# Amazon SES vs SendGrid Source: https://wraps.dev/versus/amazon-ses-vs-sendgrid SendGrid sells a company one vendor relationship covering API, SMTP relay, marketing and a procurement contact. SES sells an endpoint. The question is not which delivers better, it is how much of that relationship you actually consume. SendGrid is the email vendor most teams tried first, and the one an unusual number are quietly costing out an exit from. That is not because it delivers badly. It is because it is sold as a company-wide relationship — relay for the systems nobody owns, a marketing console for people who do not write code, a procurement contact, a certification pack — and a team that only sends application email from one codebase is paying for the whole relationship to use a tenth of it. SES is the other extreme: an endpoint and a bill, with everything above it left to you. ## At a glance |   | Amazon SES | SendGrid | | --- | --- | --- | | What it is | Raw sending infrastructure from AWS | Twilio's email platform — API, SMTP, and marketing | | Published pricing | $0.10 per 1,000 emails on à la carte. New accounts are defaulted to the Essentials plan at $0.16 per 1,000; Pro is $105/mo + $0.22 per 1,000 and Enterprise $500/mo + $0.23 per 1,000. Plans are per-account, per-Region. | Essentials $19.95 at 50K and $34.95 at 100K. Pro $89.95 at 100K, $249 at 300K, $499 at 700K. Dedicated IP is Pro and above; extra IPs $30/mo. Marketing Campaigns is priced as its own plan, not an add-on. | | Best for | Teams already on AWS with an engineer willing to own the plumbing. | Enterprises wanting one procurement relationship, SMTP everywhere, and marketing in the same console. | | Watch out for | Production access is an application AWS can refuse, and new accounts start sandboxed to verified addresses. No dashboard, no templates, and nothing stores your events until you build it. | The free tier died on 27 May 2025, replaced by a 60-day trial at 100/day. Support latency and abrupt account suspensions dominate its public reviews. | Prices are each vendor's published list prices, verified August 2026. Check the vendor's own page before you sign anything. ## Where they actually differ ### What the first afternoon looks like Amazon SES You create an IAM policy, verify a domain with three DKIM records, pick a region and call a send API. There is a console, but it exists to manage identities and quotas rather than to look at mail, and nothing in it answers did message X reach the recipient. Every part of an email vendor a non-engineer would recognise is absent by design, and you will either build it or go without it. SendGrid You get an account, an API key, SMTP credentials, a template editor, a suppression manager, an activity feed and a catalogue of integrations for platforms you have not thought about yet. The breadth is the product. It also means the first afternoon involves choosing between four ways to do the same thing, and the default for most teams ends up being whichever one the documentation opened on. ### Neither one still has the free tier people remember Amazon SES The old arrangement where mail sent from EC2 was effectively free is gone, and so is the assumption behind a great many tutorials. New accounts now land on a metered plan rather than the flat per-message rate, plans are scoped per account and per region, and the à la carte rate is something you select rather than something you are given. It is still the cheapest credible option at volume by a wide margin — just not free, and not what a 2021 blog post describes. SendGrid The permanent free tier was retired in May 2025 and replaced by a time-boxed trial at a low daily cap. This matters more than it sounds: a large number of side projects, internal tools and staging environments were quietly running on that tier, and every one of them became a billing decision on a deadline. If your reason for being on SendGrid was that it cost nothing, that reason has expired. ### SMTP relay, and the systems nobody is allowed to rewrite Amazon SES SES speaks SMTP too, with a per-region endpoint and credentials derived from an IAM user through a conversion step. It works, and it is the path a lot of legacy traffic takes. The rough edges are real: credentials are an IAM artefact so rotating them is an IAM change rather than a button, the endpoint is region-specific so a failover plan has to name a second one, and appliances with opinionated TLS behaviour occasionally need a port argued over. SendGrid This is SendGrid's most under-discussed strength. A decade of being the default relay means there is a documented path for WordPress, for Jira, for a Salesforce org, for the CI system and for the warehouse printer, usually written by the other vendor rather than by SendGrid. When the question is how do we get mail out of a system we are not allowed to modify, breadth of prior art is worth more than unit price. ### The marketing half, which is a separate purchase Amazon SES SES has no equivalent and does not pretend to. There are no campaigns, no segments, no audience, no visual builder and no report a marketer would open. Teams that need that alongside SES run a second tool for it, which is a coherent architecture — but it is two vendors, two suppression concepts and two answers to the question of who unsubscribed. SendGrid SendGrid has a real marketing product with lists, segments, a drag-and-drop designer and campaign stats. The part that surprises people at renewal is that it is priced as its own plan rather than bundled into the sending tier, so the invoice that looked like one number becomes two. The upside is genuine: the marketing side and the transactional side share a suppression list and a sending reputation, which is exactly the reconciliation problem the two-vendor architecture creates. ### Getting cut off, and who you talk to when it happens Amazon