Cloud Providers Overview: AWS, GCP, Azure in 2026
The global cloud infrastructure market crossed $400 billion in annual run rate in 2026, with Synergy Research placing Amazon Web Services at roughly 32% market share, Microsoft Azure at 23%, and Google Cloud Platform at 11%. For mobile app backends specifically, the gap narrows — Firebase alone powers over 3 million active apps, and Azure has become the default for enterprise-grade mobile projects emerging from Microsoft 365 shops.
Choosing a cloud provider is no longer a purely technical decision. It intersects with hiring (AWS talent is the deepest pool in Los Angeles), pricing negotiation (enterprise discounts range 15-40%), and compliance posture (each provider has different regional certifications). At Frenchy Digital, we've deployed mobile backends across all three major clouds and the pattern is clear: the best provider is the one your team can operate safely under pressure.
- AWS: 32% market share, deepest service catalog (240+ services), strongest LA talent pool
- Azure: 23% market share, strongest enterprise and Microsoft ecosystem integration
- GCP: 11% market share, best-in-class for data/ML workloads and Firebase mobile SDK
- Combined cloud spend in 2026 exceeds $400B annually with 21% YoY growth
- Firebase alone powers 3M+ active mobile applications globally
- Multi-cloud adoption grew to 87% of enterprises in 2026 (Flexera State of the Cloud)
| Provider | Market Share | Global Regions | Mobile SDK | Best For |
|---|---|---|---|---|
| AWS | 32% | 33 regions, 105 AZs | Amplify + Mobile SDK | Deep service catalog, mature ecosystem |
| Microsoft Azure | 23% | 60+ regions | App Center + Mobile Apps | Enterprise, Microsoft 365 shops |
| Google Cloud | 11% | 40 regions | Firebase SDK | Fast mobile MVPs, data/ML workloads |
| Alibaba Cloud | 4% | 28 regions | EMAS | APAC-focused apps |
| Oracle Cloud | 3% | 50+ regions | Limited | Oracle database shops |
The remainder of this guide dives into each provider's mobile-specific services, then pivots to cross-cutting architectural decisions — serverless vs containers, CDN strategy, and cost modeling — before closing with multi-cloud patterns and Los Angeles use cases we've shipped at Frenchy Digital.
AWS Services for Mobile Apps: Amplify, AppSync, Cognito, Lambda, S3
AWS offers the broadest mobile-specific toolkit of any cloud. For greenfield apps, AWS Amplify is the fastest entry point — it wires up authentication, data, storage, and functions through a CLI and generates client SDKs for iOS, Android, React Native, and Flutter. For custom architectures, the underlying primitives (Cognito, AppSync, Lambda, DynamoDB, S3, CloudFront) compose into virtually any shape.
AWS Amplify: Full-Stack Mobile Platform
Amplify bundles Auth (Cognito), Data (AppSync + DynamoDB), Storage (S3), Functions (Lambda), Hosting (S3 + CloudFront), and Analytics (Pinpoint) into a single CLI-driven workflow. The generated Flutter/React Native/Swift/Kotlin SDKs handle offline sync, conflict resolution, and token refresh automatically. Typical time from amplify init to a production backend: 4-8 hours for an experienced engineer.
AWS AppSync: Managed GraphQL at Scale
AppSync is a fully managed GraphQL service supporting queries, mutations, and real-time subscriptions over WebSockets. It integrates natively with DynamoDB, Lambda, Aurora, and OpenSearch as data sources. Subscriptions scale to millions of concurrent mobile clients without operational overhead — ideal for chat, live notifications, and collaborative features.
Amazon Cognito: Identity and User Pools
Cognito handles sign-up, sign-in, MFA, social identity federation (Google, Apple, Facebook), and SAML enterprise SSO. User Pools store identities; Identity Pools broker temporary AWS credentials for direct resource access (S3 uploads, DynamoDB queries). Free tier covers 50,000 MAUs — beyond that, pricing is $0.0055 per MAU.
AWS Lambda: Event-Driven Compute
Lambda runs backend code in response to HTTP calls (via API Gateway or Lambda Function URLs), S3 events, DynamoDB streams, Cognito triggers, and scheduled events. Cold starts have dropped below 100ms for most runtimes in 2026 with SnapStart. Pricing: $0.20 per million requests + $0.0000166667 per GB-second.
