Prepare and test your skills

Prepare and test your skills

Capital expenditures (CapEx) and operational expenditures (OpEx) are two different ways to pay for IT. CapEx means buying physical infrastructure upfront before it is used, while OpEx means paying for resources as they are consumed. Moving from on-premises to the cloud shifts spending from fixed CapEx to variable OpEx, which changes how total cost of ownership (TCO) (opens in a new tab) is calculated over time.
CapEx is spending on physical infrastructure that the organization owns and must plan in advance. In an on-premises data center, this includes servers, storage, routers, switches, network cabling, and software licenses. The organization also carries the ongoing cost of owning that data center, such as rack space, power, cooling, maintenance, and IT staff salaries. Because the hardware is purchased for peak capacity, much of this spending is fixed and must be paid whether or not the capacity is fully used.
On-premises TCO models estimate these ownership costs with specific inputs:
Capacity assumptions also affect how much hardware must be bought. These include CPU burst, memory burst, vCPU to core ratio, and storage redundancy policy, which determine how much raw capacity is required to support virtual workloads.
OpEx is spending that follows actual consumption, so cloud resources are billed as they are used. Google Cloud breaks usage costs into components such as Compute, Storage, Network, Operation, and operating system (OS) licenses. This makes many cloud costs variable because they rise when demand rises and fall when demand falls. Autoscaling strengthens this effect by adding resources when traffic increases and removing them when traffic decreases, which helps avoid paying for idle overprovisioned capacity.
Google Cloud also offers options that lower the OpEx rate without returning to ownership:
These options change the price paid for consumption, but the cost remains tied to usage or a usage commitment rather than to owning servers and facilities.
CapEx and OpEx differ in payment timing, cost behavior, and what must be included in a fair TCO comparison:
| Dimension | CapEx on-premises | OpEx cloud |
|---|---|---|
| Payment timing | Upfront purchase before use, spread over Hardware Life | Ongoing charges based on resource usage |
| Cost behavior | Fixed capacity set by host count and admin ratios | Variable capacity adjusted by autoscaling and rightsizing |
| TCO inputs | Hardware, raw storage, network hardware, data center space, power, cooling, maintenance, IT salaries | Compute, Storage, Network, Operation, OS licenses, migration cost per VM, support level, and discounts |
TCO is the full cost of owning and running a workload over time, not just the purchase price. A cloud TCO comparison must include hardware and software plus data center operations on the on-premises side, and usage, migration, support, and network charges on the cloud side. Network cost is commonly overlooked because on-premises routers, switches, and cabling are one-time purchases for full capacity, while cloud network utilization can be billed by use.
Exam tip: CapEx is tied to owning physical infrastructure and is spread over its Hardware Life; OpEx is tied to consuming cloud resources and is billed as usage occurs.
Financial agility is the ability to spend in small flexible steps and move money quickly toward business value. Moving from on-premises to Google Cloud creates financial agility by replacing large upfront hardware purchases with OpEx that rises and falls with demand. This shift reduces the need to overprovision for peaks and lets leaders redirect capital from running data centers toward innovation and growth.
The CapEx-to-OpEx change replaces fixed owned capacity with elastic usage-based costs. On-premises TCO includes hardware and software plus all costs of running a data center, such as power, cooling, maintenance, and support services. The network boundary works differently on each side: in a data center, routers, switches, and cabling are one-time purchases that allow use of full network capacity, while in the cloud network utilization is billed, so data-intensive workloads may need new architectures and network flows to control cost.
Elastic scalability makes OpEx more flexible than a rigid data center. Compute Engine supports rightsizing during migration, sizing recommendations after VMs are already running, and autoscaling groups of instances. These scaling choices have a large impact on running cost and should be included when calculating TCO with tools such as the Google Cloud price calculator.
Elastic scaling automatically adjusts resources to workload fluctuations so applications have capacity only while demand exists. This avoids the on-premises habit of buying extra servers as a safety net for spikes and outages. Capacity flows from the available cloud pool to the workload when demand rises, then returns to the pool when demand falls.
