Cold Email Infrastructure: How to Set It Up in 2026

Cold email infrastructure is the set of domains, mailboxes, DNS authentication, and warm-up you put in place so outreach reaches the inbox instead of the spam folder. It is the layer underneath your sending tool: separate sending domains, dedicated mailboxes on a trusted provider like Google Workspace or Microsoft 365, correct SPF, DKIM, and DMARC records, and a warm-up routine that builds each mailbox's reputation before real volume starts. Get this layer right and an average campaign lands. Get it wrong and even a great campaign never gets read.

Most deliverability problems that people blame on their software are really infrastructure problems. The tool sent the email fine; the domain was too new, the mailbox was unwarmed, or DMARC was misconfigured, so the receiving server filtered it. This guide covers what the infrastructure layer is, the four pillars it rests on, how much of each you actually need, and where the infrastructure ends and the sending service begins.

What is cold email infrastructure?

Cold email infrastructure is the sending foundation: the domains you send from, the mailboxes attached to those domains, the DNS records that prove you are allowed to send, and the reputation you build through warm-up before you scale. It is deliberately kept separate from your primary company domain so that a reputation problem on outreach never touches the email your business runs on. Think of it as the plumbing. The sending tool is the faucet; the infrastructure is everything behind the wall that decides whether water actually comes out.

A useful way to see it: infrastructure providers sell you the raw sending assets (fresh domains and pre-configured mailboxes), while a cold email service runs the campaign through those assets. Some vendors do only one of those jobs, some do both, and the split matters when you price a stack. You can send zero emails without infrastructure, no matter how good your copy is.

The four pillars of cold email infrastructure

1. Sending domains

Never run cold outreach from your primary domain. Register two or three secondary domains that look like your brand (a common pattern is your brand name with try, get, or hq in front, or a .co or .io variant), point each at your main site, and use them only for outreach. If one picks up a reputation hit, you swap it out and your real email keeps working. When you are buying separate sending domains for this, keep them close to your brand name so replies still look legitimate to a prospect. We cover the full DNS side of this in the cold email domain setup guide.

2. Mailboxes

Send from real mailboxes on an established provider, Google Workspace or Microsoft 365, not a bulk SMTP relay. Those two providers carry the sender reputation that receiving servers trust. Plan on two or three mailboxes per sending domain, and keep each one's volume low. The widely used safe ceiling is 20 to 50 cold sends per mailbox per day. That limit is the reason you need multiple mailboxes: to send 300 outreach emails a day safely, you spread them across roughly eight to ten inboxes rather than blasting all 300 from one.

The two providers behave differently once you scale, and it is worth knowing which constraints you are buying into. Google caps a Workspace mailbox at 2,000 sends a day and, more importantly, at 2,000 unique external recipients, which is set out in our guide to cold email on Gmail and Google Workspace. Microsoft is looser per mailbox but applies a tenant-wide external recipient ceiling that every mailbox you own draws on together, which is covered in detail in our guide to cold email on Outlook and Microsoft 365.

3. Authentication (SPF, DKIM, DMARC)

Authentication is how a receiving server confirms the email really came from you. Google's own sender rules require SPF or DKIM for every sender, and both SPF and DKIM plus a DMARC record once you send more than 5,000 messages a day to Gmail. A DMARC policy of p=none is acceptable to start; it monitors without blocking. Two details trip people up: SPF has a hard limit of 10 DNS lookups and exceeding it fails SPF entirely, and DKIM keys should be 2048-bit where the provider allows it. If you only remember one number, remember this: keep your spam complaint rate under 0.10%, and never let it reach 0.30%. Our SPF, DKIM, and DMARC explainer walks through each record.

4. Warm-up and sending pace

A brand-new mailbox has no reputation, so sending real volume from it on day one looks exactly like spam. Warm-up fixes that by ramping sends gradually over two to four weeks, exchanging and opening low-volume mail so the provider sees normal, engaged activity before cold outreach begins. After warm-up, hold each mailbox to its daily ceiling and rotate sends across your inboxes so no single address spikes. The mechanics are in our guide on how to warm up an email account, and an email warmup tool keeps the ramp running in the background instead of leaving it as a task you have to remember.

