TECHPERFORMANCE
Cost comparison · July 2026

What the cloud really costs.
Numbers in hand, real configurations.

We took four configurations we run in production and priced them on the public lists of AWS, Google Cloud, DigitalOcean, Kinsta, WP Engine, Matomo Cloud and ClickHouse Cloud. No invented numbers: sources cited, methodology declared, and the part others keep off the front page: who operates the machines.

Methodology

How these numbers are built.

Real configurations, not hypothetical ones

Every scenario is a configuration we run in production right now: RAM, storage, traffic and database rows are measured, not imagined.

Public price lists, cited

Cloud equivalents come from the official calculators and price lists of AWS, Google Cloud, DigitalOcean, Kinsta, WP Engine, Matomo Cloud and ClickHouse Cloud, surveyed in July 2026.

Operations cost money, so we count them

A cloud price without someone to operate it is half a price. Where needed, we add the market cost of a sysadmin: €350–700/day freelance, €500–1,500/mo on retainer, ~€4,000/mo for an in-house hire.

The scenarios

Four real workloads, five price lists.

Every scenario is a configuration in production today, with its real numbers. The bars compare the estimated total monthly cost: where the list price is not enough to actually run the workload, we say so.

Scenario 1 · Editorial WordPress network

A network doing 20 million visits a month

Real production configuration: 16 isolated environments, 124 domains, 20M+ aggregate monthly visits, ~62 GB total RAM on dedicated hardware. This is the configuration class running networks like Caffeina Media + NEM.

On managed WordPress platforms the maths is quick: you pay per site and per visit. Forty titles start at ~$1,400/mo even on entry plans, and real traffic volumes push that to several thousand. On the hyperscalers the equivalent infrastructure costs €900–1,200/mo, but it comes bare: Core Web Vitals tuning, orchestrated updates and the overnight bot flood remain the job of someone who gets paid separately.

Tech Performance
~€900–1,500/mo all included, team included
DigitalOcean DIY
~€1,800–4,000/mo infra + market-rate operations
Google Cloud DIY
~€2,300–4,800/mo infra + market-rate operations
AWS DIY
~€2,400–5,000/mo infra + market-rate operations
Kinsta / WP Engine
~$4,000–10,000+/mo per-site + visit tiers

Our bar includes: dedicated hardware, traffic, storage, offsite backups, 24/7 monitoring, continuous tuning, orchestrated updates and the team that answers.

Scenario 2 · Self-hosted Matomo analytics

A network's analytics, in-house

Real production configuration: a Matomo instance with 40 GB of RAM, 450 GB of storage and over 280 million rows of data, serving a 20M+ visits/month network.

Matomo Cloud's public price list stops at 50,000 hits a month for €29. At tens of millions of hits you enter custom-quote territory, and market estimates run to thousands of euros a month. The paradox: the bare equivalent machine on AWS costs more than our complete module, an r6i.2xlarge with 64 GB at ~€420/mo before storage, backups, or anyone to administer it.

Tech Performance
~€300–500/mo managed module, DB tuning included
DigitalOcean DIY
~€380–450/mo + ops infrastructure only
AWS DIY
~€450–550/mo + ops infrastructure only
Matomo Cloud
custom quote, estimates ≥ €2,000/mo public list stops at 50k hits

Our bar includes: InnoDB sizing on the real dataset, purge and retention configured, backups, Matomo updates, and a DBA when a report slows down.

Scenario 3 · ClickHouse warehouse

Billions of rows, a flat invoice

Self-hosted managed ClickHouse analytical warehouse: first-party ingest, interactive queries over billions of rows, multi-year retention. The configuration scales with volume; the cost model does not.

ClickHouse Cloud is an excellent product with a usage-based model: the Scale tier with two replicas starts at ~$500/mo, Enterprise runs from $2,700 to almost $10,000/mo, and egress costs $0.1152/GB every time data leaves. Self-hosted with us the data never "leaves": it sits on your dedicated hardware, inside a fee that does not move when the team runs more queries.

Tech Performance
~€250–600/mo module in the fee, volume-based
AWS DIY
~€450–550/mo + ops infrastructure only
ClickHouse Cloud Scale
from ~$500/mo + egress 2+ replicas, metered usage
ClickHouse Cloud Enterprise
~$2,700–9,700/mo custom config, compliance

Our bar includes: a cluster sized on real volume, backups, retention without surprise tiers, and an engineer who knows your schema.

Scenario 4 · WooCommerce ecommerce

An ecommerce that does not lose orders

Real production configuration: WooCommerce on a dedicated VPS with Redis object cache, a PHP-FPM pool tuned for the traffic, WAF and anti-bot rules. This is the configuration class of the ecommerce stores we operate.

