Graviton + Spot: The Cost-Cutting Stack

Graviton (ARM) is cheaper than x86. Spot is cheaper than on-demand. Combined: 70-85% off list price.

Graviton

EC2 Graviton instances are AWS's ARM-based processor family. They offer significantly better price-performance than x86 equivalents for most workloads. Combining Graviton with Spot pricing produces compounding savings: Graviton is cheaper than x86 on-demand, Spot Graviton is cheaper than on-demand Graviton. The combined discount can be 60% or more compared to x86 on-demand for compatible workloads.

What Graviton provides:

• ARM-based.: Graviton uses ARM instruction set rather than x86. The processor is designed specifically for cloud workloads. The architecture difference matters for compilation but rarely for application behavior.
• 20 to 40% cheaper than x86 equivalents.: Graviton instance types (c7g, m7g, r7g, etc.) are priced 20% to 40% lower than equivalent x86 types (c7i, m7i, r7i). The savings apply at on-demand rates; reserved and Spot pricing apply on top.
• Most modern languages support ARM out of the box.: Java, Go, Rust, Python, Node.js, Ruby all run on ARM Linux without changes. The compilation produces native ARM binaries; the runtime is identical to x86 from the application's perspective.
• Performance is competitive.: For most workloads, Graviton matches or exceeds x86 on raw performance. The combination of better performance and lower price produces strong price-performance.
• AWS-managed services support Graviton.: RDS, ElastiCache, OpenSearch, Lambda all offer Graviton variants. The savings extend beyond raw EC2; the team's whole AWS footprint can benefit.

Graviton is the right default for new workloads where ARM compatibility is not an issue. The savings are substantial; the operational impact is minimal.

Spot on Graviton

Spot pricing applies to Graviton instances just as it does to x86. Spot Graviton stacks the savings: Graviton's price-performance advantage plus Spot's discount produces extremely cheap compute for fault-tolerant workloads.

• Spot pricing for Graviton instances.: Spot prices for Graviton typically run 60% to 80% below on-demand. Combined with Graviton's already-lower on-demand rates, Spot Graviton is the cheapest practical compute on AWS.
• Stack the discounts.: The savings compound. A workload that costs $100/hour on x86 on-demand might run for $30 to $40 on Spot Graviton. The savings reduce the team's compute bill significantly.
• Spot pool is smaller.: Spot Graviton pools are smaller than equivalent x86 pools. Interruption rates can be higher; capacity availability is sometimes limited. The trade-off is real.
• Diversify across types.: The team uses multiple Graviton instance types (c7g, m7g, r7g) and multiple sizes. The diversification reduces the risk of any single pool's interruption affecting the workload. Mixed instance type policies handle this transparently.
• Combine with Auto Scaling Groups.: ASGs with mixed instance type policies can pull from many spot pools. Graviton plus x86 across multiple types produces resilient capacity at cost-optimal pricing.

Spot on Graviton is the maximum-discount strategy. The trade-offs (interruptions, smaller pools) are manageable for fault-tolerant workloads.

Compatibility

The transition to ARM requires testing. Most modern workloads work on Graviton without code changes; some have ARM-specific issues that need addressing. The discipline is testing in non-production before committing production workloads to Graviton.

• Test apps on ARM in dev.: Run the application on Graviton in dev or staging. Observe for issues; benchmark performance; verify all dependencies work. The investment is small; the production confidence is significant.
• Most just work.: The vast majority of modern applications run on Graviton without changes. Java applications, Go binaries built for ARM, Python with native dependencies that have ARM wheels. The "just works" rate is high.
• Some have ARM-specific issues.: Native dependencies that lack ARM builds. Custom assembly. Binary-only third-party libraries. Each is a specific issue that needs addressing; the alternatives are usually fixable.
• Container images: build multi-arch.: Container images can be built for both ARM and x86. Docker buildx and similar tools produce multi-arch manifests. The same image tag runs on either architecture transparently.
• Images run on either arch transparently.: The container runtime selects the right variant based on the host architecture. The deployment system does not need to know whether nodes are Graviton or x86; the right image is pulled automatically.

EC2 Spot Graviton strategy is one of the highest-leverage cost optimizations available for AWS workloads. Nova AI Ops integrates with EC2 inventory and cost data, surfaces workloads that are candidates for Graviton or Spot migration, and helps the team move toward the cheaper compute pool over time.