SES AWS will pause sending on an account whose bounce or complaint rate crosses published thresholds, and it is not gentle about it. The difference is that the thresholds are documented, the metrics that lead to them are in your own console and your own CloudWatch, and you can watch the number climb for a week before anything happens. Nothing about it is a surprise unless you never built the alert. SendGrid Abrupt suspension is the single most common theme in SendGrid's public reviews, and the accounts describing it are usually not sending anything abusive — a traffic spike, a new domain, an unfamiliar pattern. The compliance team operates on its own schedule and the first-line support tier does not overrule it. If your business stops when mail stops, the operational question is not how good is delivery, it is how long is the phone call. ### Support is a line item on one side and a plan on the other Amazon SES Basic AWS support gets you documentation and forums. A human who will look at your SES account is a paid support plan sitting on top of the sending bill, and for a small team that plan can cost several multiples of the mail itself. This is the cost most SES comparisons omit, and leaving it out is how a cheap-on-paper migration ends up merely competitive. SendGrid Support is included in the plan, with the tier you are on deciding the queue you sit in. Response latency is the other recurring complaint in its reviews, so included is not the same as fast, but there is a ticket system, a named severity scale and no additional purchase required to open one. For a team without an AWS account manager already, that is a genuine difference rather than a marketing bullet. ### Procurement, paperwork and the contract you already signed Amazon SES If the company already runs on AWS, SES arrives under an agreement legal has read, on an invoice finance already reconciles, inside a compliance boundary the security team has already assessed. Adding it is a configuration change rather than a vendor onboarding. That is not a small thing in a regulated shop, where the cost of a new supplier is measured in weeks of questionnaire rather than dollars. SendGrid SendGrid is a new vendor, with a new security review, a new data processing agreement and a new annual renewal. Twilio maintains the certification pack that makes this survivable and enterprise buyers do get through it routinely. But it is a process with a calendar attached, and a team that discovers it two weeks before launch is going to miss the launch. ### What the invoice is actually shaped like Amazon SES Per message, with no platform fee and no plan to outgrow, billed by AWS alongside everything else. The complications are the ones AWS complications always are: the plan is per account and per region, so a second region is a second decision, and dedicated addresses are a separate line. Cost scales linearly and predictably, which is precisely why large senders end up here. SendGrid A sending plan, plus a marketing plan if you use that side, plus dedicated IP charges above a certain tier. Each of those is predictable on its own and the total is the thing people underestimate. The crossover against SES is not close at high volume, and it is genuinely arguable at low volume once an AWS support plan and a week of engineering are in the comparison. ## Pros and cons, side by side ### Amazon SES Strengths - Metered per message with no platform fee and no tier to outgrow, arriving on an invoice finance already reconciles if the company is on AWS — which turns adding email into a configuration change rather than a supplier onboarding. - The sending identity, the quota and the accumulated standing sit in an account you keep, so replacing the tooling above them later is a code change rather than a warm-up you have to serve a second time. - The thresholds AWS enforces against are published and the metrics that approach them are in your own CloudWatch, so the enforcement that stops sending is one a single alarm gives you a week of notice on. - Choosing a jurisdiction is a per-account setting rather than a sales conversation, which makes a residency clause something an engineer satisfies in an afternoon. Trade-offs - Nothing above the send call exists, and a half-built replacement is worse than none: the event consumer that captures deliveries but silently drops complaints looks like it is working right up until AWS pauses you. - Getting a human to look at your account means buying an AWS support plan on top of the sending bill, and for a small team that plan is several multiples of the mail it protects. - There is no campaign product and there is not going to be one, so the marketing team's requirements become a second vendor, a second consent record and a second answer to who opted out. - A new account starts in a sandbox that only delivers to addresses you verified, and leaving it is an application a person at AWS reads on their timetable rather than yours. ### SendGrid Strengths - Fifteen years as the industry default relay means the instructions for pointing your CRM, your ticketing tool or a warehouse appliance at it already exist, usually written by that system's own vendor rather than by SendGrid. - Suppression management, an activity feed, a template editor and campaign tooling are all there on day one, so the first afternoon ends with an email operation instead of a backlog of things to build. - A support ticket is included in the plan rather than purchased as a separate contract, with a defined severity scale and a queue that exists before you need it. - Campaign traffic and application traffic share one suppression list and one standing with the mailbox providers, which is exactly the reconciliation nobody performs when those two live at different vendors. Trade-offs - The permanent free allowance ended in May 2025 and became a time-boxed