When did the m5 instance family launch?

The Amazon EC2 M5 family was introduced on November 28, 2017 at AWS re:Invent 2017. Built on custom Intel Xeon Platinum 8175 (Skylake) processors and the AWS Nitro System, M5 succeeded the M4 (2015) family in the general-purpose M line, whose lineage runs m1 (2006), m3 (2012), m4 (2015), m5 (2017), m6i/m6a/m6g (2021/2020), m7i/m7a/m7g (2023), m8g (2024), m8i/m8a (2025), and m9g (2025 preview, 2026 GA).

What was the first EC2 instance type?

The first Amazon EC2 instance type was m1.small, launched with the EC2 limited public beta on August 25, 2006, in a single Region (US East). The larger m1.large and m1.xlarge sizes followed on October 16, 2007, and the first specialized family, the compute-optimized c1 (High-CPU), arrived on May 29, 2008.

What are the AWS Graviton generations and when did each launch?

There are five AWS Graviton generations to date: Graviton1 with the A1 family at AWS re:Invent 2018 (November 26, 2018); Graviton2 with M6g/C6g/R6g (M6g GA May 11, 2020; previewed at re:Invent 2019); Graviton3 with C7g (GA May 23, 2022; previewed at re:Invent 2021), plus Graviton3E for HPC and networking (Hpc7g/C7gn, 2023); Graviton4 with R8g (GA July 9, 2024; previewed at re:Invent 2023); and Graviton5, announced with the M9g family in preview at AWS re:Invent 2025 (December 4, 2025), with M9g/M9gd reaching general availability on June 10, 2026, followed by the compute-optimized C9g/C9gd on June 30, 2026.

What do the letters and suffixes in EC2 instance names mean?

An instance family name is series + generation + options. The series is the first letter or letters (for example M = general purpose, C = compute optimized, R = memory optimized, P = GPU, I = storage). The number is the generation. The trailing option letters indicate the processor or capability: a = AMD, g = AWS Graviton, i = Intel, n = network and EBS optimized, d = local NVMe instance store, e = extra storage or memory, z = high CPU frequency, q = Qualcomm accelerators, flex = Flex instance, b = block storage optimization, and b*00 = NVIDIA Blackwell GPUs. So m5a is a general-purpose fifth-generation instance on AMD, and c6gn is a compute-optimized sixth-generation Graviton instance with enhanced networking.

What is the AWS Nitro System and when was it introduced?

The AWS Nitro System is the hardware and lightweight-hypervisor platform underlying all current-generation Amazon EC2 instances. It debuted with the C5 family, which became generally available on November 6, 2017, and was formally named the AWS Nitro System at AWS re:Invent 2017 (November 28, 2017). Nitro offloads virtualization, VPC networking, and EBS and instance storage to dedicated Nitro Cards, enforces a hardware root of trust with the Nitro Security Chip, and uses the lightweight Nitro Hypervisor, which is what makes bare metal instances, AWS Graviton instances, and hardware-accelerated networking possible.

When did the first AMD, Arm (AWS Graviton), and GPU instances launch on Amazon EC2?

The first GPU instances were the Cluster GPU (cg1) family on November 15, 2010 (NVIDIA Tesla M2050). The first AMD EPYC-based instances were m5a and r5a on November 6, 2018, and the first AMD-GPU instances were G4ad on December 9, 2020. The first Arm-based instances were the A1 family, powered by first-generation AWS Graviton, at AWS re:Invent 2018 (November 26, 2018).

What is the difference between AWS Graviton and x86 instances?

AWS Graviton instances use AWS-designed 64-bit Arm processors and carry the g suffix (for example m7g, c7g, r8g), while x86 instances use Intel processors (the i suffix, for example m7i) or AMD processors (the a suffix, for example m7a). All three run on the AWS Nitro System. AWS Graviton has advanced through five generations: Graviton1 (A1, 2018), Graviton2 (2019 to 2020), Graviton3 and Graviton3E (2022 to 2023), Graviton4 (2024), and Graviton5 (M9g preview 2025, GA 2026). Choosing between them is primarily a matter of application and binary compatibility (Arm versus x86) and price-performance for the specific workload.

Amazon EC2 Instance Types History and Timeline - Instance Families, Generations, and Introduction

First Published: 2026-07-05
Last Updated: 2026-07-05

This is a companion to the earlier AWS History and Timeline regarding Amazon EC2, which tracked Amazon EC2 as a service. This article zooms in on one axis of that history — the instance types — and follows how the Amazon EC2 instance families and generations were introduced, one launch at a time, from the single m1.small of 2006 to the AWS Graviton5, NVIDIA Blackwell, and AWS Trainium3 families of the mid-2020s.