Amazon S3: Object Storage for Media
S3 stores user-generated content — profile photos, videos, documents, backups. Transfer Acceleration speeds uploads from mobile clients globally; Intelligent-Tiering automatically moves cold objects to cheaper storage classes. With signed URLs generated by Lambda, clients upload directly without proxying through your servers — a critical pattern for scaling mobile media.
| AWS Service | Mobile Use Case | Pricing Model | Free Tier |
|---|---|---|---|
| Amplify | Full-stack backend bootstrap | Pay per underlying service | Generous for dev |
| AppSync | GraphQL API + real-time subs | $4/million query ops | 250K queries/month |
| Cognito | Auth, MFA, social login | $0.0055/MAU over 50K | 50,000 MAUs |
| Lambda | Backend logic, API handlers | $0.20/M requests | 1M requests/month |
| DynamoDB | NoSQL database | On-demand or provisioned | 25GB storage |
| S3 | User media storage | $0.023/GB/month | 5GB |
| CloudFront | Global CDN | $0.085/GB egress (US) | 1TB/month |
| Pinpoint | Push notifications, analytics | $1/M push + SMS costs | 100 endpoints |
Over 70% of new mobile apps launched on AWS in 2025 used Amplify as their starting point — developers want infrastructure that disappears into the product.
— AWS re:Invent 2025 Keynote
Google Cloud Services for Mobile Apps: Firebase, Cloud Run, BigQuery
Google Cloud's mobile story is Firebase, acquired by Google in 2014 and now the most popular mobile backend-as-a-service in the world. Firebase is optimized for speed — most apps ship a working prototype in days, not weeks. Beyond Firebase, GCP offers Cloud Run for serverless containers, BigQuery for analytics at petabyte scale, and Vertex AI for on-device and server-side ML.
Firebase: Mobile BaaS Leader
Firebase bundles Authentication, Firestore (realtime NoSQL), Cloud Storage, Cloud Functions, Hosting, Cloud Messaging (FCM), Remote Config, A/B Testing, Crashlytics, and Performance Monitoring. The mobile SDKs handle offline sync and conflict resolution out of the box — a Firestore document changed on-device while offline will sync deterministically when connection returns. Firebase's pay-as-you-go Blaze plan scales smoothly from free tier to millions of users.
Cloud Run: Serverless Containers
Cloud Run runs any HTTP container, scales from zero to thousands of concurrent requests, and charges only for actual CPU/memory used. Unlike Lambda's 15-minute timeout, Cloud Run supports request timeouts up to 60 minutes — useful for video processing, batch jobs, and long-polling mobile endpoints. Deployment: push a container image, run gcloud run deploy, done in under 2 minutes.
BigQuery: Analytics Warehouse
BigQuery is Google's serverless data warehouse handling petabyte-scale queries in seconds. For mobile apps, it's commonly paired with Firebase Analytics — events flow automatically from the app SDK into BigQuery tables where you can run SQL for retention cohorts, funnels, and ML features. The free tier includes 10GB storage and 1TB of query processing per month.
Firestore vs Realtime Database
Firebase offers two NoSQL options. Realtime Database (original, JSON tree) is simpler and slightly cheaper for small apps. Firestore (newer, document-collection model) scales further, supports complex queries, and offers stronger consistency. For any new project in 2026, choose Firestore unless you have a specific reason — Google is investing exclusively in Firestore's roadmap.
| GCP Service | Mobile Use Case | Pricing Model | Free Tier |
|---|---|---|---|
| Firebase Auth | User authentication | $0.0055/MAU over 50K | 50K MAUs |
| Firestore | Realtime database | $0.18/100K reads | 50K reads/day |
| Cloud Storage | Media files | $0.020/GB/month | 5GB |
| Cloud Functions | Serverless backend | $0.40/M invocations | 2M invocations/month |
| Cloud Run | Serverless containers | $0.00002400/vCPU-sec | 180K vCPU-sec/month |
| FCM | Push notifications | Free (unlimited) | Unlimited |
| BigQuery | Analytics warehouse | $5/TB query | 1TB queries/month |
| Vertex AI | ML inference | Varies by model | Trial credits |
Firebase's appeal is operational simplicity: one console, one billing account, one SDK per platform. Its weakness is lock-in — migrating a Firebase-native app to another backend typically requires a full rewrite of data access code. Plan the exit before you enter.
