SDR, no slide deck.
Not ready to talk? See how we migrated Hapana off AWS →
Or call +61 2 8359 9507 · Hello@aviato.consulting