Cloud Platforms: VPS, AWS, Apify, Zyte Cloud, Hostinger VPS

You have a working scraper. Where does it run? Wrong answer: the same place as your blog. Right answer: depends on scale, team, and how much "managed scraping" you want to pay for.

This lesson honestly compares the four categories, with rough cost ranges. Numbers shift quarterly; treat them as orientation.

The four categories

Category	Examples	You manage	They manage
Cheap VPS	Hostinger, Hetzner, Contabo, Vultr	Everything	Network, hypervisor
Big-three cloud	AWS, GCP, Azure	OS up	Hardware, network, region
Managed scraping platforms	Apify, Zyte Cloud, ScrapingBee, ScraperAPI	Just write the spider	Hosts, proxies, anti-bot
Serverless	AWS Lambda, GCP Cloud Run, Vercel functions	Code only	Everything below

Lesson 68 dives into serverless specifically. Here we focus on the first three.

Cheap VPS, the underrated option

A 2-vCPU / 4GB VPS at Hetzner or Hostinger runs ~5–15 USD/month. Capabilities:

• Run a Postgres + Redis + scraper container stack via docker-compose.
• Comfortably handle 10–50 req/s with a residential proxy pool.
• Suit a single-person operation, side project, or early SaaS.

What you'll do yourself: OS patching, backups, monitoring, scaling (vertical only, bigger VPS).

Hostinger VPS has a nice middle: cheaper than Hetzner, more polished panel than Contabo, predictable pricing. Suparn runs personal projects (suparnpatra.com, scrapingcentral.com) on Hostinger.

When to outgrow: when one VPS isn't enough capacity, when you need geo-distributed exit IPs, when you need real HA.

Big-three cloud (AWS / GCP / Azure)

Expensive per equivalent compute (~3–5x VPS), but you get:

• Spot / preemptible instances at 60–80% discount (great for batch scrapes).
• Managed Postgres (RDS), Redis (ElastiCache), MongoDB (DocumentDB), Kubernetes (EKS/GKE/AKS).
• Regions worldwide for geo-targeted egress.
• A long list of services (Lambda, S3, SQS) you can compose.

A typical small-scraper AWS bill:

Service	Cost/month (USD)
t4g.small EC2 (2vCPU/2GB)	~10
RDS Postgres (db.t4g.micro)	~15
100GB S3 + traffic	~5
Data egress (50GB/mo)	~5
Total	~35

You can do the same on Hetzner for $10. The premium buys managed services, scalability, region diversity, and HR conversations ("we run on AWS").

Watch out for:

• NAT Gateway ($30+/mo just for existing). If your scraper is in a private subnet, this is a tax.
• Cross-region egress is shockingly expensive.
• Data egress to internet at $0.09/GB after the first GB. A heavy scraper can rack up unexpected bills.

Managed scraping platforms, Apify, Zyte, ScrapingBee

These platforms run your scraper for you, often handling proxies and anti-bot. Pricing models vary:

• Apify: "actor" runs billed by compute units; built-in proxies; marketplace of prebuilt scrapers. Free tier modest, then ~$49+/mo for production work.
• Zyte Cloud (formerly Scrapinghub): Scrapy-native hosting. Subscription + per-request charges. Has Smart Proxy Manager and AI-based parser.
• ScrapingBee / ScraperAPI / Bright Data: API-style, you send a URL, they return rendered HTML with proxies handled. ~$0.50–2 per 1000 requests depending on features.

When they make sense:

• Spider development is your moat; ops is not.
• You want anti-bot handled (CAPTCHA solving, fingerprinting) without building it.
• Burst workloads where keeping infra idle is wasteful.

When they don't:

• Steady high volume, DIY infra dramatically cheaper at scale.
• You need bespoke parsing/storage pipelines tightly coupled.
• You don't want vendor lock-in.

A common stack: managed proxies (Bright Data, Oxylabs) on your own infra. Get pro proxies without surrendering the entire platform.

Cost-per-1000-requests rough ranges (2026 ballpark)

Setup	Cost per 1k requests
Self-hosted on Hetzner with datacenter proxies	$0.05–0.20
Self-hosted with residential proxies	$0.50–3.00
AWS-hosted with own proxies	$0.30–1.00
ScrapingBee / ScraperAPI (rendered)	$0.50–2.00
Bright Data Web Unlocker	$1.00–5.00
Apify (prebuilt actors)	varies, often $0.30–2.00

The price varies by request difficulty (anti-bot intensity, JS render). Always price for your actual mix, not the marketing landing page.

Decision shortcuts

• Solo operator, side project: Hetzner / Hostinger VPS + Bright Data or Smartproxy if you need residential.
• Small SaaS scraping: Hetzner cluster or AWS EKS with managed proxies; build anti-bot in-house as you grow.
• Need to ship now, no ops team: Apify or Zyte. Pay the markup for time-to-market.
• Enterprise scale: Mix, your own infra for steady volume, managed APIs for hard targets.

Geographic targeting

If targets serve different content per country, you need exit IPs in those countries. Options:

• Multi-region VMs: Hetzner has DE/FI/US; AWS has 30+. Run scraper instances per region.
• Residential proxy with country selection: Bright Data, Oxylabs, Smartproxy all let you pick country.
• Cheap VPS in target country: Often the cheapest geo-targeting, one $4/mo VPS per country.

Lock-in considerations

Managed scraping platforms often have proprietary SDKs. Apify's actor model is portable to a degree, but Zyte's Smart Proxy Manager configurations don't move. Plain Scrapy + your own deployment is the most portable.

Big-three cloud services lock you in via egress costs and proprietary services (SQS, CloudWatch). Stick to Kubernetes + Postgres + S3-compatible storage and you're 80% portable across providers.

What to try

Price your own scraper across three options:

1. Hetzner VPS + Smartproxy residential.
2. AWS EC2 + Bright Data.
3. ScrapingBee API only.

Estimate $/month for your actual volume. The cost difference is often a factor of 5–20, and the "right" choice is highly workload-dependent, there's no universal best.