Azure Services for Mobile Apps: App Service, Functions, Cosmos DB
Microsoft Azure's mobile strategy is tightly coupled to its enterprise identity platform (Entra ID, formerly Azure AD). For any team already invested in Microsoft 365, Dynamics, or on-prem Active Directory, Azure becomes the path of least resistance. Beyond identity, Azure offers a mature compute stack (App Service, Functions, AKS) and one of the most powerful globally-distributed databases on the market, Cosmos DB.
Azure App Service: Managed Web Hosting
App Service runs web apps, APIs, and mobile backends on fully managed infrastructure. It supports Node.js, .NET, Python, Java, PHP, and custom containers. Features include auto-scaling, deployment slots for zero-downtime releases, built-in SSL, and VNet integration. App Service Mobile Apps extension adds offline sync, push notifications, and social auth connectors.
Azure Functions: Serverless Compute
Azure Functions is Microsoft's answer to Lambda — event-driven compute that scales automatically. It supports HTTP triggers, Cosmos DB change feeds, Service Bus messages, timer triggers, and Event Grid events. The Consumption plan charges per execution; the Premium plan provides pre-warmed instances that eliminate cold starts for latency-sensitive mobile endpoints.
Azure Cosmos DB: Global Distribution
Cosmos DB is a multi-model, globally-distributed database supporting SQL, MongoDB, Cassandra, Gremlin, and Table APIs. Its killer feature is turnkey global distribution — write in one region, read from dozens with tunable consistency (strong, bounded staleness, session, eventual). For mobile apps with international users, single-digit millisecond reads from every region transforms UX.
Azure Entra ID (Auth)
Entra ID handles consumer and enterprise auth. Entra External ID (formerly Azure AD B2C) targets consumer mobile apps — customizable sign-up flows, social IdPs, MFA, and conditional access. For B2B apps selling to enterprises, standard Entra ID enables SSO with the buyer's existing Microsoft tenant.
| Azure Service | Mobile Use Case | Pricing Model | Free Tier |
|---|---|---|---|
| App Service | Managed backend hosting | Starting $13/month (B1) | 10 web apps (F1) |
| Functions | Serverless compute | $0.20/M executions | 1M executions/month |
| Cosmos DB | Global NoSQL database | $0.008/RU/hour | 1000 RU/s, 25GB |
| Blob Storage | Media files | $0.0184/GB/month | 5GB |
| Entra External ID | Consumer auth | $0.0055/MAU over 50K | 50K MAUs |
| Notification Hubs | Push notifications | $10/month basic | 1M pushes/month |
| Front Door | Global CDN + WAF | $0.081/GB egress (US) | 100GB trial |
| Application Insights | APM + analytics | $2.30/GB ingested | 5GB/month |
Azure's pricing is often negotiable at enterprise scale — EA (Enterprise Agreement) customers regularly secure 20-35% discounts. If your buyer is a Fortune 500 company likely to pay via their Microsoft EA, hosting on Azure can be a sales enabler beyond the technical merits.
Serverless vs Containers vs VMs: Choosing Your Compute Model
Every mobile backend ultimately runs on one of three compute models: serverless functions, containers, or virtual machines. In 2026, the pendulum has swung firmly toward serverless-first — CNCF's annual survey found 85% of new mobile apps use serverless for at least their API layer. But the choice is not binary; mature architectures mix all three.
