Amazon EC2 T8i Explained: Performance, Pricing and How It Compares

1. What Is Amazon EC2 T8i?

A new x86 option in the burstable T family

Amazon EC2 T8i is the latest x86-based member of AWS’s burstable T instance family. It runs on custom sixth-generation Intel Xeon Scalable processors, code-named Granite Rapids, and is built on the AWS Nitro System. The family currently includes nano, micro, small, and medium sizes.

T8i is designed for applications whose CPU demand is usually modest but can rise for short periods. Examples include development and staging environments, low-traffic websites, lightweight microservices, demo systems, login services, and some data-processing tasks.

The key word is burstable. Instead of requiring consistently high CPU capacity, a T8i instance provides a defined baseline level of CPU performance and can temporarily operate above that baseline when additional processing power is needed. This model is intended for variable workloads rather than applications that require continuously high CPU utilization.

2. How T8i's Burstable CPU Model Works

CPU credits connect baseline performance with temporary bursts

T8i uses CPU credits to control how much time an instance can spend above its baseline CPU performance. When CPU usage remains below the baseline, the instance earns credits. When it runs above the baseline, those credits are consumed. One CPU credit represents one vCPU running at 100% utilization for one minute.

In Standard mode, an instance can burst while credits are available. Once the balance is depleted, CPU performance is gradually limited toward the baseline.

T8i instances launch in Unlimited mode by default. In this mode, the instance can continue operating above its baseline even after its earned credits are exhausted by using surplus credits. If average CPU utilization remains above the baseline over the relevant 24-hour period, additional CPU charges can apply. This makes sustained CPU usage an important factor when evaluating the total cost of a T8i workload.

3. T8i Specifications at a Glance

Small instance sizes with newer-generation infrastructure

The T8i family currently offers four sizes, from nano to medium. Every size provides two vCPUs, while memory ranges from 0.5 GiB to 4 GiB. The baseline CPU level and credit-earning rate vary by instance size. The nano and micro configurations have lower baseline percentages and earn fewer CPU credits per hour, while small and medium both provide a 20% baseline per vCPU and earn 24 credits per hour.

Behind these compact configurations is newer infrastructure. T8i uses Intel Xeon Granite Rapids processors and Nitro v6. Network bandwidth can burst to 6.25 Gbps, compared with up to 5 Gbps on comparable T3 and T4g sizes, while EBS bandwidth can reach 5 Gbps.

These higher limits may be useful for workloads that occasionally generate short periods of heavier network or storage activity. Their practical impact, however, depends on how much network and EBS bandwidth the application actually uses.

4. T8i vs T3 vs T4g: What Actually Changes?

The biggest difference is not simply processor speed

T8i, T3, and T4g all belong to the burstable EC2 family, but they do not represent the same migration path. T8i and T3 use the x86_64 architecture, while T4g uses AWS Graviton2 processors based on Arm64. This distinction can affect operating systems, application binaries, container images, libraries, and other software dependencies. AWS currently lists Windows and Linux support for T8i and T3, while T4g is listed for Linux.

AWS reports that T8i can provide up to 70% higher compute performance, 1.25 times higher network bandwidth, 2.4 times higher EBS bandwidth, and up to 30% better price-performance than T3. These are AWS-reported maximum improvements, not a guarantee for every application. Actual results depend on workload behavior and configuration.

Moving from T3 to T8i keeps the same processor architecture and CPU-credit model, which can reduce some migration friction. Moving between x86 and T4g's Arm architecture requires an additional compatibility check. T3 and T4g also currently provide larger sizes, reaching 2xlarge, while T8i stops at medium.

5. T8i Pricing and the Cost Factors That Matter

The hourly instance rate is only one part of the calculation

T8i costs depend on the AWS Region, operating system, purchase option, runtime, and CPU usage pattern. With On-Demand pricing, there is no long-term commitment, and eligible EC2 usage is billed per second with a 60-second minimum. Spot Instances provide access to spare EC2 capacity at discounted rates, but they can be interrupted when AWS requires that capacity.

