Shared vs dedicated IPs for cold email

For cold email at realistic volumes, a shared IP — specifically the pools behind Google Workspace or Microsoft 365 — beats a dedicated IP almost every time. A dedicated IP needs consistent volume, typically thousands of emails a day, before inbox providers can hold a stable opinion of it, and cold outbound at 25–40 emails a day per inbox never gets near that floor. Let your domain carry the reputation and let Google's infrastructure carry the sending.

What is the difference between a shared and a dedicated IP?

Every email leaves a server with an IP address, and receiving providers score that address on its sending history: volume, consistency, bounce rates, spam complaints. On a shared IP you send alongside thousands of other customers of the same provider, and the pool's collective behaviour shapes the score. On a dedicated IP you are the only sender, so the score reflects your behaviour alone.

The pitch for dedicated IPs writes itself — full control, nobody else's mistakes dragging you down. That pitch is accurate for high-volume senders and wrong for almost everyone running cold outbound. As I set out in the Cold Email Deliverability guide, the infrastructure layer decides whether anything else in your outbound works, and this is one of the places where the intuitive choice fails.

Why do dedicated IPs underperform at cold email volume?

Reputation systems run on data, and low-volume senders do not produce enough of it.

The mechanism runs like this: when you send from a dedicated IP at 25–40 emails a day, then receiving providers see too few data points to form a stable reputation profile. When the profile is thin, then filters default to caution, and caution means the spam folder. When your volume also fluctuates — thirty emails on Tuesday, none at the weekend, eighty after a new campaign loads — then the profile never settles, and each send is judged as if it came from a near-stranger.

A Google or Microsoft shared pool has the opposite properties: enormous volume, years of history, and aggressive internal policing. Riding that pool means your messages arrive on infrastructure the receiving provider already trusts. You inherit a reputation you could never build alone at cold-email volume.

Doesn't a shared pool expose you to other senders' spam?

Less than you would think. Google and Microsoft police their pools hard — accounts that spam get throttled or suspended quickly, which is precisely why the pool reputation stays strong. More importantly, modern filtering has shifted weight from the IP to the sending domain. Your domain's history follows you whichever address the mail leaves from, which is why domain reputation is the asset you are actually building; the IP question is secondary.

The residual risk sits with smaller SMTP providers whose pools are policed loosely. If you use a specialist sending tool, check whose infrastructure sits underneath it. A cheap pool full of other people's cold email is the worst of both worlds: shared exposure without shared trust.

When does a dedicated IP actually make sense?

There are legitimate cases — they just don't look like cold outbound.

The pattern: dedicated IPs suit senders with enough steady volume to feed the reputation algorithms daily. If your business is big enough for that, you likely have someone whose job is deliverability. A 5–50 person B2B firm running outbound is not that sender and should not pay to become a worse version of one.

What should you optimise instead?

Put the effort where the leverage is. Send from secondary domains through warmed Google or Microsoft mailboxes, hold the line at 25–40 emails a day per inbox, authenticate properly, and verify every list before it loads. Then watch the numbers weekly — the early-warning metrics that catch decay before it becomes a burned domain — rather than debugging after replies dry up.

Those numbers deserve the same visibility as revenue. Bounce rate and reply rate sit next to pipeline value on the same page in the MD Dashboard Blueprint, because a deliverability problem is a pipeline problem with a two-week delay. The firms that treat it that way notice in days; the ones that don't notice when the quarter goes quiet.

The IP debate is mostly a distraction sold to people who want a purchase to solve a discipline problem. Rent trusted infrastructure, protect your domain, and spend the saved attention on lists and follow-up.

---

Next step: the Growth System Audit — £450, seven days, credited against any build — maps where your growth system leaks and what to build first.