CPU on Akamai Cloud: Choose Shared, Dedicated, or High Memory
Match your workload to the right compute plan with predictable pricing, native private networking, and 99.99% uptime. Deploy in minutes through Cloud Manager, API, or CLI and resize as you scale.
Choose Dedicated CPU if you need guaranteed, competition‑free cores for production apps, CI/CD, databases, game servers, or other CPU‑intensive/latency‑sensitive services.
Choose Shared CPU if you want the lowest cost for dev/staging, low–medium traffic apps, workers, or workloads tolerant of occasional contention.
Choose High Memory if your app is memory‑bound (in‑memory DB/caching, analytics) and still benefits from dedicated CPUs.
Plans, fit, and key details
Dedicated CPU
Guaranteed, competition‑free cores for consistently high performance. Ideal for nearly all production apps that need predictable compute.
Best for: high‑traffic sites and APIs, CI/CD and build servers, busy databases, scientific computing, big data analysis, real‑time game servers, media transcoding, and CPU‑optimized ML/AI steps.
Generations:
G8 Dedicated (featured): Zen 5 cores, new 1:4 VM shapes, larger memory options, and usage‑based network transfer billing. Strong fit for enterprise‑grade, latency‑sensitive, and resource‑intensive apps. Learn more
G7 Dedicated: Zen 3 cores for consistent, enterprise‑grade performance.
G6 Dedicated: Balanced performance; good for development and basic production.
Our most affordable VMs with a strong price‑to‑performance ratio. CPU cores are shared; short bursts to 100% are fine, but sustained usage should average under 80%.
Best for: development/staging, blogs/marketing sites, forums, worker nodes, and production apps not affected by resource contention.
Memory‑optimized instances on dedicated CPUs for in‑memory workloads.
Best for: in‑memory databases and caches (Redis/Memcached), memory‑heavy analytics, and apps that need lots of RAM without paying for extra vCPUs or storage.
Compute plans include bundled egress in most regions (pooled across instances). Additional egress is typically $0.005/GB in the majority of regions. See the full pricing list.
If you’re comparing against hyperscalers like Azure, many customers find total costs lower with Akamai due to flat, predictable instance pricing and low egress fees. For a side‑by‑side projection, use the Cloud Cost Calculator.
Distributed compute: core, metro, and edge
Distributed compute places dedicated CPU instances in major metro locations to push latency‑sensitive services closer to users—especially where traditional cloud regions are distant or limited.
When to use: real‑time APIs, multiplayer game servers, media processing, CDN‑adjacent services, and regionalized apps that benefit from single‑digit‑millisecond RTT improvements.
Access: Enabled per account; talk to our team to deploy. Contact sales
Use GPU‑equipped instances for parallelized, accelerator‑friendly AI/ML work:
NVIDIA RTX PRO 6000 Blackwell Server Edition: High‑throughput, low‑latency inference, multimodal pipelines, concurrent model serving, and real‑time decision systems where TTFT and tokens/sec matter.
NVIDIA RTX 4000 Ada: Cost‑efficient entry to mid‑range inference, media processing, visualization, and dev/test of GPU pipelines.
NVIDIA Quadro RTX 6000: Larger VRAM and bandwidth for bigger model footprints, heavier batch inference, advanced rendering, and long‑running GPU workloads.
Best‑fit AI scenarios include LLM and VLM inference, real‑time RAG services, vision workloads (detection, tracking, OCR), speech‑to‑text/text‑to‑speech, and GPU‑accelerated data preprocessing. See GPU product details and GPU plans and pricing.
Note: Akamai Accelerated Compute uses NETINT VPUs and is purpose‑built for video transcoding pipelines. For AI acceleration, choose GPU plans.
Shared vs. Dedicated CPU for enterprise workloads
Performance consistency: Dedicated provides competition‑free cores and predictable latency; Shared favors cost and can see minor variance under contention.
Sustained usage: Dedicated supports 100% CPU all day; Shared should average under 80% sustained use (bursts to 100% are fine).
