Explain It in 30 Seconds
A traditional enterprise may buy:
server
storage
network equipment
database licenseinstall them in its own datacenter and maintain them for years.
Azure offers another model:
customer needs computing
↓
Microsoft provides datacenter capacity
↓
customer uses compute / storage / database / AI
↓
customer pays based on contract / reservation / usageMicrosoft said Azure revenue surpassed $100 billion for the first time in FY2026. [S2]
That makes Azure one of Microsoft's largest businesses and one of the main reasons Microsoft is now physically capital intensive.
Where It Sits in Microsoft
Azure is the infrastructure layer underneath many Microsoft products.
Azure also sells the same infrastructure and services to customers building their own applications.
This creates both:
external product
The Problem It Solves
Companies historically had to buy infrastructure before they knew exactly how much they would need.
Cloud computing changes that.
A customer can:
- provision capacity faster;
- scale up and down;
- use managed services;
- deploy globally;
- access AI accelerators without owning them directly.
The economic trade-off is:
more dependence on cloud provider
How It Works — Simple Version
The more workloads the customer moves into Azure, the more Azure can monetize:
- compute;
- storage;
- databases;
- networking;
- security;
- AI.
Infrastructure as a Business Model
Cloud looks like software to the customer.
Underneath, it is a massive industrial system.
Microsoft must procure:
- land;
- datacenter buildings;
- servers;
- CPUs;
- GPUs;
- networking equipment;
- cooling;
- electricity.
The FY2026 10-K explicitly says cloud and AI strategy require significant and accelerating infrastructure investment. [S1]
This creates a different economic profile from Microsoft's earlier packaged-software business.
Azure and AI
AI made infrastructure even more strategically important.
Training and serving models require:
- GPUs / accelerators;
- high-speed networking;
- memory;
- large-scale power.
Azure distributes:
- OpenAI models;
- Microsoft MAI models;
- other models and AI services.
Microsoft's current AI strategy is increasingly multi-model. [S3][S24]
OpenAI and Azure
Microsoft's OpenAI relationship has several layers:
investment
+
IP license
+
cloud partnership
+
product integrationThe April 2026 amended agreement says Microsoft remains OpenAI's primary cloud partner while OpenAI gains more flexibility to use additional clouds. Microsoft retains a non-exclusive license to OpenAI IP through 2032. [S24]
The commercial implication is:
OpenAI can be important to Azure without Azure being only an OpenAI cloud.
What It Depends On
Datacenter construction
Cloud demand cannot be monetized without physical capacity.
Power
Electricity availability is increasingly a binding constraint.
GPUs / accelerators
AI demand can be limited by specialized hardware supply.
Networking
Large distributed systems depend on high-speed network infrastructure.
Software reliability
Customers move mission-critical workloads to Azure and expect high availability.
Developers and enterprise adoption
A cloud platform is valuable only if people build and run workloads on it.
What Depends on Azure
Microsoft Copilot
Many AI workloads need Azure infrastructure.
GitHub Copilot and developer services
Developer tools can connect to cloud and model services.
Dynamics / Microsoft 365 cloud services
The broader Microsoft Cloud stack depends on hyperscale infrastructure.
Third-party software companies
Independent vendors build SaaS products on Azure.
Why It Matters to Microsoft
Azure changed Microsoft's business from:
sell software that runs on customer hardwaretoward:
operate infrastructure
+
sell software / platforms / AI on topThat gave Microsoft:
- recurring consumption revenue;
- deeper enterprise relationships;
- a platform for modern application development;
- a physical infrastructure base for AI.
It also increased capital requirements dramatically.
Before / After Azure
| Dimension | Traditional Microsoft server model | Azure model |
|---|---|---|
| Hardware ownership | Customer | Microsoft |
| Revenue | License + support | Consumption + subscriptions |
| Capacity risk | Customer | Microsoft |
| Delivery | On-premises | Global cloud |
| Upgrade cycle | Customer-managed | Continuous service |
| AI access | Customer buys hardware | Azure exposes models / compute |
| Capital intensity | Lower for Microsoft | Much higher |
What Most People Misunderstand
"Azure is just Microsoft renting servers."
Too narrow.
Azure also sells managed databases, data platforms, AI, security and developer services.
"Cloud is asset-light software."
Not at hyperscale.
The provider owns or leases a huge physical infrastructure layer.
"Azure equals Microsoft Cloud."
No.
Microsoft Cloud also includes cloud components of Microsoft 365, LinkedIn and Dynamics. [S28]
"Microsoft depends on OpenAI for all Azure AI."
No.
Azure supports a broader model ecosystem and Microsoft develops its own MAI models. [S3]
Key Milestones
2008 Microsoft announces Windows Azure. [S6]
2010 Azure becomes generally available. [S6]
2014 Satya Nadella becomes CEO after leading Microsoft's Cloud and Enterprise group. [S7][S8]
2010s Azure expands from infrastructure into data, application and developer services.
2019–2026 AI infrastructure becomes a major Azure growth layer.
FY2026 Azure annual revenue surpasses $100 billion. [S2]
Evidence
- Microsoft said Azure annual revenue exceeded $100 billion in FY2026. [S2]
- Server products and cloud services generated $129.425 billion of FY2026 revenue. [S1]
- Microsoft identifies land, power, networking and servers / GPUs as cloud-capacity constraints. [S1]
- Microsoft remains OpenAI's primary cloud partner under the April 2026 agreement while its IP license is non-exclusive. [S24]
- Microsoft reports Azure demand continuing to exceed available capacity in some areas. [S3]
Connected Entities
- Microsoft
- Microsoft Copilot
- GitHub
- Satya Nadella
- How Microsoft Shifted from Windows to Cloud Computing
- How Microsoft Built AI into Its Product Stack