IBM server providers (IBM Cloud, IBM Power Virtual Server, IBM LinuxONE, etc.) bill compute resources mainly based on time + allocated capacity, not just usage. This is important: you are usually paying for reserved CPU + memory capacity while it is running, even if it’s not fully utilized.

Here’s a clear breakdown:

🖥️ 💰 1. Core compute billing model

⚙️ A. Hourly billing (most common)

You are charged for:

• 🧠 vCPU / Power cores per hour
• 🧠 RAM (GB) per hour
• ⏱️ Billed continuously while server is running

📌 Formula:

Cost = (CPU hourly rate × hours) + (RAM hourly rate × hours)

📅 B. Monthly billing (PAYG aggregated)

Even in pay-as-you-go:

• IBM calculates usage hourly
• Then totals it monthly (~730 hours)

So:

You are still billed based on hourly consumption, not flat monthly usage.

📦 C. Reserved / subscription compute (discount model)

If you commit for 1–3 years:

• You reserve CPU + memory capacity
• Pay fixed discounted rate

💡 Savings: 15%–60% cheaper than on-demand

🧠 2. What exactly is billed in compute

🔹 CPU (vCPU / Power cores)

• Charged per core-hour
• Dedicated or shared cores
• SAP/enterprise workloads usually use dedicated cores

👉 Example:

• 8 cores × hourly rate × runtime

🔹 Memory (RAM)

• Charged per GB-hour
• Major cost driver for databases and SAP HANA

👉 Example:

• 256 GB RAM × hourly rate × runtime

🔹 OS + platform overhead

Depending on setup:

• AIX / IBM i / Linux included or bundled
• Some enterprise images include OS cost in compute rate

🔹 GPU / special compute (if used)

• Billed separately per GPU-hour
• Higher cost for AI/ML workloads

📊 3. Example compute billing (realistic IBM model)

Medium enterprise server:

• 8 cores
• 128 GB RAM
• Running 1 month (~730 hours)

Assume typical IBM rates:

• CPU: ~$0.50/core-hour
• RAM: ~$0.02/GB-hour

Calculation:

CPU cost:

• 8 × 0.50 × 730 ≈ $2,920

RAM cost:

• 128 × 0.02 × 730 ≈ $1,868

💵 Total compute cost:

👉 ~$4,800/month (compute only)
(Storage + bandwidth extra)

⚙️ 4. Important billing characteristics

🔴 1. You pay for allocated capacity, not usage

• Even if CPU is idle → still billed
• Even if memory is unused → still billed

🔴 2. Billing continues while VM is running

• Start = billing starts
• Stop = billing stops (for most PAYG VMs)

🔴 3. Overprovisioning increases cost

• Bigger RAM = linear cost increase
• Bigger cores = linear cost increase

🔴 4. Enterprise setups may include minimum commitments

• SAP / Power systems often require minimum reserved capacity

📌 5. Simple rule of thumb

💡 IBM compute cost ≈ $100 – $600 per core per month (effective blended cost)
plus memory charges

So:

• Small VM → $20–$300/month
• Enterprise VM → $1K–$10K/month
• SAP Power system → $10K–$100K+/month

🧾 6. Final summary

IBM providers bill compute resources using:

• 🧠 CPU cores → per core-hour
• 🧠 Memory → per GB-hour
• ⏱️ Time → hourly usage (billed continuously)
• 📦 Optional → reserved capacity discounts (1–3 years)

👉 Key idea:

You are paying for reserved hardware capacity, not actual CPU usage