Buy first, integrate never

HostingBrain analyst brief · August 2026 · data snapshot 2026-08-16 · aggregate-level, no named providers

Roll-ups are usually described as a cycle. Buy a national hoster, move its customers onto the group platform, take the cost out, then buy the next one. The migration is where the synergy lives, so a buyer with a queue of half-digested acquisitions is supposed to slow down until the queue clears.

Public infrastructure signals show where each acquired brand's customers are actually served from, so the queue is measurable from outside. It is real. It is also not acting like a brake.

57 of 87acquired brands served from the buyer's own platform
1.30Mlive domains still served where they were bought, two or more years on
7 of 7owners carrying that residue kept acquiring

Integration is real — for two brands in three

Nine European hosting groups both operate a shared platform and have acquisitions we can read at the infrastructure layer. Between them they hold 87 acquired brands, and 57 of those — 65.5% — have their customers served from a platform belonging to the group rather than the one they arrived on. Whatever else is true, the machine works.

That matters before anything else, because the interesting number is a residue, not a failure rate. The question is not whether these groups integrate. It is what happens to the part that does not.

The residue does not clear

The other 30 brands fall into two groups, and only one of them is a story. 7 were bought inside the last two years and carry 149,568 live domains between them: a brand bought last quarter is mid-integration, not un-integrated, so they are counted as neither.

That leaves 23 brands whose customers are still served from the infrastructure they were bought with, two or more years after the deal. They are not a rounding error: together they carry 1,297,780 live domains. 21 of them have moved nothing at all; the other 2 moved their mail onto the group's platform and left the hosting where it was, and the younger of those two deals is 69 months old.

Those 21 have been sitting there for a median of 49 months — just over four years. The oldest was acquired 254 months ago, a little over two decades.

brands not served from the buyer's platform, by years since the deal (bar width = live domains) 2-3 years10 brands · 888,701 domains 4-5 years7 brands · 189,745 domains 6+ years6 brands · 219,334 domains 23 brands in total · snapshot 2026-08-16 21 of the 23 have moved nothing at all; the other two moved only their mail

The obvious objection is that these are migrations in progress, caught mid-flight. They are not. Alongside the headline reading of each brand's infrastructure there is a second one: the share of its individual domains already served from the group's platform, which is what a migration in progress looks like from outside. For 20 of the 21 that have moved nothing, that share is under 5%. The work has not started.

A backlog that is four years old is not a queue. It is a decision that nobody has had to justify.

And it does not gate the next deal

Seven of the nine groups carry at least one of these brands. All seven announced further acquisitions afterwards — 78 validated deals between them, counted only from the date of the oldest brand each owner still holds entirely on its own infrastructure. Not one of the seven paused while its queue stood still. The remaining two groups carry no such brand at all: every acquisition we can read for them is served from a group platform.

It would be tidier if the owners were simply dormant, but the second instrument says otherwise. Over the last 90 days, between 0.47% and 22.18% of these same groups' books moved between platforms the owner already controls — the ordinary churn of reorganising a portfolio. The books are being worked. The acquired backbones are the part that stays where it is.

What this changes for a buyer of one of these groups

Integration cost is the most common bridge between an asking price and a model. The record here says four things about that bridge.

  1. Ask which acquisitions were integrated, not what the integration rate is. Two in three is a respectable number and it hides the shape: the residue is concentrated, old, and made of businesses that were bought with their own infrastructure and still run it.
  2. Count the halfway cases on the un-integrated side. A brand whose mail has moved and whose hosting has not is not an integrated brand; it is two platforms with one of them shared. Where a group reports an integration percentage, ask what a partial move counts as.
  3. Treat "planned migration" as a dated commitment. A brand that has moved under 5% of its customers four years after the deal has a plan, not a project. The distinction is worth a question in a management session and a line in the model.
  4. Price the parallel estate. Two platforms mean two renewal cycles, two support surfaces and two sets of infrastructure contracts. Where a group has carried that for years without visible urgency, the run-rate is the base case and the migration is the option.

None of which says the buyers are wrong. Keeping a strong local brand on its own infrastructure is a defensible strategy — customers rarely leave for it, and a migration is the moment they might. But it is a strategy that should be stated and priced, rather than modelled as a synergy that arrives on schedule.

What we cannot see. Whether an owner intends to migrate, what it costs them not to, and what the platform contract says. Infrastructure signals show where customers are served from and when the brand changed hands; they do not show a plan, a budget or a decision. Everything above is a statement about observed sequence — un-integrated brand, then further deals — and not about causation.

How the population is defined

Three conditions decide who is in this brief, and each of them excludes somebody. A group counts as European when more than half of its live customer book sits on a European country domain — measured, not inferred from where the company is registered, because the record holds groups registered in Europe whose customers are overwhelmingly not. Seven groups fall out on that test. A group needs a shared platform of its own: without one, a brand running its own infrastructure has nothing to be integrated onto, which is a different story, and it removes one more. And a brand is only counted where its infrastructure is readable — brands served from behind a large cloud cannot be attributed, so we exclude them rather than count an unreadable brand as un-integrated; two further groups have no readable acquisition left after that.

Method & caveats. "Served from the buyer's platform" is one specific reading: the infrastructure a brand's customers are most commonly served from is one the group itself operates. A brand whose mail or name service has moved to the group while its hosting has not is counted on the un-integrated side, and named separately above, because its customers are still served where they were bought. The second reading is the share of that brand's individual domains already on a group platform. Both are public-signal observations, so a migration that keeps the same customer-facing naming is invisible to us, and a brand can be operationally integrated in ways the network does not show. Acquisitions come from our curated ownership record, built from primary announcements, counting validated acquisitions and mergers only; it is a floor, and small or undisclosed deals are under-represented. "Two or more years" is measured to the announcement date, with year-only dates read as mid-year. The European test counts country domains only, so a European operator whose customers mostly hold .com addresses reads as out of scope, and second-level national namespaces such as .com.pl are not counted toward it. Domain counts are the live business book attributed by infrastructure signature. Aggregate figures only: this brief names no group, and the per-group detail is the paid product. Snapshot 2026-08-16.

Reproduce this — or ask it yourself

Every figure here is queryable through the HostingBrain connector. In Claude or any MCP-compatible assistant, this is the whole brief in one prompt:

Prompt · paste into an MCP client with HostingBrain connected

“Using HostingBrain, show me how deeply each European hosting group has integrated its acquisitions: which acquired brands still run on their own infrastructure, how long ago they were bought, and how much of each brand's book has actually moved. Then show me how much of those same groups' books moved internally in the last 90 days.”

Resolves to integration_depth and group_movement (both Pro). definitions(term='integration_depth') explains the stage vocabulary and the weights, on the free tier.

Reproduce this analysis: the integration_depth tool returns the per-brand stage, acquisition date and migration share behind every count above — Pro tier, with the free tier covering market structure, concentration and the definitions.

Underwriting a consolidator? HostingBrain answers questions like this — with the date, denominator and caveats attached — inside Claude and any MCP-compatible assistant.

Get free access

Share this pageShare on LinkedIn