cloud-models-providers-and-primitives

Cloud Computing: Models, the Provider Landscape, and the Primitives

Part 1 of 5 of the Cloud Computing reference (plugin cloud-computing), covering §0–§3. Sibling skills: cloud-architecture-and-resilience (§4–§6), cloud-cost-security-and-operations (§7–§9), cloud-migration-sovereignty-and-ai-workloads (§10–§14), cloud-reference (§15–§20). Section numbers are shared across the set; a reference written as §N → skill points into that sibling skill.

Currency: Verified August 2026. See §17 → cloud-reference for the currency snapshot and what goes stale first.

How to read this. Reference, not a certification guide. Three markers:

• [DURABLE] — architectural principles and trade-offs that survive provider churn.
• [VERSIONED] — market data, pricing, service names, regulation. Verify before quoting.
• [CONTESTED] — genuine disagreement.

⚠️ GOTCHA boxes mark the mistakes that produce outages or surprise invoices.

Provider-neutral by default, with AWS/Azure/GCP names where the concept needs an anchor. Service names change; the primitives don't.

The three framings that organize everything below:

1. The cloud is someone else's computer, rented, with an API and a metered bill. Every genuine advantage (elasticity, global reach, managed services, capex→opex) and every genuine failure mode (cost surprise, lock-in, shared fate, opaque outages) follows from that one sentence.
2. ⚠️ Your failure domain is always larger than your architecture diagram says. The 2025 outages taught this expensively: multi-AZ deployments failed because of control-plane dependencies nobody had drawn (§6 → cloud-architecture-and-resilience). Map what you actually depend on, not what you think you depend on.
3. Cost is an architectural property, not an operational afterthought. It is determined by design decisions — data placement, service selection, egress paths — and discovered on an invoice 30 days later (§7 → cloud-cost-security-and-operations). Design for it up front or pay for it forever.

---

§0. Routing

Asked about...	Go to
Service and deployment models	§1
Which provider, and market reality	§2
Compute, storage, networking, identity	§3
Architecture and Well-Architected	§4 → cloud-architecture-and-resilience
Containers, Kubernetes, serverless	§5 → cloud-architecture-and-resilience
Resilience, and the 2025 outage lessons	§6 → cloud-architecture-and-resilience
Cost engineering and FinOps	§7 → cloud-cost-security-and-operations
Security and shared responsibility	§8 → cloud-cost-security-and-operations
Operations and observability	§9 → cloud-cost-security-and-operations
Migration strategy	§10 → cloud-migration-sovereignty-and-ai-workloads
Sovereignty, the EU Data Act, egress	§11 → cloud-migration-sovereignty-and-ai-workloads
Multi-cloud and lock-in	§12 → cloud-migration-sovereignty-and-ai-workloads
AI workloads and their economics	§13 → cloud-migration-sovereignty-and-ai-workloads
Sustainability	§14 → cloud-migration-sovereignty-and-ai-workloads
"Don't do this"	§15 → cloud-reference
"Which side is right?"	§16 → cloud-reference
"Is this still current?"	§17 → cloud-reference
Books and resources	§18 → cloud-reference

---

§1. Models

1.1 Service models

On-prem  →  IaaS  →  CaaS  →  PaaS  →  FaaS  →  SaaS
◄─────────── you manage more        you manage less ───────────►
◄─────────── more control           less operational burden ───►
◄─────────── more lock-in risk?     ⚠️ see §12 — it's not that simple

[DURABLE] The trade-off is real in both directions, and the common mistake is optimizing only one end. Managed services cost more per unit and less in engineering time; raw IaaS is cheaper per unit and you pay for it in people. ⚠️ The comparison people skip is the fully-loaded one — a managed database's premium is frequently less than one engineer's fraction of time keeping a self-hosted one patched, backed up, and monitored.

1.2 Deployment models

Public, private, hybrid (⚠️ the actual state of most large enterprises, whatever the strategy deck says), multi-cloud (§12 → cloud-migration-sovereignty-and-ai-workloads), and edge. Sovereign cloud is now a distinct category with regulatory meaning rather than a marketing label (§11 → cloud-migration-sovereignty-and-ai-workloads).

1.3 The five NIST characteristics

On-demand self-service, broad network access, resource pooling, rapid elasticity, measured service. [DURABLE] Dated 2011 and still the cleanest definition — and "measured service" is the one that quietly determines your architecture (§7 → cloud-cost-security-and-operations).

---

§2. The Provider Landscape

[VERSIONED — and read the ⚠️ in this section before quoting any number.]