trial, so every staging environment and internal tool quietly parked there turned into a billing decision with a deadline attached. - Marketing Campaigns is metered as its own plan rather than folded into the sending tier, so a cost model drawn from the sending price alone will be short by an entire line. - Being cut off without notice is the theme that dominates its public reviews, and the compliance function that does it is not reachable through the first-line queue you would be escalating in. - It is a new supplier: a security review, a data processing agreement and an annual renewal, which in a regulated shop is weeks of questionnaire that nobody put on the launch plan. ## Pick in thirty seconds ### Choose Amazon SES if - The company already runs on AWS, so SES is a configuration change inside an existing contract rather than a supplier onboarding with a security questionnaire attached. - Volume is high enough that a per-thousand rate beats a plan ladder even after you price an AWS support plan and the engineering time to build a console. - You need delivery events in your own warehouse on your own retention schedule, for an audit trail or a dispute process rather than for debugging. - Nobody at the company needs a visual campaign builder, because nothing about SES will ever provide one. ### Choose SendGrid if - Mail has to leave systems you do not control — a CRM, a ticketing tool, an appliance — and a documented relay path for each of them is worth more than the unit price. - A marketing team and an engineering team both need to send from the same domain and you would rather they shared one suppression list than reconciled two. - You need a support ticket to be included rather than purchased, and an enterprise certification pack that a procurement team can read in one sitting. - Volume is modest and stable, so the plan price is noise against the engineering cost of building what SES omits. ## What actually moves when you switch In roughly the order you would do it. The expensive parts of an email migration are never the send call. 1. Apply for SES production access before writing a line of integration code. The form asks how you collect addresses and how you process bounces, a person reads the answer, and a vague one buys a round of follow-up questions rather than a refusal. This is the step that moves launch dates. 2. Verify the sending domains in SES and publish its DKIM records alongside SendGrid's rather than instead of them. Both sets coexist happily, and removing the incumbent's records early is how a rollback quietly stops being available halfway through the cutover. 3. Export all four of SendGrid's suppression collections — global unsubscribes, bounces, spam reports and blocks — and load every one into the SES account-level suppression list. They live behind separate endpoints, and a team that exports only the unsubscribe group will mail people who previously filed a complaint. 4. Rewrite anything built on SendGrid dynamic templates. The handlebars-style substitution is evaluated inside SendGrid and the SES template API is not a comparable destination, so most teams end up rendering finished HTML in the application and handing SES a string. Plan for that rather than for a translation. 5. Replace the event consumer rather than reconfiguring it. SendGrid posts a batched array to a single URL; SES publishes individual notifications through SNS, EventBridge or Firehose under different field names, so the webhook handler is new code and the delivery-status mapping is a decision somebody has to make. 6. Choose where the campaign mail is going before the transactional cutover, not after. SES will not take it, and a marketing send due the week after the switch is the most reliable way to get a migration reversed by somebody who was not in the planning meeting. 7. Convert the SMTP senders one system at a time. SES credentials are derived from an IAM user through a conversion step instead of being issued directly, and the endpoint belongs to one region, so a failover story has to name a second. Budget a day for each appliance with opinions about TLS. 8. Treat the dedicated IP as unrecoverable. The address and everything the mailbox providers learned about it stay with SendGrid, so raise volume on the new identity across several weeks, leading with the recipients who open most, and keep the old path live until placement data says you can stop paying for it. ## Questions to ask before you sign Put these to a sales rep, or find the answer in the documentation before anyone quotes you a number. 1. Which AWS support plan would we genuinely need, and what does it add to the sending bill — because on the default tier a mail problem gets you documentation and a forum thread? 2. Does the SendGrid quote in front of us cover Marketing Campaigns, or is that a second plan on a second meter that nobody added to the spreadsheet? 3. How many of our senders are systems we are not permitted to modify, and does a documented relay path exist for each of them on the side we are proposing to move to? 4. On the morning sending stops, who do we contact, what response time are we contractually owed, and does that commitment exist on the plan we are actually buying rather than the one above it? 5. Where do delivery events land, how long are they kept, and could we answer a billing dispute about a message sent eleven months ago without asking the vendor for a favour? 