On paper the prices look alike: a Kinsta Business plan costs ~$115/mo, a DIY stack on AWS €150–250. The difference shows at 3 AM, when the bot flood arrives: on one of our configurations a real attack went from load 40 to 0.24 without losing a single order, because the WAF, the tuning and the person who intervenes were already in the fee. On the cloud, that night is a ticket, an overage, or a sysadmin called in at full emergency rate.

Tech Performance
~€100–300/mo team and incident response included
Kinsta Business
~$115/mo + overage 100k visits, then $2/1,000 extra
AWS DIY
~€150–250/mo + sysadmin infrastructure only
WP Engine
~$96–400/mo + overage visit tiers

Our bar includes: WAF and bot mitigation, Redis and PHP-FPM tuning on the real cart, backups, 24/7 monitoring, and someone who answers when checkout matters.

Bars are proportional to the midpoint of the stated range. USD prices left in dollars where the price list is in USD.

The hidden lines

What is not in the cloud price.

The homepage price is the price of the machines. These four line items arrive later, and they usually all arrive.

Egress

AWS and Google Cloud bill outbound traffic: ~$0.09/GB on the first 10 TB. On editorial traffic volumes that is a four-figure yearly line item even behind a CDN. In our fee, traffic is included, full stop.

Visit overages

Managed WordPress platforms meter visits: on WP Engine, excess traffic costs $2 per 1,000 visits. A viral article or a match-day becomes a surprise on the invoice. With us a spike is a technical event, not an accounting one.

Operations

Someone has to update, tune, watch and intervene. Market rates in Italy: €350–700/day for a freelance, €500–1,500/mo on retainer, ~€4,000/mo total cost for an in-house hire. In our fees, this column is already inside.

Support that is not support

AWS Business Support costs 10% of the monthly bill and covers their platform, not your application. The slow query, the broken deploy and the overnight flood remain yours. Our real incidents of recent months (bot floods, full disks, post-upgrade regressions) each took 2–8 hours of specialist work: hours that, with us, never generate an extra invoice.

The approach

Artisanal does not mean small. It means made to measure.

Enterprise is a business model: standardise everything, scale the sales, put a ticket system between you and the people who know the answer. Craftsmanship is a different process: measure first, size accordingly, keep the same engineers on your stack for years. We run infrastructure the second way.

The enterprise model
  • Self-service by design

    The console is the product. Architecture, sizing and mistakes are yours to figure out.

  • Tickets, in tiers

    Severity levels, escalation queues, a different person every time. Nobody on the other side has ever seen your stack.

  • Platform support, not application support

    They answer for the hypervisor and the API. Your slow query, your broken deploy, your bot flood: out of scope.

  • Prudential oversizing

    Nobody audits your workload, so you buy headroom "to be safe" and pay for it twelve months a year.

  • A bill that moves

    Egress, IOPS, snapshots, visit overages. Your best traffic day is your worst billing day.

  • Lock-in by service

    Every proprietary managed service is one more reason you can never leave.

The Tech Performance model
  • Audit before quote

    We measure your real workload first. The fee is sized on what you actually run, not on a price list fantasy.

  • A call, not a ticket

    The engineer who answers is the one who built your stack and tuned it last month. Context included.

  • Whole-stack ownership

    From the hardware to the PHP pool to the CDN rule: one team, accountable end to end.

  • Sized on measurements

    We tune RAM, database buffers and PHP workers on observed traffic. You pay for what the workload needs.

  • A flat, boring fee

    Server, traffic, storage, backups, interventions: included. The invoice looks the same every month. On purpose.

  • No lock-in, by contract

    Standard open-source stack, documented playbooks that remain yours. You can leave. That is why clients stay.

Enterprise is a business model. Quality is a process. We chose the process.

Free audit

Let’s talk about your stack,
free, no strings attached.

30 minutes with a Romiltec architect. Together we figure out whether Tech Performance is a fit, and if it isn’t, we tell you straight away. No cold pitch, no black-box quote.

Book a call cal.com/romiltec/tech-performance · 30 min call

Indicative estimates, not quotes. Sources: public price lists and calculators of AWS (EC2 on-demand, EBS gp3, Data Transfer Out, Business Support), Google Cloud (Compute Engine), DigitalOcean (Droplets), Kinsta, WP Engine, Matomo Cloud and ClickHouse Cloud, surveyed in July 2026. Dollar prices are left in USD where the source list is in USD; values may vary by region, exchange rate and promotions. Cloud equivalents are sized on the actual resources of our production configurations; operations costs are market rates for sysadmin/DevOps roles in Italy. Tech Performance bands match our public pricing: the actual fee is calculated after the audit.