Together these options improve performance, save cost, and use resources more efficiently. By reducing downtime and improving reliability, elastic scaling also helps prevent revenue loss and improve customer satisfaction.
Self-service optimization lets teams provision, manage, and optimize their own cloud resources without waiting for central purchasing. When teams receive self-service cost tools, observability tools, and resource management platforms, they can test ideas quickly and respond to demand changes on their own. This ownership accelerates innovation because experiments do not wait for manual hardware orders.
Reliable delivery practices keep that speed from creating waste and outages. DORA metrics for software delivery, such as change failure rate (CFR), time to detect (TTD), and time to restore (TTR), help find and fix bottlenecks in development and deployment. Site Reliability Engineering (SRE) practices focused on automation, observability, and incident response reduce downtime and recovery time, which removes the need to overprovision extra capacity just to survive failures.
FinOps (opens in a new tab) is the collaborative operating model that keeps flexible OpEx spending accountable to business value. FinOps creates shared accountability where engineering, finance, and business owners balance cost against value. This culture encourages experimentation because teams can see the cost effect of their choices and correct course quickly.
A value-driven and TCO-aware mindset looks beyond upfront price to lifetime cost and return. Value stream mapping visualizes the flow of value through software delivery to find improvements, while unit costing and unit economics connect cloud spending to business measures such as revenue per customer or transaction volume. When cost is compared with revenue in this way, the organization can verify that higher spend comes from profitable growth and distinguish healthy scaling from operational inefficiency that should be fixed.
Exam tip: On-premises network hardware is a one-time CapEx for full capacity, while cloud network use is an OpEx billed on utilization — include network flow in TCO for data-heavy workloads.
Total cost of ownership (TCO) is the full cost of running workloads over time, not just the purchase price. On-premises TCO includes direct costs such as hardware and software plus indirect costs such as electricity, cooling, facilities, and staff. Google Cloud TCO includes usage-based service costs plus migration, support, and discount effects, so a multi-year comparison shows the long-term financial impact of moving to the cloud.
On-premises TCO includes direct costs that appear on purchase invoices and indirect costs spread across facilities and staff. Direct costs include servers, storage, network equipment such as routers and switches, cabling, and software licenses. Indirect costs include power, cooling, rack space, maintenance, support services, and salaries for IT operations employees.
The Quick TCO Estimator models these costs with specific assumptions:
Performance and capacity assumptions also affect on-premises TCO, including CPU burst, memory burst, vCPU to core ratio, and storage redundancy policy such as FTT1/R1. A rigid data center can force overprovisioning as a safety net, which adds cost without adding value.
Google Cloud TCO is built from integrated service costs billed on usage rather than owned hardware. The Pricing breakdown divides costs into Compute, Storage, Network, Operation, and OS licenses. Additional inputs include migration cost per VM, support level, committed use discounts (CUDs), Enterprise Discount Program (EDP) or Compute Engine discounts, partner funding, and spend milestone credits.
Premium OS licenses are a separate TCO input and include Red Hat Enterprise Linux (RHEL), RHEL 7 Extended Lifecycle Support (ELS), Ubuntu Pro, SUSE Linux Enterprise Server (SLES), Windows Server, and Microsoft SQL Server tiers such as Enterprise, Standard, and Web. Network cost is commonly overlooked because data center network hardware is often a one-time purchase for full capacity, while Google Cloud bills for network utilization and egress. Data-intensive workloads may therefore need new architectures and network flows to lower networking costs, including attention to inter-region traffic.
Comparing TCO means comparing how each environment handles hardware, facilities, staff, network, scaling, and discounts over the same period:
| Area | On-premises | Google Cloud |
|---|---|---|
| Hardware and software | Upfront purchase amortized over hardware life | Usage-based Compute, Storage, and OS licenses |
| Facilities and staff | Monthly data center cost per host plus IT salaries and admin ratios | No owned data center; managed services reduce overhead but support tier is a cost |
| Network | Routers, switches, and cabling are one-time costs for full capacity | Billed for utilization and egress; inter-region traffic adds cost |
| Scaling and discounts | Rigid capacity with possible overprovisioning and limited discounts | Elastic scaling, rightsizing, CUDs, enterprise discounts, and Spot VMs |
A 5-year TCO comparison gives a clearer view than a one-year estimate because migration waves take time to plan, schedule, and execute. Google Cloud elastic scalability makes costs easier to reduce than a rigid data center through rightsizing during migration with Migrate for Compute Engine, recommendations after VMs are running, autoscaling groups, and Spot VMs that are more than 80% cheaper than regular VMs.
