Five Uncomfortable Truths about the European Web

Cloudflare appears as #3 in both Top 10 DNS and Top 10 Hosting providers in Europe. In both cases, the number overstates its commercial relevance - for different reasons. Treating Cloudflare's measured footprint as a unified market position leads to misallocation of competitive attention.

Rather than treating Cloudflare as a single competitor, ShareShift segmented all domains with a Cloudflare signal into four commercially distinct populations, each requiring a different response.

• DNS management only: 11.6%
• CDN layer on top of another host 33.5%
• DNS + CDN, security layer shielding actual infra 47.4%
• Full stack (DNS + proxy + Cloudflare compute) 7.5%

Cloudflare Trojan?

The headline dominance of Cloudflare is a modern-day wooden horse left on the beach.

Today, their massive scale is startlingly hollow; the footprint is bloated with free-tier users and system defaults. These raw deployment signals carry almost zero immediate commercial value.

However, dismissing the structure entirely is a mistake. That 7.5 percent full-stack compute segment is a loud signal of what is to come. Free DNS and proxy services are merely a wedge product, designed to build trust and capture the network control plane. As Cloudflare aggressively ships an expanding ecosystem of new products, the switch risk for traditional hosts quietly multiplies. For a hosting provider, explicitly bundling Cloudflare CDN for your customers might feel like a cheap performance win today. In reality, the niche vendor may eventually walk away with the entire customer relationship.

Six domains move in for every one relationship kept.

Across the last 13 months, hosting migrations onto Hetzner infrastructure have outpaced nameserver migrations by roughly 6 to 1, and the gap holds on a net basis too: Hetzner's net hosting footprint grew about 7x faster than its net nameserver footprint over the same period. The compute layer is genuinely surging. The domain relationship (and the renewal revenue) mostly isn't following. Hetzner functions less like a sticky endpoint accumulating a growing footprint, and more like a high-throughput waystation: gross monthly churn runs into the hundreds of thousands in both directions, while net accumulation is a fraction of that.

It's not a front door. Brand-new domains choosing Hetzner as their first-ever host are outnumbered 5.4x by domains migrating existing hosting there. Hetzner captures the move phase of a site's life far more than the launch phase. That's consistent with what Part 1 already shows: Hetzner is not, and isn't trying to be, an end-to-end supplier.

When customers disappear, they choose to.

Only about 1 in 5 domains that stop resolving to Hetzner do so through deactivation. The other 4 in 5 are active migrations to a named alternative. That's a notably low structural-churn share for a budget-adjacent VPS provider, and it points to a customer base that is technical, price-aware, and willing to move.

That sensitivity shows up directly in the flow data. Hetzner raised prices in April and again in June this year. First, a broad increase applied to existing customers with no grandfathering, and a steeper, narrower increase effective June 15 that applied only to new orders. The flow data tracks both moves with unusual precision. March looks like a cleanup operation: big jump in NS-level deletions before bills arrive for old unused accounts. June shows a double effect of deletions and outbound transfers. Two price actions, structured two different ways, produced two opposite and dateable flow signatures.

New capacity can mean different things.

Everything above is reported at the DNS layer; however, switching nameservers is quick and easy compared to a hosting migration. To check whether Hetzner's infrastructure was really absorbing what the migration data promised, we moved one layer down: tracking new IP capacity directly.

Specifically, we tracked how many new IP address blocks Hetzner spun up each month, their utilization and density changes over time. The deployment curve does not always match the domain volume linearly except in Q2 with a clear case of a DNS-then-Hosting migration sequence: a wave that began with 17,150 nameserver migrations the week of May 4, and peaked at 26,392 (a record!) in the same week as the new IP address capacity peaked, June 8.

To further contextualize Hetzner's capacity expansion, we compared it to two competitors with drastically different business models. IONOS, a large and structurally dense shared-hosting operator, shows what ordinary capacity expansion looks like: a handful of new blocks provisioned to fill with shared hosting accounts over time. Hostinger gave us a more direct comparison: caught up in the same VPS-demand narrative as Hetzner, Hostinger's growth looked similar in volume and density pattern; at least, at the surface. However, a closer look at the April spike revealed freshly provisioned shared hosting infrastructure, not VPS. Yet another reminder that quality of growth matters.

