3.2 Costs and Advantages of Cloud, Hybrid, and On-Premises

Quick Answer: Buying servers, storage, and a datacenter is CapEx. Paying for Microsoft 365 as you go is OpEx. Public cloud wins on speed, burst capacity, and global reach. On-premises wins on leftover hardware, ultra-low latency, and some regulatory placements. Hybrid is the usual Microsoft 365 answer when both sets of constraints are true at once. None of those benefits remove your duty to protect data, identities, and devices.

Exam context

The April 30, 2025 outline asks you to compare costs and advantages of cloud, hybrid, and on-premises services. MS-900 will not ask you to calculate a bill or invent a dollar amount. It will give a finance, operations, or compliance story and ask which spending model or placement fits. Keep this section conceptual. Licensing SKUs and billing channels belong in Chapter 14. Service-credit mechanics belong in Chapter 15. Here you need the vocabulary and the trade-off.

CapEx versus OpEx

Microsoft Learn’s consumption-based unit uses two budgeting words:

• Capital expenditure (CapEx) is up-front spending on physical infrastructure — servers, network hardware, and datacenter space. The purchase is an asset you depreciate over years. You pay whether the hardware is busy or idle.
• Operational expenditure (OpEx) is ongoing spending on services over time. Because you pay for cloud services as you consume them, cloud computing is classified as an operating expense.

A Microsoft 365 subscription, add-on licenses, and metered Azure resources are OpEx. A rack of Exchange servers, a SAN, and a generator are CapEx. Hybrid usually produces both: you still depreciate remaining servers while you pay monthly for cloud mailboxes and Teams.

Learn lists the consumption-model benefits without prices: no up-front hardware cost, no need to buy capacity that sits unused, the ability to add resources when demand rises, and the ability to release them when demand falls. That is the financial story behind elasticity.

Capacity planning: guess versus track

Traditional datacenters force you to estimate future need. Overestimate and you overspend on idle kit. Underestimate and performance collapses until you can order, rack, and cool more hardware. Cloud capacity can follow actual demand: scale out for a product launch or a busy tax season, then scale back. You pay for what you used, not for a three-year guess.

That does not mean cloud is always cheaper. A fully depreciated, already-powered server farm that runs a stable overnight batch job may have a low incremental cost. The exam cares about which cost type and which risk you are accepting, not about a made-up savings percentage.

Advantages the exam actually names

Scalability is the ability to adjust resources to meet demand so you are not permanently over-provisioned. Vertical scaling changes the size of one resource (a bigger Cloud PC or mailbox plan). Horizontal scaling changes the count (more meeting capacity, more OneDrive users, more app instances).

Elasticity is rapid, often automatic scaling that follows demand. Seasonal retailers, university enrollment spikes, and a manufacturer’s year-end close are elasticity stories. Microsoft 365 itself is elastic from the customer’s point of view: you assign or remove licenses instead of unboxing servers.

Reliability is the ability of a system to recover from failures and keep functioning. Microsoft designs Microsoft 365 with redundant storage, geographic replication, and service-continuity processes. High availability is the uptime face of that design.

Service-level agreements (SLAs) are the contractual expression of availability. Microsoft publishes financially backed SLAs for Microsoft 365 online services (core workloads such as Exchange Online are commonly committed at 99.9% monthly uptime). An SLA is a commitment with defined downtime and service credits — it is not a promise of zero outages. Microsoft also publishes recent worldwide quarterly uptimes in the service description; those historical figures sit near 99.9% and above but are not the contractual SLA itself. Chapter 15 covers credits and Service health. For this chapter, remember: cloud vendors publish SLAs; a homegrown on-premises stack only has the availability you can staff and fund.

Geographic availability. Microsoft 365 runs in Microsoft’s global datacenter network. A new office in another country can get mail, Teams, and files without you shipping a SAN. Tenant geo and data-residency choices still matter for compliance (Chapter 13). The fundamentals point is simpler: public cloud already has regions; on-premises has the buildings you own.

Speed to deploy. Creating a user and a license is minutes. Standing up a new mailbox server, certificate, and backup job is weeks to months. Agility is why startups and merger “day-one” collaboration land on Microsoft 365 even when factories stay local.

Predictability versus burst. Learn splits predictability into performance (autoscaling and load handling so users get a consistent experience) and cost (you can forecast consumption and watch it in real time). On-premises cost looks predictable after the purchase — until a refresh cycle or a surprise peak. Cloud unit prices are visible, but a burst month can raise OpEx. Hybrid lets steady, local, or regulated work stay on owned capacity while bursty collaboration and remote access use the public cloud.

Why organizations stay hybrid

Hybrid is not a consolation prize. Official Microsoft 365 hybrid guidance and cloud-economics material keep returning to the same reasons:

• Regulation and data placement. A workload or a copy of data must stay in a specific country, on dedicated hardware, or under a regulator’s eye. Mail and chat can still move to Microsoft 365.
• Latency and locality. Shop-floor controllers, hospital imaging, or a print/scan cluster next to a press may need a local file share or server. Headquarters staff can still live in OneDrive and SharePoint.
• Sunk servers. Hardware is paid for and still under support. Hybrid lets you keep using it while you migrate mailboxes or files in waves.
• Staged migration. Exchange hybrid exists specifically so you can move mailboxes over time, keep a shared address book, and even move a mailbox back if needed.
• Application readiness. A line-of-business app is not cloud-capable yet. Identity can still be hybrid so that app and Microsoft 365 share the same user.

Considerations — advantages have a flip side

Connectivity dependency. Microsoft 365 is designed to be used over the internet. If the WAN or ISP fails, cloud mail, Teams, and OneDrive pause unless you have offline caches or a local fallback. Hybrid mailbox moves and directory sync also consume bandwidth; Microsoft’s Exchange hybrid planning guidance tells you to measure average mailbox size and throughput before you schedule waves. ExpressRoute can add a private path to Microsoft cloud services for specific scenarios, but Microsoft 365 is still intended to work securely over the internet.

Shared responsibility. Moving to SaaS does not move ownership of data, accounts, or access decisions. Microsoft’s shared-responsibility model is explicit: for every cloud type, you own your data and identities. Microsoft 365 (SaaS) means Microsoft operates physical datacenters, physical network, hosts, OS, and most network security. You still classify and protect data, manage users, apply MFA and Conditional Access, and secure endpoints. Client devices are a shared SaaS concern: Microsoft can offer device-management capabilities; you still own endpoint protection and compliance. A question that says “the cloud vendor handles everything, including our passwords and retention” is wrong.

Subscription sprawl. OpEx is easy to start and easy to forget. Unused Microsoft 365 licenses, overlapping add-ons, and shadow teams that buy extra capacity create waste. Cloud cost advantage depends on license hygiene (Chapter 14) just as on-premises cost depends on not buying idle racks.

Operational complexity of hybrid. You now run two environments: certificates and hybrid mail flow on-premises, plus cloud identity and service health. That is a real cost in skills and tooling. Hybrid flexibility is not free.

On the exam: “We already paid for the servers” and “the regulator will not let that database leave the building” point to hybrid or on-premises, not to “never use Microsoft 365.” “We need 200 contractors next month and zero the month after” points to public-cloud OpEx.