The Quick TCO Estimator is a Gemini-powered tool in Google Cloud Migration Center (opens in a new tab) that estimates TCO for migrating VMs and infrastructure to Compute Engine. It is available in us-central1, europe-west1, europe-west2, and europe-west4. For databases, the tool evaluates TCO from the VM infrastructure provided and can include SQL Server vCPU counts for licensing, but it does not accept database-specific inputs such as engine type or size.
You can create an estimate by entering infrastructure totals manually or by uploading RVTools exports. RVTools files must be csv files such as vInfo.csv and vDisk.csv or XLSX workbooks, with a maximum of 10GB per source. Individual XLSX tabs must not exceed 1,048,576 rows, and files must contain standard RVTools headers for automatic identification. After the estimate is created, results appear in three views: Summary shows average GCE TCO, workload scope, and 5-year TCO; Technical breakdown compares GCE configuration, on-premises estimations, and detailed metrics; and Pricing breakdown shows cost components, annual TCO, and 5-year TCO.
Managing TCO over time follows a fixed order where monitoring comes first, control comes second, and reduction comes last:
Discount and placement choices further affect long-term TCO. Enterprise discounts are negotiated on minimum total spend, resource-based CUDs commit to Compute Engine quantity for one or three years in specific regions, and spend-based CUDs commit to product spend at the billing account level. Region choices matter because prices vary by region, multi-region resources cost more than single-region resources, and inter-region traffic incurs egress costs. A FinOps practice ties these decisions together by finding waste with Active Assist, centralizing monitoring with Cloud Billing reports and the FinOps hub, and allocating spend with tags to drive accountability.
Exam tip: On-premises TCO includes indirect costs such as power, cooling, maintenance, and IT operations salaries, while Google Cloud TCO includes usage-based network egress and support costs.
Capital expenditures (CapEx) means buying owned physical infrastructure upfront before it is used, while operational expenditures (OpEx) means paying for cloud resources as they are consumed. Moving from on-premises to the cloud shifts spending from fixed CapEx spread over Hardware Life to variable OpEx billed on usage, which changes how total cost of ownership (TCO) is calculated over time.
On-premises total cost of ownership (TCO) includes direct costs such as servers, storage, routers, switches, cabling and software licenses plus indirect costs such as rack space, power, cooling, maintenance and information technology (IT) operations salaries. Google Cloud total cost of ownership (TCO) includes usage-based Compute, Storage, Network, Operation and operating system (OS) licenses plus migration cost per virtual machine (VM), support level, network egress and discounts.
Committed use discounts (CUDs) lower prices in exchange for committing to a level of usage over one or three years, while sustained use discounts apply automatically in Compute Engine for consistent use. Spot virtual machines (VMs) provide access to unused virtual machine capacity at a lower price for fault-tolerant work and are more than 80 percent cheaper than regular virtual machines (VMs).
A financial services company is evaluating the five-year Total Cost of Ownership (TCO) of migrating its on-premises data center and self-hosted databases to Google Cloud managed services such as Cloud SQL. Currently, the finance team's TCO model only compares two factors:
Based on this initial comparison, the finance team concludes that remaining on-premises is less expensive.
Which additional cost components must the company include in its TCO analysis to accurately determine the long-term financial impact?
The indirect and ongoing operational costs of the on-premises environment, including power, cooling, facility maintenance, and the engineering effort required for patching and disaster recovery
The upfront capital expenditure required to purchase and install physical routers, switches, and network cabling inside Google Cloud data centers
The standard per-gigabyte network fees charged by Google Cloud for all ingress data transfers entering the cloud environment from the public internet
The mandatory three-year Committed Use Discounts (CUDs) required before provisioning any virtual machines or managed databases on Google Cloud
Last updated