The official Amazon EC2 documentation lists only the current generation of instance families; the introduction dates of each generation are scattered across nearly two decades of announcements. This article gathers them into one chronological reference so that questions such as "when did the m5 family launch?", "how many AWS Graviton generations are there?", or "what does the a in m5a mean?" can be answered from a single page.

This article focuses on the arrival of instance families and generations, the naming convention, and the processor lineages as a Current Overview, Functions, Features of Amazon EC2 Instance Types. It intentionally does not duplicate the service-level history (purchasing options, regions, networking) that the Amazon EC2 service timeline already covers.

Background and Method of Creating Amazon EC2 Instance Types Historical Timeline

Amazon EC2 has expanded along many axes since 2006, but the instance types are where its hardware evolution is most visible: new CPU generations (Intel, AMD), the AWS Nitro System, AWS Graviton Arm processors, purpose-built accelerators (AWS Inferentia, AWS Trainium), third-party accelerators (NVIDIA, AMD, Xilinx/AMD FPGA, Habana, Qualcomm), and Apple Silicon for macOS all surface first as new instance families.

Tracking the history of Amazon EC2 instance families and organizing the transition of generations
Summarizing the current instance family catalog, the naming convention, and the processor lineages

This timeline primarily references the following official sources for Amazon EC2.

A note on granularity (important): each row in the timeline below represents the introduction of an instance family generation (for example, the general availability of the m5 family). To keep the reference readable, additional instance sizes, per-Region rollouts, and minor derivative variants (for example m5n, m5zn, c5a, c6gd, t3a) are not given their own rows — representative derivatives such as the first AMD family (m5a) and the first 100 Gbps family (c5n) are included, and the rest are noted in prose.

The content posted is limited to major instance family launches related to the current Amazon EC2 and necessary for the family catalog and naming overview.
In other words, please note that the items on this timeline are not all Amazon EC2 instance type updates, but are representative launches that I have picked out.
There may be slight variations in dates due to differences between the What's New posting date and the AWS News Blog posting date for the same launch.

Amazon EC2 Instance Types Historical Timeline (Updates from August 25, 2006)

Here is a timeline of Amazon EC2 instance family and generation launches, from August 2006 to the present. As of the time of writing, this history spans nearly 20 years.

2006 | 2007 | 2008 | 2009 | 2010 | 2011 | 2012 | 2013 | 2014 | 2015 | 2016 | 2017 | 2018 | 2019 | 2020 | 2021 | 2022 | 2023 | 2024 | 2025 | 2026

* The table can be sorted by clicking on the column names.

