How Much Does Server-Side GTM Really Cost in 2026?

Going server-side, sure, but at what price? The server-side GTM cost is the first question everyone asks before migrating, and it is also the one that gets the most misleading answers. Most comparisons simply copy the pricing grids posted on the hosts’ websites. The problem: the advertised price is almost never the real cost. This guide prices the three main options for 2026 (self-hosting on Google Cloud Run, or using a managed host like Stape or Addingwell), breaks down the hidden costs, explains the request-counting trap, and gives you a clear decision framework based on your profile.

If you haven’t migrated yet, start with the why and how to move to server-side GTM guide, which covers the technical side. This article is its budget and tool-selection companion.

Why the cost question matters now

Server-side is no longer a niche setup in 2026, it is practically the standard. The scheduled death of third-party cookies, Consent Mode v2 now required for measurement in GA4 and Google Ads, and the loss of browser-side signal are pushing a wave of companies to migrate. The result: everyone is asking at the same time how much it costs and which tool to pick.

There are three very different models, and their billing logic has nothing in common:

• Self-hosted on Google Cloud Run: you pay Google directly for compute instances. You control everything, you maintain everything.
• Managed, Stape style: an international player handles hosting for you, billed in request tiers, with value-added features.
• Managed, Addingwell style: a French player (Didomi group), hosted in the European Union, on the same tier model, with a data-sovereignty angle.

The advertised prices, side by side

Here are the public grids as of June 26, 2026. Currencies are not harmonized (Cloud Run and Stape bill in dollars, Addingwell in euros): keep that in mind when comparing, and remember these prices change over time.

At first glance, Cloud Run looks unbeatable at $45. That is exactly where the trap springs shut.

The real cost of Cloud Run, broken down

The $45/month covers a single instance (1 vCPU, 0.5 GB) running in “CPU always allocated” mode, the one Google recommends for server-side to avoid cold starts. But Google also recommends a minimum of two instances for resilience. So your realistic floor is not $45 but closer to $90/month, before a single traffic spike.

Then comes autoscaling. Between 2 and 10 servers, you handle roughly 35 to 350 requests per second depending on how many tags run inside your container. The more heavy tags you have, the more CPU you burn, the higher the bill climbs. The price therefore becomes variable and hard to forecast.

And above all, there are the hidden costs nobody puts in the grid:

• Cloud Logging: beyond about 1 million requests, logs become billable. You have to remember to disable them manually through the log router, otherwise the bill quietly inflates.
• Load balancer: needed as soon as you deploy multi-region for performance, that is an extra line item.
• Custom domain and network configuration.
• Engineering time, the most underestimated item: :stable image updates, monitoring, scaling management, debugging. With no vendor support, and with real Google Cloud skills required.

In other words, Cloud Run should not be compared against a list price but against a total cost of ownership (TCO). The $90 of compute can quickly become $90 plus several engineering hours a month. For a team that already has GCP skills in house, it pays off. For everyone else, the “seemingly free” option is expensive.

The request-counting trap

This is the point most comparisons miss, yet it is the most important one for estimating your budget with a managed host.

Both Stape and Addingwell bill by volume of incoming requests, not by GA4 sessions or events. A request is every HTTP call that hits your server-side container. That changes everything:

• A single user hit can fire several tags (GA4, Google Ads, Meta CAPI, and so on) but still counts as one incoming request. Good news.
• On the other hand, the script loads served from your domain (gtm.js, gtag.js via the container) can count. Bad news.
• Preview mode and bots also generate requests that add up.

The takeaway: a request is neither an event nor a session. To estimate your real volume, start from your monthly pageviews, add script reloads and a margin for non-human traffic. A safe rule: count more generously than your GA4 intuition suggests, because GA4 filters out part of what your container actually absorbs.

One last structural trap: with both hosts, one container equals one site. If you run several domains, you need one plan (or a paid additional container) per site. For an agency or a multi-brand group, that is a cost multiplier to factor in from day one.

Three priced scenarios by profile

Numbers only really speak once tied to a concrete case. Here are three typical profiles (volumes in monthly requests, not sessions).

Small site, tight budget (~300K requests/month)

At this level, paying for Cloud Run and its engineering time makes no sense. Stape Pro covers the volume easily for the price of a lunch.

Growing e-commerce (~5M requests/month)

Here it comes down to two questions: do you have a technical team, and is EU hosting a constraint? If sovereignty matters, Addingwell. If you want minimum friction with power-ups, Stape. If you have a GCP engineer with spare cycles, Cloud Run.

Agency or high-volume multi-site (~20M requests, several domains)

At scale, “one container equals one site” becomes the real cost driver. Addingwell add-ons (extra container at +€30/month, domain at +€10/month) and Stape Enterprise plans earn their keep through support and pooled management.

The decision framework

There is no universal best tool, only a best choice for your profile. Here is how to call it:

• Low volume, tight budget: Stape Pro (or either host’s Free tier to test). No need to pay for Cloud Run.
• Technical team with Google Cloud skills: Cloud Run directly, if and only if you properly value maintenance time in your math.
• GDPR constraint or EU hosting requirement: Addingwell (Didomi group, hosted in Europe). For many French and European players, this is the deciding argument.
• Multi-site or agency: compare the total container cost (one per site) across hosts; add-ons and Enterprise plans add up fast.
• High volume: negotiate the upper tiers (Stape Enterprise, Addingwell 25M and above) or seriously model a Cloud Run setup with autoscaling.
• Need for specific power-ups (Cookie Keeper to extend cookie lifetime, Enricher, Custom Loader): Stape stands out on these features.

The cost of server-side is also a reason to look at the return. Properly configured, it powers your server-side enhanced conversions, able to recover 10 to 20% of conversions lost on the browser. It is often that ROI, more than the monthly price, that should guide the decision.

Checklist before you sign

Before choosing, answer these questions:

1. What is my real volume in requests (not sessions), bots and script loads included?
2. Do I have Google Cloud skills in house, and how many hours a month am I willing to spend on it?
3. Is EU hosting a legal or commercial constraint for me?
4. How many sites do I need to cover (one container per site)?
5. Which power-ups are genuinely useful to me, and which are just nice to have?
6. What is the total cost of ownership over twelve months, not just the list price?

Answer these six questions honestly and the right choice becomes obvious. The trap is never the advertised price: it is comparing a list price against a real cost.

Other managed hosts exist on the market; I focus here on the three most representative options to stay readable. If you are still unsure about the architecture itself, my Google Tag Gateway vs server-side GTM comparison will help you decide upfront whether you really need a full server-side container.