← Read the full guide
All infographics

Backend Platforms 2026

Railway vs AWS vs Azure vs GCP vs Firebase vs Cloudflare — picked for an AI iOS app.


$5-30
Indie /mo on Railway
200+
AWS services
90%
Of cost = AI API
1
Pick to start
By cost at small scale (a few thousand MAU)
DEFAULT
Railway
$5-30/mo · Predictable. Push code, get URL.
PaaS
Vercel
$0-20/mo · Hobby tier generous. Web-first.
Edge
Firebase
$0-50/mo · BaaS — least backend code. Lock-in real.
BaaS
Cloudflare Workers
$0-10/mo · Edge-first. Global low latency.
Serverless
AWS
$30-150/mo at small scale. Service breadth. Cost-trap risk real.
Power
Azure
$30-150/mo · Microsoft ecosystem, Azure OpenAI Service.
Enterprise
GCP
$15-80/mo · Cloud Run = cleanest serverless containers.
Dev DX
Pro Tip
Backend hosting is rarely your biggest infra cost. AI API spend (Claude / GPT) typically dwarfs it 5-10×. Optimize there first.
Common Trap
Choosing AWS "in case you scale" before you have users. Burns months learning vs shipping. Start simple — migrate when you must.
Pick by use case
By app maturity stage
  1. Solo iOS dev, Claude proxy
    → Railway
  2. Real-time chat / social
    → Firebase
  3. Global low-latency API
    → Cloudflare Workers
  4. F500 customer mandated
    → AWS / Azure
  5. GPU / ML training
    → AWS / GCP / specialty
  1. Stage 1 (0-100 users)
    Railway / Firebase / Vercel
  2. Stage 2 (100-10K)
    Same — optimize cache + sub plan
  3. Stage 3 (10K-1M)
    Most stay on PaaS. Some migrate.
  4. Stage 4 (1M+)
    Hyperscaler with SRE practice