Market structure as of Q1 2026: the Big Three hold roughly 60–68% of global cloud infrastructure spend, with the market reaching ~$129B in Q1 2026, growing ~35% YoY — described as the ninth consecutive quarter of increasing year-over-year growth.

⚠️ GOTCHA — the share figures genuinely conflict across sources, and you should know why before citing them. Synergy Research puts Q1 2026 at AWS 28%, Azure 21%, Google Cloud 14%. Other widely-circulated figures give AWS 30% / Azure 24–25% / GCP 13%, and others still AWS 32% / Azure 23% / GCP 11%.

The discrepancy is methodological, not an error. One analysis explains it directly: Azure's higher share estimates exceed a straight revenue calculation because Synergy weights IaaS-and-PaaS spend, where Azure is stronger — and separately, Microsoft does not break out Azure revenue at all, only its growth rate, so any "Azure revenue" figure is derived. Both kinds of figure are valid; they measure different scopes. Cite the source and the scope, or don't cite the number.

Growth is the more interesting story than share: Azure and Google Cloud have been growing substantially faster off smaller bases, with Google Cloud posting the highest rates of the three. AI is the driver — AI-related spending was reported at ~19% of total cloud spend in 2026, up from ~8% in 2023.

⚠️ The capex arms race is the structural fact underneath all of it. 2026 guidance figures in circulation: Amazon ~$200B, Microsoft ~$190B, Alphabet ~$180–190B. This functions as a moat — the cost of building global infrastructure excludes almost everyone — and it carries genuine risk: the Big Three are betting hundreds of billions that AI demand will justify the build-out, and if it disappoints the write-downs would be enormous. For a buyer, large capex commitments are both a supply signal and a lock-in signal.

Choosing — the honest criteria, roughly in order: existing skills and contracts (⚠️ usually decisive and rarely stated), the specific services you need, regional and data-residency footprint (§11 → cloud-migration-sovereignty-and-ai-workloads), pricing for your actual shape of workload, enterprise agreements and discounts, and regulatory posture. Beyond the Big Three: Oracle Cloud, IBM, Alibaba, and the "neoclouds" — GPU-focused providers (CoreWeave, Lambda and others) estimated collectively at a small but real share segment. Cloudflare, Fastly, Vercel, Fly.io for edge; DigitalOcean, Hetzner, Scaleway for straightforward compute at markedly lower prices.

---

§3. The Primitives

[DURABLE] Every provider has these under different names. Learn the primitive, map the name.

Compute: VMs (⚠️ still the majority of cloud spend), containers (§5 → cloud-architecture-and-resilience), serverless functions, GPU/accelerator instances (§13 → cloud-migration-sovereignty-and-ai-workloads), bare metal. Pricing modes matter more than instance choice: on-demand, reserved/committed (1–3 year, large discounts), savings plans, spot/preemptible (⚠️ 60–90% cheaper and interruptible — enormously underused for batch, CI, and fault-tolerant work), dedicated hosts.

Storage, and ⚠️ picking the wrong class is one of the most common cost errors:

Type	Use	⚠️ Watch
Object (S3, Blob, GCS)	The default for almost everything: static assets, data lakes, backups	Request costs at high volume; egress (§11 → cloud-migration-sovereignty-and-ai-workloads)
Block (EBS, Managed Disks)	Filesystem for a VM	⚠️ Provisioned, not consumed — you pay for the size you asked for, idle or not
File (EFS, Azure Files)	Shared POSIX access	Expensive per GB
Archive (Glacier, Archive tier)	Long-term retention	⚠️ Retrieval time and retrieval cost — and early-deletion fees

⚠️ Lifecycle policies are free money and routinely unconfigured. Object storage tiering from hot → cool → archive on an age rule takes minutes to set and runs forever.

Networking: VPC/VNet, subnets, security groups/NSGs, load balancers (L4 and L7), DNS, CDN, private endpoints, peering, transit gateway/hub-spoke, NAT gateways (⚠️ a surprisingly large and invisible line item at volume).

Identity — [DURABLE] the most important primitive and the most commonly misconfigured (§8 → cloud-cost-security-and-operations): workload identity (⚠️ use it — long-lived access keys are the single most common cloud breach vector), roles and policies, federation, and least privilege as an actual practice rather than an aspiration.

Managed data services: relational, NoSQL, cache, search, queues and streams, data warehouse. ⚠️ Each is a lock-in decision as well as an architecture decision (§12 → cloud-migration-sovereignty-and-ai-workloads).