Date	Summary
2006-08-25	Amazon EC2 launches as a limited public beta with a single instance type, m1.small. This is the first EC2 instance type — a resizable Linux virtual server billed by the hour, in one Region (US East).
2007-10-16	Amazon EC2 adds its first larger sizes, m1.large and m1.xlarge. The first expansion of the instance lineup beyond m1.small, adding 64-bit instances.
2008-05-29	Amazon EC2 introduces the first compute-optimized High-CPU instances, c1.medium and c1.xlarge. The first time EC2 offered a hardware profile specialized for a workload class.
2009-10-27	Amazon EC2 introduces the High-Memory family (m2.2xlarge and m2.4xlarge). The first memory-optimized instances, for large in-memory workloads such as databases and caches.
2010-07-13	Amazon EC2 introduces Cluster Compute instances (cc1.4xlarge). The first HPC-oriented family, using 10 Gigabit Ethernet and cluster placement groups for low-latency communication.
2010-09-09	Amazon EC2 introduces Micro instances (t1.micro). Low-cost burstable capacity for low-throughput applications — the conceptual ancestor of the later T family.
2010-11-15	Amazon EC2 introduces the first GPU family, Cluster GPU (cg1.4xlarge). Powered by two NVIDIA Tesla M2050 GPUs, this was the first accelerated-computing instance on EC2.
2011-11-14	Amazon EC2 introduces second-generation Cluster Compute instances (cc2.8xlarge). Sandy Bridge processors with more cores for tightly coupled HPC.
2012-07-18	Amazon EC2 introduces High I/O instances (hi1.4xlarge) with local SSD storage. The first SSD-backed storage-optimized family, for high random IOPS workloads.
2012-08-01	Amazon EC2 introduces EBS-optimized instances. Alongside Provisioned IOPS EBS, these dedicate throughput between the instance and Amazon EBS; m1.large, m1.xlarge, and m2.4xlarge were the first supported types.
2012-10-31	Amazon EC2 introduces second-generation general-purpose instances (m3). m3.xlarge and m3.2xlarge with newer Intel Xeon processors; EBS-backed only at launch, with SSD instance storage added in the January 2014 M3 refresh.
2012-12-21	Amazon EC2 introduces High Storage instances (hs1.8xlarge) with 48 TB of local HDD. Targeted at data warehousing and large sequential-I/O workloads.
2013-01-21	Amazon EC2 introduces High-Memory Cluster instances (cr1.8xlarge). 244 GiB of memory with Intel Xeon E5-2670 for in-memory analytics.
2013-11-05	Amazon EC2 introduces the G2 GPU family. NVIDIA GRID (Kepler) GPUs for graphics-intensive and streaming applications.
2013-11-14	Amazon EC2 introduces the C3 compute-optimized family and Enhanced Networking with SR-IOV. SR-IOV, first available on C3, delivers higher packets-per-second and lower jitter.
2013-12-19	Amazon EC2 introduces the I2 storage-optimized family. SSD-backed instances for high random I/O such as transactional and NoSQL databases.
2014-04-10	Amazon EC2 introduces the R3 memory-optimized family. Ivy Bridge processors with SSD instance storage, succeeding the m2 and cr1 lines.
2014-07-01	Amazon EC2 introduces the T2 burstable family with the CPU credit model. Instances accrue CPU credits while idle and spend them during bursts.
2015-01-12	Amazon EC2 introduces the C4 compute-optimized family. Custom Intel Xeon E5-2666 v3 (Haswell) processors.
2015-03-31	Amazon EC2 introduces the D2 dense-storage family. Up to 48 TB of HDD per instance for data warehousing and Hadoop, succeeding hs1.
2015-06-11	Amazon EC2 introduces the M4 general-purpose family. Custom Intel Xeon E5-2676 v3 (Haswell) processors, up to 40 vCPUs (later 64).
2016-05-18	Amazon EC2 introduces the X1 family (x1.32xlarge) with 2 TB of memory. The largest EC2 memory-optimized instance at the time, targeted at SAP HANA and other large in-memory workloads.
2016-06-28	Amazon EC2 introduces the Elastic Network Adapter (ENA). A high-performance SR-IOV network interface, initially delivering up to 20 Gbps on X1 and scaling to 100+ Gbps on later generations.
2016-09-29	Amazon EC2 introduces the P2 GPU family (NVIDIA Tesla K80). Aimed at machine learning training and HPC.
2016-11-30	Amazon EC2 introduces the R4 memory-optimized family. Broadwell processors; the F1 FPGA family and the C5 family were also previewed at AWS re:Invent 2016.
2017-02-23	Amazon EC2 introduces the I3 storage family with local NVMe SSD. Higher IOPS and throughput than I2 for storage-bound databases.
2017-04-19	Amazon EC2 F1 FPGA instances become generally available. Xilinx UltraScale+ FPGAs for genomics, analytics, and video acceleration — the first FPGA family.