6. Who inside the company needs a visual campaign builder, and has anybody asked them — because one side of this comparison has no answer and is not building one? 7. How many calendar weeks does a new-supplier security review take here, and is that shorter or longer than waiting on an AWS production-access decision? ## A third shape of answer The specific trap on this pair is that the two ends are so far apart. A team leaving SendGrid over price is usually not leaving over the product, and discovers at month two that it has traded an invoice for a backlog: an event consumer, a suppression view, a template pipeline, somewhere to look up a message. That is the gap a platform layer over your own SES account fills, and Wraps is one way to buy it while the account, the reputation and the sending data stay in AWS you control. The honest costs: you still need an AWS account and SES production access, which is an AWS approval on AWS's timetable and not ours to grant, our SDKs cover TypeScript and Python only, contacts and templates live in our database rather than yours, and we are not SOC 2 certified — so a procurement team that chose SendGrid for the certification pack has not been answered by this. Wraps publishes this page. We are not one of the two vendors being compared here, and this note is on it because the tension above is one owning the AWS account actually addresses — not because it is on every page. ## Common questions ### Is SendGrid built on Amazon SES? No, and this is the main place the SES comparison differs from the one people make with Resend. SendGrid operates its own mail infrastructure and has done since well before it was acquired by Twilio, so moving between the two genuinely changes the network your mail leaves from, the IP pools it uses and the relationships behind them. Arguments about deliverability between these two are therefore about real differences, not about two accounts on the same substrate. ### How much cheaper is SES really, once everything is counted? At high volume, several multiples, and the gap widens as you grow because one side is metered and the other is a ladder. At low volume the honest answer is that it depends on two costs nobody puts on the slide: an AWS support plan if you want a human to answer, and the engineering weeks to build the console, event pipeline and suppression tooling that arrive free with SendGrid. Below roughly a few hundred thousand messages a month, those two can eat the entire saving. ### Can I keep SendGrid's SMTP integrations if I move to SES? Usually yes, because most of those integrations only want a host, a port and a username and password. SES provides all four. The work is in the details: SMTP credentials are generated from an IAM user rather than issued directly, the endpoint is specific to one region so a failover plan has to name a second, and a few older appliances are fussy about TLS negotiation on the submission port. Budget a day per awkward system, not a week. ### What happens to my marketing emails if I switch to SES? They need a home, because SES will not provide one. The common patterns are a dedicated marketing tool alongside SES, or a platform layer that adds campaigns on top of it. What you should not do is have a developer rebuild segmentation and a visual builder, because that project has a well-documented habit of consuming a quarter and producing something the marketing team refuses to use. ### Will my deliverability drop if I leave SendGrid? For a while, yes, and not because SES is worse. You would be a new sending identity with no history, which every mailbox provider treats cautiously until it has evidence. Ramp volume over a few weeks, start with your most engaged recipients, and have DKIM, SPF and DMARC aligned before the first message rather than after the first spam-folder complaint. Teams that move a large list in one evening are the ones who write the blog posts about how SES has bad deliverability. ### Does AWS suspend accounts the way SendGrid does? AWS will pause sending, and it does so on published bounce and complaint thresholds you can watch approach in your own metrics. That is the substantive difference: the mechanism is documented and the leading indicator is visible to you, so a team with a single CloudWatch alarm gets a week of warning. Whether the recovery conversation is faster depends entirely on whether you are paying for a support plan, which is the cost that keeps reappearing in this comparison. ### Does SES have anything equivalent to SendGrid subusers? Not as one feature. The pieces exist scattered across AWS primitives — separate verified identities, separate configuration sets, separate IAM principals, separate dedicated addresses — and assembling them into per-tenant credentials with per-tenant reporting and per-tenant suppression is a project rather than a setting. Agencies and platforms sending on behalf of customers should price that project honestly, because it is the single largest thing SendGrid gives you that SES does not, and it is considerably more work than the send path it sits around. ### Can we run both during the migration, and how long should the overlap be? Yes, and you should. Put each provider on its own subdomain so the DKIM keys, the reputation and the DMARC alignment stay independent, then move one traffic class at a time and leave the highest-stakes mail until last. The overlap is dictated by placement data rather than by engineering progress: you are watching complaint and bounce rates on the new identity climb into normal range, which usually takes several weeks at meaningful volume. Budget for two invoices during that period explicitly, because the alternative is discovering it at the end of the month. ### Do SendGrid's event webhooks map cleanly onto SES events? No, and assuming they do is a common way to lose a week. SendGrid posts a batch of events as a JSON array to one endpoint you nominate. SES emits individual notifications through SNS, EventBridge or Firehose, with a different envelope, different field names and a different set of event types — and you choose the destination and the delivery guarantees yourself. The consumer is new code, and somebody has to decide explicitly how the two vocabularies line up before the reporting built on the old shape can be trusted again.