The data tells a different story: sovereignty may be impacting boardroom conversations, not yet server placement.

If we consider the volume of outbound migrations from AWS and Azure (US hyperscalers), the share of European alternatives is stable at c. 6% with no upward trend in 2026. That's right, the combined share of AWS migrations to OVHCloud, Hetzner, IONOS, Scaleway, etc. is tiny.

March 2026 might be one small exception. We saw a genuine spike in AWS migration volume, and OVHCloud captured most of it. Timing coincides with a major outage though. Could this be about reliability?

By contrast, the Q2'26 spike in outbound migrations from Azure is not a sovereignty story. What's actually happening is one vendor rolling out a Cloudflare proxy across .de/.uk/.dk, inflating the Azure denominator without touching the EU-directed numerator.

Important note: conclusions here are drawn based on analysing Nameserver migrations, and absolute counts are low. Domain-level analysis for web hosting (A-record) is available to paying ShareShift customers.

Exactly how the web is broken

ShareShift’s Health RAG scale (red, amber or green) applies across DNS, mail, web and performance issues. Definitions of warnings and errors closely match industry standards like CENTR to ensure our intelligence is fully audit-ready.  

With this methodology, every second domain in Europe is carrying a critical error. The chart breaks them down by frequency:

• SSL is by far the largest contributor. Missing, broken and expired certificates on domains with a valid website account for the vast majority of errors.
• Old PHP is affecting 1.5 million websites. Versions 7.* and lower were deprecated years ago. Any site with this technology today is carrying vulnerabilities and performance issues.
• 24% of sites are simply inactive. That's 13.5 million sites waiting for a buyer, or a project, to bring them to life. Who can save them?

Brand Specifics

DNS errors, inactive domains and expired SSL certificates are endemic across the ecosystem. Once you exclude those from the analysis, patterns emerge.

Without naming and shaming specific brands, we see clear choices that different providers make - or choose to ignore - when it comes to the health of their customers' content and their own infrastructure.

For hosting operators:

The insights above don't land the same way on every operator. Where you sit in the stack determines which one is most actionable for you.

For investors:

M&A teams, credit organizations and analysts tracking this space underwrite the European infrastructure investments at a scale that forces a different set of strategic choices. What's a given asset worth today, and what value can it create tomorrow?

Illustrative example: a potential target with 0.5mio domains; the teaser claims €30mio revenue and expects bids at 3x on that.

How can an analyst understand the target better, and give informed guidance before any LOI is prepared? Let's run the same 500k domains through two different quality profiles for a bottom-up recompute.

If 50% the portfolio is inactive (= below benchmarked range), and 80% of the rest runs with no intermediaries (i.e. no CDN, security shield or other hoster threatening to absorb the relationship), mostly hosting low-quality sites expressed as a digital footprint quality index of 0.75, then the target looks like a "basic hoster" profile. Many parked domains and websites with a low level of sophistication. We can reasonably apply an illustrative ARPU assumption of €150 on such assets to get to a estimated ARR of €22.5mio.

Finally, to see beyond the current "point in time" snapshot, we consider the net effect after migrations and churn. It shows 0.9, so an annual decline of 10% per year, meaning the business has a terminal value.

With a 3x top-line multiple, the valuation lands at €60.75mio.

By contrast, another target of the same size with better metrics can fetch double the price:

500k * 75% active * 80% no intermediaries * at-benchmark digital footprint of 1 * gross:net churn co-efficient of 1 (aka stable base) * 150 ARPU * 3x multiple = €135mio

The takeaway:

Thirty million in ARR and 500,000 domains under management isn't a valuation. It's a starting point for the diligence that decides which of these two businesses you actually bought.

Sovereign infrastructure spend keeps climbing regardless of whether real switching does. Gartner has European sovereign cloud spend roughly tripling by 2027, and even AWS has stood up a legally separate European entity in response. We cannot ignore these signals, even if our daily flow data shows they have not yet changed short-term customer behavior.

What's becoming crystal clear is this. Companies like Hetzner and Cloudflare are rapidly gaining high-margin revenues, regardless of where the domain renewal fee is paid. So while the domain remains the front door to the web, owning the front door means very little if someone else is hosting the party inside.

The market will keep rewarding those who look past the vanity metrics and measure what actually matters: the unglamorous, everyday infrastructure running beneath the surface.