2017-07-13	Amazon EC2 introduces the G3 GPU family (NVIDIA Tesla M60). For graphics workstations and 3D visualization.
2017-09-13	Amazon EC2 introduces the X1e family (x1e.32xlarge) with up to 3,904 GiB of memory. An even higher memory-per-vCPU ratio than X1.
2017-10-25	Amazon EC2 introduces the P3 GPU family (NVIDIA Tesla V100). A major step up for deep-learning training.
2017-11-06	Amazon EC2 C5 instances become the first family built on the AWS Nitro System. Previewed in November 2016, C5 (Intel Xeon Platinum, Skylake) moved virtualization, networking, and storage onto dedicated Nitro hardware.
2017-11-28	The AWS Nitro System is formally introduced and bare metal instances (i3.metal) enter preview. At AWS re:Invent 2017, AWS named the Nitro Cards, Nitro Security Chip, and Nitro Hypervisor that underpin current-generation instances. Bare metal i3.metal reached general availability in 2018.
2017-11-28	Amazon EC2 introduces the M5 general-purpose family (Nitro). Custom Intel Xeon Platinum 8175 (Skylake), the first Nitro-based general-purpose family. The H1 dense-HDD storage family was also introduced at re:Invent 2017.
2018-07-25	Amazon EC2 introduces the R5 memory-optimized family. Intel Xeon Platinum (Skylake), with higher memory-per-vCPU than R4.
2018-07-25	Amazon EC2 introduces the z1d high-frequency family. Custom Intel Xeon running at up to 4.0 GHz sustained, for EDA and licensing-constrained workloads.
2018-08-21	Amazon EC2 introduces the T3 burstable family (Nitro). Successor to T2 with Unlimited mode enabled by default.
2018-09-27	Amazon EC2 introduces High Memory instances with 6, 9, and 12 TB of memory. Bare metal instances purpose-built for the largest SAP HANA in-memory databases (the u-*tb1 line).
2018-11-06	Amazon EC2 introduces its first AMD EPYC-based instances (m5a and r5a). The `a` suffix denotes AMD processors, giving customers a lower-cost x86 option.
2018-11-26	Amazon EC2 introduces A1 instances powered by first-generation AWS Graviton (Arm) processors. At AWS re:Invent 2018, the first Arm-based EC2 family; the c5n family with 100 Gbps networking was also introduced.
2019-05-08	Amazon EC2 introduces the I3en storage family. Dense NVMe SSD with higher storage-per-vCPU and up to 100 Gbps networking.
2019-09-20	Amazon EC2 G4dn GPU instances (NVIDIA T4) become generally available. Previewed in March 2019, aimed at cost-effective ML inference and graphics.
2019-12-03	Amazon EC2 introduces Inf1 instances powered by the AWS Inferentia inference chip. The first EC2 family built on an AWS-designed ML accelerator; AWS Inferentia had been announced at re:Invent 2018.
2020-05-11	Amazon EC2 M6g instances become the first generally available AWS Graviton2 family. Arm Neoverse N1; the C6g and R6g families followed on 2020-06-11. Graviton2 had been previewed at re:Invent 2019.
2020-09-14	Amazon EC2 introduces the T4g burstable family (AWS Graviton2). Arm burstable instances launched with a free-trial program.
2020-11-02	Amazon EC2 introduces the P4d GPU family (NVIDIA A100). Deployed in EC2 UltraClusters with 400 Gbps EFA networking for large-scale ML training.
2020-11-30	Amazon EC2 Mac instances (mac1, Intel) are introduced for macOS workloads. macOS on Apple Mac mini hardware connected to the AWS Nitro System, run on Dedicated Hosts.
2020-12-01	Amazon EC2 introduces the D3 and D3en dense-storage families. Cascade Lake processors with dense HDD; AWS Trainium was also announced at re:Invent 2020.
2020-12-09	Amazon EC2 introduces the G4ad family (AMD Radeon Pro V520). The first AMD-GPU-based EC2 family, for graphics workstations.
2021-03-16	Amazon EC2 introduces the X2gd memory family (AWS Graviton2). Arm memory-optimized instances with local NVMe.
2021-08-23	Amazon EC2 introduces the M6i family (3rd Gen Intel Xeon, Ice Lake). The Intel counterpart to the Graviton2 M6g generation.
2021-09-13	Amazon EC2 introduces the VT1 video transcoding family (Xilinx Alveo U30). Purpose-built for live video transcoding.
2021-10-26	Amazon EC2 introduces the DL1 training family (Habana Gaudi). Intel/Habana Gaudi accelerators for cost-efficient deep-learning training.
2021-10-28	Amazon EC2 introduces the C6i compute-optimized family (Ice Lake). The Intel counterpart to the Graviton2 C6g generation.
2021-11-12	Amazon EC2 introduces the G5 GPU family (NVIDIA A10G). For graphics and ML inference.
