Cloud Repatriation in 2025: What the Numbers Actually Say

TL;DR for busy readers:

• 70% of enterprises plan to repatriate workloads by 2026 — you’re not crazy for considering it
• 42% of IT pros already moved from cloud back to dedicated servers last year
• Typical savings: 30-50% on infrastructure costs
• Startups have more freedom than enterprises — no vendor politics or procurement cycles
• Our test: €7.59/month VPS vs €10,560/month AWS Lambda for same workload (1,400x difference)
• Spending €5k-15k/month on cloud? We can help you evaluate options

What if the infrastructure you’re paying for could cost you half as much—with better control?

That’s not a pitch. It’s what the data is showing across companies of all sizes: a quiet but accelerating shift away from public cloud, back toward sustainable, hybrid architectures.

This isn’t about hating AWS or declaring “cloud is dead.” It’s about recognizing that for many organizations, the economics have flipped—and the trade-offs have changed.

Let’s look at what’s actually happening.

What Is Cloud Repatriation?

Cloud repatriation is the process of moving workloads, applications, or data from public cloud providers (AWS, GCP, Azure) back to on-premises infrastructure, co-location facilities, or dedicated servers.

It’s not a return to the pre-cloud era. Modern repatriation usually means:

• Dedicated servers (Hetzner, OVHcloud, Vultr) instead of EC2
• Simple container orchestration (Docker Compose, Kamal, Coolify) instead of Kubernetes
• Predictable monthly bills instead of usage-based pricing with egress surprises
• Full root access instead of managed-service abstraction layers

The goal isn’t to abandon all cloud services—it’s to place workloads where they make economic and operational sense. We call this sustainable infrastructure: a hybrid approach that uses cloud where it adds value, and dedicated resources where it saves money.

What You’re Trading

Enterprises often have big contracts with cloud providers, driven by top-down choices based on business relationships and procurement cycles. These contracts tend to prioritize politics over cost-efficiency. Startups, on the other hand, can experiment more freely and make decisions suited to their unique challenges, untouched by the constraints of scaling across thousands of employees.

To be clear, repatriation involves trade-offs:

For many teams, especially those with predictable workloads, the trade is worth it.

The Data: What’s Actually Happening

The repatriation trend isn’t anecdotal—it’s showing up in industry surveys, financial disclosures, and company announcements.

Enterprise Signal: VMware Private Cloud Outlook 2025

VMware (now part of Broadcom) surveyed 1,800 senior IT leaders across 17 countries (Sept–Oct 2024):

Top drivers for the shift:

• Security and data sovereignty — 68%
• Predictable costs (no egress surprises) — 61%
• Regulatory and compliance requirements — 59%

Source: VMware Private Cloud Outlook 2025

IT Professionals: Liquid Web Dedicated Server Study 2025

A survey of 1,009 IT professionals across industries:

The quote that stuck with me:

“The biggest misconception is that we don’t need [dedicated servers] anymore because of cloud adoption. But dedicated servers still handle our most critical operations due to reliability, control, and predictable costs.” — IT professional, Liquid Web survey

Source: Liquid Web dedicated servers study in a cloud-first world.

The a16z Analysis: The Cloud Paradox

Andreessen Horowitz published a widely-cited analysis in 2021 that has proven remarkably accurate. Their core insight:

“You’re crazy if you don’t start in the cloud; you’re crazy if you stay on it.”

Translation: Cloud makes sense for speed at the start—but the economics flip as you scale. The problem is, by then you’re locked in.

Their estimate: $100B+ in market cap lost across the top 50 public software companies due to cloud costs eating into gross margins.

The Dropbox example: Improved gross margin from 33% to 67% after infrastructure optimization (a mix of repatriation and internal tooling), saving $75M cumulatively in the two years before their IPO.

I saw this dynamic firsthand when working at Unity Technologies. In their public financial results, cloud infrastructure is listed as the “second-largest expense category.” When your hosting bill runs into the hundreds of millions annually, every percentage point of margin matters.

