Compute that spins up before your coffee does.
CPU instances, bare-metal servers, and high-memory nodes on one platform. Deploy in seconds, scale by the core, and pay only for what you run.
CPU Cloud
Rent CPU cores by the second
Spin up a single vCPU or a 192-core node on demand. Latest-gen AMD EPYC and Intel Xeon, NVMe-backed storage and private networking out of the box.
- AMD EPYC & Intel Xeon instances
- Per-second billing, no minimums
- Pre-baked Ubuntu, Docker and build images
- CPU
- AMD EPYC 9554
- Cores
- 64c / 128t
- Clock
- 3.1–3.75 GHz
- RAM
- 256 GB DDR5
- Storage
- 1.5 TB NVMe
- Network
- 25 GbE
Dedicated
Bare-metal servers, fully yours
Single-tenant machines with no hypervisor tax. Full physical cores, dedicated RAM and local NVMe, reserved for hours or for years.
- Up to 384 dedicated cores per node
- 10–100 GbE private interconnect
- Hourly, monthly or reserved terms
High-Memory
RAM by the terabyte
Memory-optimized instances with up to 4 TB of DDR5 for in-memory databases, caches and large-scale analytics. Huge pages on, latency down.
- 128 GB to 4 TB DDR5 per instance
- Tuned for Redis, Spark and in-memory DBs
- Transparent huge pages + NUMA pinning
Pricing
Straightforward, pay-as-you-go
No egress fees, no lock-in. Pick a category and see live sample rates.
| c7a.large | 2 | 4 GB | $0.034 | High |
| c7a.xlarge | 4 | 8 GB | $0.068 | High |
| c7a.4xlarge | 16 | 32 GB | $0.272 | High |
| m7i.8xlarge | 32 | 128 GB | $0.610 | Medium |
| c7a.16xlarge | 64 | 128 GB | $1.088 | High |
| c7a.24xlarge | 96 | 192 GB | $1.632 | Medium |
| c7a.48xlarge | 192 | 384 GB | $3.264 | Low |
| Tier | CPU | RAM | Storage | $/hr |
|---|---|---|---|---|
| BM-Standard | 1× EPYC 9354 (32c) | 256 GB | 2× 1.9TB NVMe | $1.60 |
| BM-Compute | 2× EPYC 9554 (128c) | 768 GB | 4× 3.8TB NVMe | $3.90 |
| BM-Memory | 2× EPYC 9654 (192c) | 1.5 TB | 2× 3.8TB NVMe | $5.20 |
| BM-Storage | 1× EPYC 9124 (16c) | 256 GB | 12× 16TB HDD | $2.40 |
| Instance | RAM | vCPU | $/hr |
|---|---|---|---|
| r7a.4xlarge | 128 GB | 16 | $0.42 |
| r7a.8xlarge | 256 GB | 32 | $0.84 |
| r7a.16xlarge | 512 GB | 64 | $1.68 |
| x7a.32xlarge | 1 TB | 96 | $3.90 |
| x7a.metal-48 | 2 TB | 192 | $7.80 |
| u7a.hi | 4 TB | 224 | $15.20 |
Sample on-demand rates · billed per second.
How it works
From zero to deployed in three steps
- 01
Pick your compute
Choose a CPU instance, a bare-metal tier, or a high-memory node from the console or the CLI. Cores, RAM and price are shown live before you commit.
- 02
Deploy in seconds
Launch from a pre-baked image or your own container. Nodes boot in under a minute with storage, networking and SSH already wired up.
- 03
Scale and pay per use
Autoscale cores and memory across regions, snapshot anytime, and tear down when you are done. Compute is billed by the second.
Built on solid metal
Infrastructure you can trust
Every node runs in tier-IV facilities with the redundancy your workloads expect.
Datacenter
Concurrently maintainable, fault-tolerant facilities.
Network
Non-blocking fabric with private VPC peering.
Security
Encrypted at rest and in transit, isolated tenancy.
SLA
Backed by service credits on every compute tier.
Storage
Up to 60 GB/s per node, plus regional object storage.
Support
Humans on call with a one-hour response target.
Your next build is one command away.
Create an account, drop in your SSH key, and ship to a live node in under a minute.