How much cold email infrastructure do you need?

Work backwards from your daily send target. Divide the number of cold emails you want to send per day by a conservative 30 sends per mailbox, and that is roughly how many mailboxes you need. Then group two to three mailboxes per domain. A worked example makes it concrete.

Start smaller than you think you need. It is far better to run three well-warmed mailboxes at their safe limit than to spin up thirty and burn them all with impatience. Add capacity as your reply pipeline justifies it, not before.

Do I need dedicated IPs for cold email?

For most senders, no. When you send through Google Workspace or Microsoft 365, you are on their shared sending IPs, and those IPs carry strong reputation that a small sender benefits from. Dedicated IPs only make sense at high, consistent volume where you can keep the IP warm on your own, and where you want full control over its reputation. Chasing a dedicated IP too early usually hurts, because a cold IP with low volume looks worse to filters than a shared one. Focus on domain and mailbox reputation first; the IP question can wait until you are sending tens of thousands a month. If you want to watch that reputation rather than guess at it, an email deliverability tool tracks spam rate, authentication and inbox placement per mailbox, and our cold email deliverability guide covers what to do when a number moves the wrong way.

Infrastructure versus the sending service

The infrastructure layer gives you domains, mailboxes, and authentication. The sending layer is what runs the actual campaign through them: writing the message, warming and rotating the mailboxes, pacing the sends, following up, and handling replies. Buying great infrastructure and pairing it with a weak sender wastes the foundation, and vice versa. This is why a cold email service that includes warm-up, inbox rotation, and reply handling natively removes a whole category of integration work; the deliverability controls live in the same tool that sends, so nothing falls through the gap between an infrastructure vendor and a separate sequencer.

The practical takeaway: decide first whether you want to assemble infrastructure yourself (buy domains, provision mailboxes, set DNS) and bolt a sender on top, or run a platform that connects to mailboxes you own and handles warm-up, rotation, and sending in one place. Smaller teams almost always come out ahead consolidating, because every extra vendor is another place deliverability can quietly break. If you are leaning that way, cold email marketing software is the category that bundles the copy, warm-up, rotation and follow-up together, and how many mailboxes per domain covers how to size the mailbox and domain layer underneath it.

One route is worth ruling out before you price it, because it looks like the cheapest option on the list. Pushing campaign volume through a bulk relay or transactional API is prohibited by the providers themselves, not merely discouraged: SendGrid, Mailgun, Postmark, Mailjet, SMTP2GO and Resend all ban sending to recipients who did not opt in, in their own published terms. Our page on SMTP for cold email quotes each policy verbatim and covers why a shared relay also breaks the reply path that a cold sequence depends on.

Common cold email infrastructure mistakes

• Sending from the primary domain. One complaint can affect the domain your invoices and contracts go out on. Always use secondary domains.
• Too much volume per mailbox. Pushing 100+ cold sends from a single inbox is the fastest way into the spam folder. Respect the 20 to 50 ceiling and add inboxes instead.
• Skipping or rushing warm-up. A mailbox that has not been warmed has no reputation to trade on. Give it the two to four weeks.
• Broken authentication. An SPF record over the 10-lookup limit fails silently, and a missing DMARC record caps how much Gmail will accept. Validate the DNS before the first send.
• Not cleaning the list. The best infrastructure in the world cannot save a list full of dead addresses; verify before you send so your bounce rate stays under 2%.

The bottom line

Cold email infrastructure is not a nice-to-have you add after the copy; it is the thing that decides whether the copy is ever seen. Separate domains, dedicated mailboxes on a trusted provider, correct SPF, DKIM, and DMARC, and disciplined warm-up and pacing are the four pillars, and they are all cheaper than the cost of a burned domain. Set them up once, keep volume within the safe limits, and let the sending layer do the rest.