IaaS · Bare metal against EC2
AWS alternativepriced against EC2, core for core.
Server Room is an AWS alternative for workloads that run all month. A dedicated server with the same number of cores and at least the same memory as an EC2 instance costs 76–86% less than that instance on demand, and 52–75% less than a three-year Savings Plan. Traffic on the port is unmetered.
- No egress bill
- IPv4 in the price
- Monthly, no contract
- Root on the whole machine
Below: the price table, what AWS charges for traffic out, how each figure was worked out, and when AWS is still the better buy.
IaaS pricing
IaaS pricing, machine by machine
Each row pairs an EC2 instance with one of our builds that has the same physical core count and at least as much memory. Our prices are what the checkout charges for that build today. AWS prices are its published Linux rates in US East (N. Virginia), read on 22 October 2026.
| Size | Server Room build | Server Room a month | EC2 instance | EC2 on demand a month | EC2 3-year plan a month |
|---|---|---|---|---|---|
| Same CPU generation on both sides | |||||
| 24 cores 96 GiB+ |
| $303.99 | c6a.12xlarge48 vCPU · 96 GiB 3rd gen AMD EPYC (Milan) | $1,363.13−77% | $669.97−54% |
| 64 cores 128 GiB+ |
| $374.98 | c6g.16xlarge64 vCPU · 128 GiB Graviton2 (Arm Neoverse N1) | $1,611.33−76% | $789.86−52% |
| AWS runs a newer CPU generation | |||||
| 4 cores 32 GiB+ |
| $62.58 | m7i.2xlarge8 vCPU · 32 GiB 4th gen Xeon (Sapphire Rapids) | $317.19−80% | $172.12−63% |
| 8 cores 32 GiB+ |
| $73.47 | c7i.4xlarge16 vCPU · 32 GiB 4th gen Xeon (Sapphire Rapids) | $564.87−86% | $295.31−75% |
| 8 cores 64 GiB+ |
| $107.67 | m7i.4xlarge16 vCPU · 64 GiB 4th gen Xeon (Sapphire Rapids) | $611.52−82% | $321.39−66% |
| 64 cores 512 GiB+ |
| $754.98 | m6i.32xlarge128 vCPU · 512 GiB 3rd gen Xeon (Ice Lake) | $4,528.77−83% | $2,318.28−67% |
The percentage under each AWS figure is how much less our build costs. AWS month = hourly rate × 730 hours, plus gp3 storage for the same usable capacity as our disks at $0.08 per GB, plus one public IPv4 at $0.005 an hour. The 3-year plan is a Compute Savings Plan with no upfront payment. Our side is billed monthly with no contract. Each build above is priced as a configured build, which pays a one-time setup fee of $29. Traffic is not in either column: see the next section.
Method
How the comparison is built
An EC2 alternative is only worth the move if the comparison is fair. This one matches cores, gives AWS the benefit of the doubt on memory, and says where its hardware is newer.
- 01
Cores, not vCPUs
On Intel and AMD instances one vCPU is one thread, half a physical core. A 16 vCPU instance is matched with an 8-core server. On Graviton one vCPU is a full core, so 64 vCPU meets 64 cores.
- 02
Memory at least equal
Our build takes the smallest memory step that is at least the instance's memory. On the EPYC row that gives us 128 GB against AWS's 96 GiB, and we still price the bigger build.
- 03
Disk and address counted
An instance comes with no disk. We add gp3 volumes equal to our usable capacity, a mirrored pair counted once, and the public IPv4 AWS bills by the hour. Our first IPv4 is in the price.
- 04
The generation gap, named
Two rows are like for like: EPYC Milan against EPYC Milan, and Neoverse N1 against Neoverse N1 (Graviton2 and Ampere Altra use the same Arm core). On the other four AWS has the newer Xeon, faster per core. Size for that before you move.
Traffic
AWS egress costs against an unmetered port
AWS charges for every gigabyte that leaves for the internet once the first 100 GB of the month are used: $0.09 per GB for the first 10 TB, then $0.085, $0.07 and $0.05 as volume grows. Every build in the table above comes with an unmetered 1 Gbps port, and there is no per-gigabyte line on our invoice.
| Sent out in a month | AWS egress bill | Server Room |
|---|---|---|
| 1 TB | $81 | $0 |
| 10 TB | $891 | $0 |
| 50 TB | $4,292.70 | $0 |
| 100 TB | $7,812.20 | $0 |
| 324 TB* | $20,086.20 | $0 |
| 3,240 TB** | $165,886.20 | +$529 |
1 TB = 1,000 GB. AWS tiers as published for US East (N. Virginia); the 100 GB allowance is shared across all your AWS services. * 324 TB is a 1 Gbps port running flat out for 30 days. ** 3,240 TB is 10 Gbps for 30 days; on our side that is the 10 Gbps port upgrade at a flat monthly price, still with no meter. Ports go up to 100 Gbps on some machines; see how 10 Gbps ports are priced.
