Everyone selling cold email infrastructure will tell you a dedicated IP is the upgrade. It is the easiest thing to charge for and the easiest thing to put a checkmark next to on a comparison table.
The claim is half true. A dedicated IP gives you control, and control is worth real money once you send enough for it to matter. Below that point it does the opposite: it hands you an asset with no history and asks you to build a reputation on a volume of mail too small to build one with.
So the question is not which is better. It is where your volume sits relative to the crossover, and whether the pool you would otherwise sit on is one anybody is actually managing.
Why low volume breaks a dedicated IP
Inbox providers score an IP on sustained behaviour. Volume, consistency, bounce rate, complaint rate, how recipients treat the mail. That scoring needs a stream of data to work with.
A brand new IP arrives with none of it. To Gmail it is not neutral, it is unknown, and unknown is closer to suspicious than to trusted. The way out is to send consistently until there is enough history to form a view.
If you send 800 emails a month from a dedicated IP, you never get there. The signal is too thin and too intermittent. Every send looks like a first send. You have taken on all the responsibility of owning an IP with none of the benefit, and you are competing against senders on well-established pooled infrastructure that has years of history behind it.
The practical threshold is roughly 1,500 to 2,000 sends per IP per month, sent consistently. Not total volume across your estate. Volume per IP, spread across the week. An IP sending 80 a day every weekday builds reputation. An IP sending 2,000 on the first Monday of the month and nothing after does not.
Where the line actually sits
| Monthly volume per IP | Use | Why |
|---|---|---|
| Under 1,500 | Managed shared pool | Not enough signal to establish reputation. Borrowed history beats no history. |
| 1,500–5,000 | Either | Judgement call. Shared if your list quality is inconsistent, dedicated if it is clean and you want the isolation. |
| 5,000–30,000 | Dedicated | Enough consistent volume to build reputation, and enough at stake to want problems isolated to you. |
| 30,000+ | Multiple dedicated IPs, segmented | Split by campaign type or client so one bad list cannot take everything down. |
There is a second input that matters as much as volume: how much you trust your own list. A dedicated IP isolates you from other senders. It also isolates you with your own mistakes. If your bounce rate is unpredictable because your sourcing is unpredictable, a pool absorbs some of that. A dedicated IP concentrates it.
Teams sometimes reach for a dedicated IP because deliverability is bad, when the actual problem is a list they bought. Moving that list onto its own IP makes the damage cleaner to observe and no smaller.
Not all shared pools are the same thing
This is the part the standard advice skips, and it is why “shared IPs are bad” is too blunt to be useful.
When you send from Google Workspace or Microsoft 365, you are on pooled IPs shared with an enormous population of senders who were admitted by having a valid credit card. Nobody screens what they send. Nobody enforces a ramp. Your placement moves with the behaviour of people you will never identify, and when it drops you cannot prove why.
A managed pool is a different product wearing the same word. On a well-run pool, senders are vetted before they are admitted, complaint and bounce rates are monitored per sender, ramps are enforced rather than suggested, and senders who degrade the pool are moved off it. The pool is also usually segmented, so newer or lower-quality senders are not sitting on the same IPs as established ones.
That is a meaningful difference in outcome. It is also invisible from the outside, which is why you have to ask.
Five questions to ask before you accept a shared pool
- Who else is on it, and how did they get on? Open enrolment and vetted admission produce different neighbours. If the answer is “anyone who signs up”, you are on the Google model with none of Google’s scale to absorb it.
- What happens when a sender on the pool goes bad? You want a specific answer: monitored thresholds, automatic throttling, removal from the pool. “We investigate” is not a process.
- Are ramps enforced or advised? Advised ramps get ignored, and one sender pushing a new domain to full volume on day two affects everyone on that IP.
- Is the pool segmented? New senders and established senders on the same IPs means your reputation is capped by the worst recent arrival.
- Who handles a blocklisting, and on what timeline? On a shared IP, delisting is the provider’s job. Find out whether that means hours or a support ticket.
If a provider cannot answer those five clearly, the pool is not managed, it is just shared.
What an IP does not fix
Worth saying plainly, because dedicated IPs get sold as a deliverability product and they are not one.