We strongly recommend reading the full analysis: a16z: The Cost of Cloud, a Trillion Dollar Paradox

The Famous Example: 37signals

DHH and the team at 37signals (Basecamp, HEY) are the most vocal about their cloud exit. The numbers from their public disclosures:

• $3.2M/year cloud spend in 2022 (already heavily optimized)
• $10M projected savings over 5 years after migration
• Zero new hires to manage the new infrastructure
• Built and open-sourced Kamal for zero-downtime deploys

Their take:

“The benefits have been vastly overstated. The cloud is often just as complicated as running things yourself, and it’s usually ridiculously more expensive.” — DHH

Source: 37signals Cloud Exit and Leaving the Cloud podcast

Enterprise vs Startup Freedom

Enterprises often have big contracts with cloud providers, driven by top-down decisions based on business relationships and procurement cycles. These contracts can prioritize politics over optimal economics. Startups, on the other hand, have the flexibility to explore and experiment with infrastructure choices. The challenges and decisions are entirely different from those faced by a 10,000-employee enterprise.

This freedom is an advantage—don’t follow what the big companies do when their situations don’t apply to yours.

The Gap: Why Growing Teams Get Stuck

Here’s the problem.

The enterprise data is compelling—but if you’re a growing startup with a 10-engineer team spending €70–90k/year on AWS, you’re not going to spin up a data center or hire a platform team.

And if you’re an indie hacker on a $5/month DigitalOcean droplet, you’re already lean.

The teams that get stuck are in the middle:

These teams know they’re overspending. They just don’t have the time, expertise, or confidence to change it.

A quote from an Indie Hackers thread captures it perfectly:

“I’m a Software Engineer and I don’t know DevOps stuff that well. I’d like to build safe, scalable projects and don’t worry that much about infrastructure… right now when I’m managing that, it takes too much time.”

That’s not a knowledge gap—it’s a bandwidth gap.

The Opportunity: Same Economics, Different Execution

The good news: repatriation at startup scale is faster, cheaper, and lower-risk than the enterprise version.

You don’t need:

• A multi-year migration plan
• $500k in hardware
• A dedicated platform team

You need:

• A focused audit of your current spend (48–72 hours)
• A shortlist of workloads that are predictable and steady
• A migration path that doesn’t block product work

Real Examples

pirsch.io (privacy-friendly analytics) handles 150 million page views per month on Hetzner infrastructure for €300/month. Their setup: a dedicated database server (€250/mo) plus 4 VMs and a load balancer (€50/mo). Simple. Predictable. Fast.

Source: pirsch.io Tech Stack and 2021 Recap

Our own test: We ran a production workload on a single Hetzner CAX21 (€7.59/month) and sustained 484 requests per second with P99 latency under 200ms. The equivalent AWS Lambda cost? Estimated at €10,560/month.

That’s not a rounding error. That’s a 1,400x difference.

Source: Our full analysis

What This Means

Cloud repatriation isn’t a contrarian stance—it’s a data-driven re-evaluation of where your infrastructure budget delivers the most value.

The pattern is consistent across company sizes:

1. Start in the cloud — it makes sense early. Speed matters more than cost.
2. Costs compound — as traffic and complexity grow, so does the bill.
3. Lock-in deepens — every managed service you add makes migration harder.
4. At some point, the math flips — and staying becomes more expensive than leaving.

For enterprises, that tipping point comes at $500k–$3M/year in cloud spend.

For startups, it can come much earlier—especially if you’re running predictable workloads on unpredictable pricing.

What We’re Doing

We’re documenting this shift—and helping small teams make the move when it makes sense.

If you’re spending €5k–15k/month on cloud infrastructure and wondering whether there’s a simpler path, we’re running focused 72-hour infrastructure audits to map out your options. No commitment, just data.

→ Request an infrastructure audit

Or follow along as we publish more. Next up: the step-by-step playbook for scaling from €8 to €800/month on simple infrastructure.

Published December 23, 2025