| Model | AWS | GCP | Azure | Best For | Cost Profile |
|---|---|---|---|---|---|
| Serverless Functions | Lambda | Cloud Functions | Azure Functions | Spiky traffic, async jobs | Pay per invocation |
| Serverless Containers | App Runner, Fargate | Cloud Run | Container Apps | Custom runtimes, bursty | Pay per vCPU-sec |
| Managed Containers | ECS, EKS | GKE, GKE Autopilot | AKS | Long-running services | Pay per node hour |
| Virtual Machines | EC2 | Compute Engine | VMs | Legacy, specialized HW | Pay per hour |
| Edge Compute | Lambda@Edge | Cloud Functions at Edge | Front Door Rules | Low-latency transforms | Pay per request |
When to Choose Serverless Functions
- API endpoints with variable or unpredictable traffic
- Asynchronous background jobs (image resizing, email sending, webhooks)
- Event-driven processing (DynamoDB/Firestore/Cosmos change streams)
- Low-to-moderate sustained traffic (below ~100 RPS/function)
- Small engineering teams wanting zero infrastructure ops
- Rapid prototyping and MVPs where speed trumps cost optimization
When to Choose Containers (Cloud Run / Fargate / AKS)
- Sustained high traffic where reserved capacity is cheaper than per-invocation
- Long-running requests (>15 minutes — exceeds Lambda timeout)
- WebSocket-heavy real-time apps (chat, live collaboration)
- Custom runtimes or system dependencies not available in Lambda
- Microservices architectures with service mesh requirements
- Teams with existing Docker/Kubernetes expertise
When to Choose VMs
- Legacy applications not yet refactored to containers
- GPU workloads (ML training, video transcoding) not yet in serverless
- Specialized software requiring specific OS-level access
- Workloads with unusual networking or storage requirements
- Cost optimization at very large scale with reserved/spot instances
The fastest path to a profitable mobile backend in 2026 is: start on serverless functions, move hot paths to containers when your CFO asks about cost, and never touch a VM unless legacy forces you to.
— Frenchy Digital Engineering Team
- Cold starts on Lambda/Functions now typically 50-200ms — rarely user-visible for mobile
- Cloud Run supports scale-to-zero with ~500ms cold start (minimum instances eliminate this)
- Container startup time often the dominant factor — optimize image size first
- Reserved instances (EC2, GCE, Azure Reserved VM) cut costs 40-60% for sustained workloads
- Spot/Preemptible instances reduce costs 70-90% for fault-tolerant batch jobs
CDN Strategies: CloudFront, Cloud CDN, and Azure Front Door
A CDN is non-negotiable for any mobile app serving static assets (images, videos, JS bundles) or cacheable API responses. For apps with users outside the host region, a CDN cuts p99 latency by 50-80% and egress costs by 30-60%. All three major clouds offer competitive CDN products, though their architectural models differ meaningfully.
| CDN | PoPs | Egress Pricing (US) | Key Features | Best For |
|---|---|---|---|---|
| CloudFront (AWS) | 600+ PoPs, 100+ countries | $0.085/GB (first 10TB) | Lambda@Edge, Origin Shield, signed URLs | AWS-native apps |
| Cloud CDN (GCP) | 200+ PoPs, 150+ countries | $0.08/GB (first 10TB) | Anycast IP, HTTP/3, cache invalidation API | GCP-native apps |
| Azure Front Door | 192+ PoPs, 118+ countries | $0.081/GB (first 10TB) | WAF, routing rules, private link | Azure-native, enterprise |
| Cloudflare (3rd party) | 330+ PoPs, 120+ countries | $0.00 egress (Workers) | Workers edge compute, Zero Trust | Multi-cloud, startups |
| Fastly (3rd party) | 90+ PoPs, 60+ countries | $0.12/GB | VCL, instant purge, edge logic | Media-heavy apps |
CloudFront: AWS's Edge Network
CloudFront integrates deeply with S3, Lambda, API Gateway, and AWS Shield (DDoS protection). Lambda@Edge lets you run request/response transforms at 600+ edge locations — A/B testing, auth header injection, image resizing. Origin Shield adds a regional cache layer between edge and origin, reducing backend load for apps with many viewers but few origins.
Google Cloud CDN: Anycast Simplicity
Cloud CDN uses Google's global Anycast network — the same infrastructure serving Google Search and YouTube. Enable it with a single flag on a Global HTTP(S) Load Balancer. Strength: extremely low latency in regions where Google has PoPs. Weakness: fewer advanced features (no edge compute as rich as Lambda@Edge).
Azure Front Door: Global Load Balancer + CDN + WAF
Azure Front Door Standard/Premium combines global load balancing, CDN caching, and Web Application Firewall in one product. Strong fit for enterprises needing single-pane-of-glass security (WAF rules, bot mitigation, rate limiting). Premium tier adds private link to origins — traffic never touches the public internet.