Your domain reputation is separate from your IP reputation, and for most cold senders it carries more weight. Gmail in particular leans heavily on domain-level signals for senders who are not at enormous volume. A fresh dedicated IP attached to a domain with a poor history changes nothing.
Authentication is not affected by IP model at all. SPF, DKIM, DMARC and MX either pass or they do not, and misconfigured records are still the single most common cause of cold email landing in spam. If you have not checked, check before you spend anything on IPs.
List quality is unaffected. Bounces and complaints hurt you identically on either model. They are simply more visible on a dedicated IP, which is a real benefit, but visibility is not a fix.
Roughly speaking, the IP model is worth arguing about after authentication is clean, the list is verified and warmup has run. Before that it is the wrong problem.
What each model actually costs
Pricing across the market is inconsistent enough that headline per-mailbox rates stop being comparable, mostly because IPs are priced separately by some providers and bundled by others.
| Model | Typical cost | What to watch |
|---|---|---|
| Managed shared pool | $0.80–$2.00 per mailbox / month | Cheapest per mailbox. You are buying the provider’s pool management, so the five questions above are the whole purchase. |
| Dedicated IP, bundled | $1.30–$2.50 per mailbox / month | Check whether “dedicated” means per workspace or genuinely per mailbox. |
| Dedicated IP, as an add-on | Around $50–$100 per extra IP / month | Ten IPs adds $500–$1,000 that never appeared in the per-mailbox figure you compared. Price IPs as a separate line. |
| Google / Microsoft mailboxes | $4–$7 per mailbox / month | Pooled only, per-seat licence, no IP option at any price. |
The comparison people get wrong is a shared pool against Google. Both are pooled, so the checkmark on a table looks identical, but a monitored pool and an open one are not the same risk. The comparison that matters is managed against unmanaged, not shared against dedicated.
Switching later
Neither choice is permanent, and neither is free to reverse.
Moving from a pool to a dedicated IP means warming from zero again. Your domain reputation carries over, which helps, but the IP itself starts unknown and needs the same ramp any new IP does. Budget three to four weeks before it is carrying full volume.
Moving from dedicated back to a pool is faster because you inherit established history immediately, but you also give up the ability to isolate a problem to yourself.
The sensible pattern for a growing team is to start on a managed pool, watch volume per sending domain, and move to dedicated IPs when you are consistently past a few thousand sends a month and want the isolation for debugging. Agencies usually go further and run one IP set per client, so one client’s aggressive list cannot damage another client’s placement.
The short version
Under about 1,500 sends a month per IP, a dedicated IP is the wrong purchase. There is not enough volume to build reputation and you are better off borrowing established history from a pool somebody is actively managing. Past roughly 5,000 it flips, and the isolation becomes the thing that turns deliverability from guesswork into something you can debug.
Between those numbers it depends on your list more than your volume. Clean and predictable, take the dedicated IP. Inconsistent, fix the list first.
And whichever way you go, the IP is not the lever most people think it is. Authentication, domain reputation and list quality decide more of your placement than the IP model ever will.
Next: what a dedicated IP costs and how to warm one, or how many mailboxes you actually need.
Frequently asked questions
Is a shared IP bad for cold email?
Not inherently. An unmanaged pool is bad, because your placement moves with senders nobody is screening. A monitored pool with vetted senders and enforced ramps is a reasonable place to send from, and below roughly 1,500 sends a month per IP it will usually outperform a dedicated IP you cannot generate enough volume to warm.
How much volume do I need to justify a dedicated IP?
Around 1,500 to 2,000 sends per IP per month as a floor, sent consistently across the week rather than in bursts. Past 5,000 the case is clear.
Does a dedicated IP guarantee inbox placement?
No. It isolates your reputation from other senders, which makes problems debuggable. It does not affect authentication, domain reputation or list quality, and those account for more of your placement than the IP does.
Does warmup work differently on a shared IP?
Yes. On a pool you are warming the domain and the mailbox, and the IP already has history. On a dedicated IP you are warming all three at once, which is why the ramp takes longer and why skipping steps is more expensive.
Can I switch from shared to dedicated later?
Yes, but the new IP starts with no history and needs a full ramp, so allow three to four weeks before it carries target volume. Your domain reputation carries across, which shortens it somewhat.