At launch, AWS made T8i available through On-Demand and Spot Instances and stated that Savings Plans support would follow. Because purchasing options can change after launch, current T8i eligibility should be verified when estimating production costs.

Unlimited mode adds another cost variable. AWS currently charges surplus T8i CPU credits at $0.06 per vCPU-hour for Linux, RHEL, and SLES and $0.1152 per vCPU-hour for Windows and Windows with SQL Web. These CPU-credit rates are consistent across T8i sizes, Regions, and supported purchase options.

As a result, two identical T8i instances can generate different monthly costs if their CPU usage patterns differ. Comparing expected total cost is therefore more useful than considering the advertised instance rate alone.

6. Where T8i Fits, and Where It May Not

Workload behavior matters more than the label

T8i is designed around workloads that spend much of their time at low or moderate CPU utilization but occasionally need additional processing capacity. AWS identifies development and test environments, microservices, low-traffic websites, login gateways, CI/CD pipelines, data-processing jobs, and short-duration compute tasks among its intended workloads.

The fit becomes less straightforward when CPU demand remains high for long periods. Sustained utilization can consume CPU credits and, in Unlimited mode, may introduce surplus-credit charges. Applications that require more than 4 GiB of memory or an instance larger than t8i.medium also fall outside the current T8i range.

Architecture is another consideration. Software that already depends on x86 can remain within that architecture with T8i, while applications that support Arm can also be evaluated against alternatives such as T4g. The useful comparison is therefore based on actual CPU patterns, memory requirements, software compatibility, and expected operating cost rather than instance generation alone.

7. Current Availability and Limitations

T8i is generally available, but the initial footprint has boundaries

As of September 21, 2026, T8i is available in 16 AWS Regions across the United States, Asia Pacific, Canada, and Europe. The initial availability includes four US Regions, seven Asia Pacific Regions, Canada Central, and four European Regions. AWS may expand T8i availability over time, so Region support should be confirmed before planning a deployment.

The family currently stops at t8i.medium and supports shared tenancy only. AWS states that T8i does not support Dedicated tenancy or Dedicated Hosts. At launch, T8i was available through On-Demand and Spot Instances, with Savings Plans support announced as coming soon. t8i.micro and t8i.small are also available under the AWS Free Tier, subject to applicable Free Tier terms.

These limitations are worth checking early, particularly for workloads that require larger instance sizes, dedicated infrastructure, or purchasing strategies based on commitment discounts.

8. A Practical T8i Decision Checklist

Start with the workload, then evaluate the instance

Before choosing T8i, look at how the application behaves rather than starting with the instance specifications. Is CPU utilization usually low with occasional peaks, or does it remain high for extended periods? How frequently would the workload operate above its CPU baseline, and could sustained bursting result in additional Unlimited-mode charges?

Next, check the infrastructure requirements. Does the application depend on x86 software, or can it also run on Arm-based instances? Does the workload fit within T8i's current 0.5 to 4 GiB memory range and its nano to medium size range? Is T8i available in the required AWS Region and tenancy model?

Finally, compare the expected monthly cost using the workload's actual CPU behavior against relevant alternatives, such as an existing T3 deployment or T4g where Arm compatibility exists. These checks provide a more useful basis for evaluating T8i than processor generation or headline performance figures alone.

Conclusion: Where T8i Fits in the EC2 T Family

T8i brings newer Intel processors, updated Nitro infrastructure, and higher networking and EBS capabilities to AWS’s burstable x86 instance family. Those improvements expand the options available for variable workloads, but they do not remove the need for workload-specific evaluation.

CPU utilization patterns, memory requirements, software architecture, regional availability, and expected monthly cost should all be considered together. T8i is another option within the EC2 T family, and whether it fits depends on how closely its characteristics match the application running on it.