- Always enable HTTP/3 (QUIC) — 10-30% latency reduction on mobile networks
- Use long cache TTLs (1 year+) with versioned URLs for static assets
- Deploy Origin Shield or equivalent to protect origins during traffic spikes
- Compress at edge (Brotli preferred, gzip fallback) — 20-40% bandwidth savings
- Signed URLs for private content (user uploads, premium videos)
- Log CDN traffic to S3/GCS/Blob for security and performance analysis
A common pattern: host static assets on CloudFront or Cloud CDN (near-zero cost for cache hits), and put Cloudflare in front of API endpoints for its free egress on Workers-originated traffic. This multi-CDN pattern can cut total egress bills by 40-60% at scale.
Cost Comparison and Estimation: Real Numbers for Real Apps
Cloud pricing pages obscure as much as they reveal. The only reliable way to estimate cost is to model your specific workload against real usage patterns. Below are three reference architectures with itemized monthly costs for a hypothetical social mobile app — you can substitute your own numbers.
Scenario: 10,000 MAU Social App (Small)
Assumptions: 10,000 MAU, 100K API calls/day, 500MB avg storage per user, 50MB avg CDN egress per user/month.
| Component | AWS Cost | GCP Cost | Azure Cost |
|---|---|---|---|
| Auth (10K MAU) | Free | Free | Free |
| Database (25GB, ops) | $45 (DynamoDB) | $55 (Firestore) | $60 (Cosmos DB) |
| Compute (serverless) | $30 (Lambda) | $25 (Cloud Run) | $35 (Functions) |
| Storage (5TB) | $115 (S3) | $100 (GCS) | $92 (Blob) |
| CDN egress (500GB) | $43 (CloudFront) | $40 (Cloud CDN) | $41 (Front Door) |
| Push notifications | $10 (Pinpoint) | Free (FCM) | $10 (Hubs) |
| TOTAL / month | ~$243 | ~$220 | ~$238 |
Scenario: 100,000 MAU Mobile App (Medium)
Assumptions: 100K MAU, 1M API calls/day, 1GB avg storage per user, 200MB avg CDN egress per user/month.
| Component | AWS Cost | GCP Cost | Azure Cost |
|---|---|---|---|
| Auth (100K MAU) | $275 (Cognito) | $275 (Firebase) | $275 (Entra) |
| Database | $350 (DynamoDB) | $400 (Firestore) | $420 (Cosmos) |
| Compute (serverless) | $280 (Lambda) | $240 (Cloud Run) | $310 (Functions) |
| Storage (100TB) | $2,300 (S3) | $2,000 (GCS) | $1,840 (Blob) |
| CDN egress (20TB) | $1,700 (CloudFront) | $1,600 (Cloud CDN) | $1,620 (Front Door) |
| Observability | $150 (CloudWatch) | $130 (Stackdriver) | $140 (App Insights) |
| TOTAL / month | ~$5,055 | ~$4,645 | ~$4,605 |
Scenario: 1,000,000 MAU Enterprise App (Large)
Assumptions: 1M MAU, 10M API calls/day, 2GB avg storage per user, with reserved capacity discounts applied.
| Component | AWS Cost | GCP Cost | Azure Cost |
|---|---|---|---|
| Auth (1M MAU) | $2,750 | $2,750 | $2,750 |
| Database (reserved) | $2,800 | $3,200 | $3,400 |
| Compute (mixed) | $3,500 | $3,200 | $3,800 |
| Storage (2PB) | $32,000 | $28,000 | $25,800 |
| CDN egress (400TB) | $28,000 | $26,000 | $26,500 |
| DR + Backups | $3,200 | $2,800 | $2,900 |
| Observability | $1,500 | $1,300 | $1,400 |
| TOTAL / month | ~$73,750 | ~$67,250 | ~$66,550 |
The difference between a well-architected cloud backend and a neglected one is typically 2-4x on the monthly bill. Right-sizing, reserved capacity, tiered storage, and aggressive caching pay for themselves in the first quarter.
— Frenchy Digital FinOps Practice
- Storage egress dominates bills at scale — invest in CDN and caching first
- Reserved instances (1 or 3-year commitments) cut compute costs 40-60%
- Savings Plans (AWS) and Committed Use Discounts (GCP) offer flexibility vs RIs
- Enterprise agreements with Microsoft can cut Azure costs 20-35%
- Cost anomaly detection (AWS, GCP, Azure all offer it) catches 90% of bill spikes early
- Tag every resource — untagged resources are 3x more likely to be abandoned and billed forever
Multi-Cloud Strategies and Los Angeles Use Cases
Multi-cloud is one of the most misunderstood decisions in modern architecture. Flexera's 2026 State of the Cloud found 87% of enterprises run workloads on multiple clouds — but only 23% genuinely integrate across them. The rest operate isolated silos (most commonly AWS for product + GCP for analytics). At Frenchy Digital, we rarely recommend true multi-cloud for mobile apps under $5M ARR.