Workload impact: Dedicated is recommended for production apps, real‑time services, and CPU‑intensive jobs. Shared suits development, low–medium traffic apps, and tolerant workloads.
Network model: G6/G7 Dedicated include bundled transfer; G8 uses usage‑based billing. Shared includes bundled transfer that scales with plan size.
Comparisons and buyer guidance
Akamai vs Azure/Google for distributed and edge‑enabled workloads: Akamai adds metro‑level distributed compute to reduce latency and egress costs with simple, predictable pricing. Hyperscalers offer expansive managed services and broad ecosystems. If your priority is price‑performance at the edge and low egress, Akamai is compelling; if you require deep managed PaaS breadth, hyperscalers may fit better. Use our Cloud Cost Calculator to model TCO.
Akamai vs DigitalOcean for developer‑friendly compute and Dedicated CPU: Both focus on simplicity for developers. Akamai adds distributed compute in metro locations, low egress fees, and pooled transfer (plus usage‑based transfer on G8), with Dedicated CPU generations up to 256 vCPUs on G8 for higher‑end scaling needs.
Monitor and alert. Use Cloud Pulse metrics and Alerts. Monitoring
Reference architecture: distributed compute on Akamai Cloud
Core region (system of record): Dedicated/High Memory instances for databases, stateful services, and control planes.
Metro/distributed regions (latency‑sensitive tiers): Dedicated CPU for app servers, real‑time game servers, or media workers placed near users.
Edge and delivery: Optionally front with Akamai delivery and NodeBalancers for regional load balancing.
Private networking and security: VPC segmentation, Cloud Firewalls, and least‑privilege access.
Data and storage: Persistent volumes via Block Storage; object data via Object Storage.
Resilience: Health checks and failover for instances and IPs, plus backups and cross‑region restore. Configure failover
RFP criteria and KPIs for a distributed compute platform
Price‑performance: $/vCPU/$/GB RAM, included storage, and egress ($/GB), plus pooled transfer or usage‑based options.
Latency and placement: Regions plus metro/distributed availability; observed RTT to target markets.
Scale and shapes: vCPU counts (up to 256 on G8 Dedicated), memory/vCPU ratios (1:2 compute‑optimized, 1:4 general purpose), GPU/VPU options.
Network: Guaranteed bandwidth per plan, private networking (VPC), firewalls, IPv6, and failover options.
Reliability and support: 99.99% uptime, 24/7 support, maintenance transparency.
Operations: API/CLI tooling, metrics and alerting, backups/snapshots, and straightforward resizing/migrations.
Compliance and data locality: Ability to meet residency requirements using core and metro regions.
KPIs to track post‑deployment: p95/p99 latency, error rates, CPU steal/ready time (should be near zero on Dedicated), throughput per vCPU, cost per request/session/GB, and egress cost per GB.
Operational playbook for performance, uptime, and cost SLOs
Plan selection: Start with Dedicated CPU for production latency‑sensitive tiers; use Shared for dev/test; use High Memory for caches/analytics.
Baseline and rightsize: Benchmark CPU, RAM, and bandwidth; choose 1:2 (compute‑optimized) or 1:4 (general purpose) shapes accordingly.
Scale: Resize instances or add nodes; for containerized apps, add node pools to LKE.
Resilience: Enable backups; configure failover and health checks; periodically test recovery runbooks.
Observe and alert: Use Cloud Pulse metrics and Alerts to enforce SLOs.
Optimize cost: Leverage pooled transfer where available; understand usage‑based transfer on G8; minimize egress by regionalizing services; use the Cloud Cost Calculator.
Next steps
Compare plans and pricing. See compute pricing
Create your first instance. Get started
Need help choosing or want distributed regions enabled? Contact sales
Explore adjacent options:
- GPUs for AI/ML and visualization. GPU product details
- Accelerated Compute for media transcoding with VPUs. Accelerated Compute