Infrastructure as a service
An IaaS provider that rents you the whole machine
Infrastructure as a service means renting compute, storage and network instead of buying them. On a hyperscaler that is a slice of a host, billed by the second. Here it is the host: every core, all the memory and the disks, on a monthly bill that does not move with traffic.
Build it in the server configurator: processor, memory, disks and RAID level, port speed, operating system and the city. Our dedicated servers are in New York, Bucharest, San Francisco, Amsterdam and Miami, on our own network.
- RootFull root or Administrator on the server. Install a hypervisor and run your own VMs if you want them.
- KubernetesA managed Kubernetes cluster on bare metal, for teams leaving EKS.
- Virtual machinesSmaller workloads fit a cloud VPS on the same network.
- PortsUnmetered on every machine, from 1 Gbps on the builds above.
- SupportOur AI on chat answers in seconds and can check your account and your server; a person takes over the moment you ask. Faster hardware response is on the management plans.
Fair comparison
When AWS is still the better choice
The saving comes from running a machine all month. Where that is not the shape of the work, stay where you are, or move only part of it.
Load that comes and goes
A job that runs two hours a day is cheaper on per-second billing. A server is paid for whether it is busy or not.
Managed services you rely on
RDS, Lambda, DynamoDB and the rest have no drop-in twin on a server. You would run the database yourself.
Regions we do not have
AWS has regions in Asia, South America, Africa and Australia. If your users sit in Asia or South America, our cities are not the closest.
The common answer is both: keep the bursty part on AWS and move the steady base, the databases, encoders, build runners and storage that are busy around the clock, onto servers. Inbound traffic to AWS is free, so the link between the two only costs you on the way out.
Moving
Moving off AWS in three steps
- 01
Size it
Take the instance you run today, match the cores and memory in the table above or in the configurator. A ready machine is delivered in about 30 minutes after approval.
- 02
Copy the data
Getting your data out of AWS is billed as egress, once. The migration planner works out that bill and the overlap month before you start.
- 03
Cut over
Run both side by side, move DNS, then switch off the instances. Want a second opinion on the design first? Book a call with us.
Answers
IaaS and AWS questions
Is a dedicated server cheaper than AWS?
For a machine that runs all month, yes. In the comparisons on this page a dedicated server with the same cores and at least the same memory costs 76–86% less than the EC2 instance on demand and 52–75% less than a three-year Savings Plan, before any egress is counted.
What is IaaS?
Infrastructure as a service is renting servers, storage and network from a provider instead of owning them. You run the operating system and everything above it; the provider runs the hardware, power and network. AWS EC2 is IaaS, and so is a rented dedicated server.
How much does IaaS cost?
It depends on how it is billed. Cloud IaaS charges by the hour or second for the instance, by the GB-month for disks, and by the GB for traffic out. A dedicated server is one monthly price with the disks, an IPv4 address and an unmetered port included; the table on this page puts both side by side.
What is an IaaS server?
A server you rent and control, either a virtual machine on a shared host or a whole physical machine. Ours are physical: the processor, memory and disks belong to your workload alone.
What does AWS charge for egress?
In US East (N. Virginia), the first 100 GB a month are free, then $0.09 per GB up to 10 TB, $0.085 for the next 40 TB, $0.07 for the next 100 TB and $0.05 beyond that. Sending 50 TB out costs about $4,293 a month. Traffic on our ports is not metered.
Why is AWS so expensive for steady workloads?
Because you pay for the ability to stop. Per-second billing, instant scaling and dozens of regions are built into every hourly rate. A workload that never stops pays for that flexibility without using it.
Can I keep AWS and add bare metal?
Yes. Many teams keep the parts that scale up and down on AWS and move the busy, steady parts to servers. Traffic into AWS is free; only what AWS sends back out is billed.
Who helps when something breaks?
Our AI on live chat answers in seconds, can look up your account and your server, and hands you to a person whenever you ask. Hardware faults are covered on every server; the management plans shorten the response time.
Price your AWS bill on bare metal
Pick the cores, memory and disks you run on EC2 today and see the monthly price before you order.