When Multi-Cloud Makes Sense
- Regulatory isolation: Healthcare, defense, or FedRAMP workloads requiring geographic separation
- Vendor risk mitigation: Enterprise contracts requiring no-single-vendor clauses
- Best-of-breed services: BigQuery for analytics + AWS for product, or Vertex AI on top of Azure apps
- M&A integration: Acquired companies arrive on different clouds and full migration is deferred
- Customer data residency: EU GDPR data on one provider, US data on another
When Multi-Cloud Is a Mistake
- 'Avoiding lock-in' — all clouds lock you in; the question is which lock-in you accept
- Cost optimization alone — volume discounts on one cloud usually beat best-price cherry-picking
- Resume-driven development — team wants to learn every cloud simultaneously
- Startup teams under 20 engineers — operational complexity destroys velocity
- Without an abstraction strategy — running native services on both clouds doubles the surface area
Every multi-cloud architecture we inherit from a previous team takes 6-12 months to simplify. The cost of going multi-cloud is paid every day in operational friction. Only commit if the benefit is concrete and the alternative is worse.
— Frenchy Digital Cloud Practice Lead
Below are real (anonymized) architectures Frenchy Digital has delivered for Los Angeles clients — each illustrates a different cloud strategy.
LA Use Case 1: Beverly Hills Healthcare Telemedicine (AWS)
- HIPAA-compliant iOS/Android app on AWS with Cognito + AppSync + DynamoDB + S3
- PHI encrypted at rest (KMS) and in transit (TLS 1.3); BAA in place with AWS
- Lambda@Edge for geographic routing between US-West-2 (primary) and US-East-1 (DR)
- CloudFront with signed URLs for medical imaging — no PHI on public edge caches
- Monthly cost: ~$4,200 at 35,000 MAU; 99.99% uptime SLA to clinical partners
LA Use Case 2: Hollywood Entertainment App (GCP + Firebase)
- Firebase Auth + Firestore + Cloud Functions for user-facing mobile app (React Native)
- Cloud Run for media transcoding (FFmpeg containers, 20-minute jobs)
- BigQuery for Firebase Analytics data + ad performance cohort analysis
- Firebase Cloud Messaging for push (free, global, 10M+ monthly sends)
- Monthly cost: ~$1,800 at 120,000 MAU; shipped from concept to production in 9 weeks
LA Use Case 3: Santa Monica Fintech Platform (Multi-Cloud)
- AWS for primary app backend (Cognito, Lambda, Aurora, S3) — PCI DSS compliant
- GCP BigQuery for transaction analytics and fraud detection ML (Vertex AI)
- Data pipeline: AWS DMS streams to S3 → Dataflow → BigQuery hourly
- Cloudflare fronts both — unified security perimeter, one WAF config
- Monthly cost: ~$18,000 at 250,000 MAU; 15% overhead accepted for best-of-breed ML
- Default to single-cloud for apps under 100K MAU — simplicity compounds
- Default to AWS or GCP for new mobile apps; Azure if enterprise/Microsoft shop
- Use Terraform or Pulumi from day one — never manage cloud resources in a console
- Instrument FinOps (Cost Explorer, GCP Cost, Azure Cost Management) before launch
- Compliance (HIPAA, PCI DSS, SOC 2) adds 20-40% architectural complexity — plan for it
- Frenchy Digital delivers LA-based, AWS-certified architects for design, build, and operate
Architect Your Cloud Backend with Frenchy Digital
Free architecture reviews for LA businesses. We audit existing workloads for cost, performance, and security — or design new backends from scratch on AWS, GCP, or Azure.
Need a Production-Grade Cloud Backend?
Frenchy Digital's AWS-certified architects design, deploy, and operate mobile backends on AWS, GCP, and Azure. Free architecture reviews available for LA businesses.
1517 S Bentley Ave Unit 204, Los Angeles CA 90025