2021-11-22	Amazon EC2 introduces the R6i memory family (Ice Lake). The Intel counterpart to the Graviton2 R6g generation.
2021-11-29	Amazon EC2 introduces the G5g family (AWS Graviton2 plus NVIDIA T4G) and the M6a family (3rd Gen AMD EPYC). G5g pairs Arm CPUs with GPUs; M6a is the AMD counterpart of M6i.
2021-11-30	Amazon EC2 introduces the Im4gn and Is4gen storage families (AWS Graviton2, Nitro SSD). Dense NVMe storage on Arm, with vCPU-to-memory ratios of 1:4 (Im) and 1:6 (Is).
2022-01-10	Amazon EC2 introduces the Hpc6a HPC family (3rd Gen AMD EPYC). The first family in the dedicated Hpc series, optimized for tightly coupled HPC.
2022-02-14	Amazon EC2 introduces the C6a compute family (3rd Gen AMD EPYC). The AMD counterpart of C6i.
2022-03-10	Amazon EC2 introduces the X2idn and X2iedn memory families (Ice Lake). Large-memory Intel instances succeeding X1/X1e (the high-frequency X2iezn arrived on 2022-01-26).
2022-04-27	Amazon EC2 introduces the I4i storage family (Ice Lake, AWS Nitro SSD). Successor to I3 with AWS Nitro SSDs.
2022-05-23	Amazon EC2 C7g instances become the first generally available AWS Graviton3 family. Arm Neoverse V1 with DDR5 memory; previewed at re:Invent 2021.
2022-07-19	Amazon EC2 introduces the R6a memory family (3rd Gen AMD EPYC). The AMD counterpart of R6i.
2022-10-10	Amazon EC2 introduces Trn1 instances powered by the AWS Trainium training chip. The first EC2 family built on the AWS-designed training accelerator; Trainium had been announced at re:Invent 2020.
2022-11-29	Amazon EC2 introduces the Hpc6id HPC family (Ice Lake). Optimized for data-intensive, memory-bandwidth-bound HPC.
2023-02-13	Amazon EC2 introduces the M7g and R7g families (AWS Graviton3). The general-purpose and memory-optimized Graviton3 generations.
2023-04-13	Amazon EC2 Inf2 instances (AWS Inferentia2) become generally available. For large-scale generative-AI inference; the Trn1n networking-optimized training variant launched the same day.
2023-06-20	Amazon EC2 introduces the C7gn and Hpc7g families (AWS Graviton3E). Graviton3E adds higher vector and network throughput, with up to 200 Gbps networking.
2023-07-26	Amazon EC2 introduces the P5 GPU family (NVIDIA H100). For large generative-AI training and HPC in EC2 UltraClusters.
2023-08-02	Amazon EC2 introduces the M7i family and the new Flex (m7i-flex) instances. Sapphire Rapids Intel Xeon; the `flex` type trades a lower baseline for lower price on bursty workloads.
2023-08-16	Amazon EC2 introduces the M7a family (4th Gen AMD EPYC, Genoa). The AMD counterpart of M7i.
2023-08-17	Amazon EC2 introduces the Hpc7a HPC family (4th Gen AMD EPYC). DDR5 with up to 300 Gbps EFA networking.
2023-09-14	Amazon EC2 introduces the C7i compute family (Sapphire Rapids). Along with the R7a (2023-09-12), R7iz high-frequency (2023-09-07), C7a (2023-10-04), and R7i (2023-10-16) families, this completed the 4th-generation Intel and AMD lineup.
2023-10-25	Amazon EC2 introduces Apple M2 Mac instances (mac2-m2). Following mac2 (Apple M1, GA 2022-07-07) and mac2-m2pro (2023-09-19), extending macOS build capacity to Apple Silicon.
2023-11-15	Amazon EC2 introduces the DL2q family (Qualcomm Cloud AI 100). Qualcomm accelerators for inference and edge-AI development.
2023-11-28	AWS announces Graviton4 with R8g instances in preview at AWS re:Invent 2023. Arm Neoverse V2 (Armv9); AWS Trainium2 was announced the same week.
2024-04-04	Amazon EC2 introduces the G6 GPU family (NVIDIA L4). For graphics and mid-range ML inference.
2024-05-29	Amazon EC2 High Memory u7i instances (up to 32 TiB) become generally available. Sapphire Rapids with up to 896 vCPUs; previewed at re:Invent 2023.
2024-07-09	Amazon EC2 R8g instances become the first generally available AWS Graviton4 family. Arm Neoverse V2 with more cores and memory bandwidth than Graviton3.
2024-08-15	Amazon EC2 introduces the G6e family (NVIDIA L40S). Higher-end GPU for graphics and generative-AI inference.
2024-09-25	Amazon EC2 introduces the M8g and C8g families (AWS Graviton4). The general-purpose and compute-optimized Graviton4 generations; the X8g memory family launched on 2024-09-18.
2024-12-01	Amazon EC2 introduces the I7ie (Intel) and I8g (AWS Graviton4) high-density storage families. Both use third-generation AWS Nitro SSDs; announced at re:Invent 2024.
2024-12-03	Amazon EC2 Trn2 instances (AWS Trainium2) become generally available. With Trn2 UltraServers for the largest generative-AI training jobs.
2024-12-12	Amazon EC2 introduces the F2 FPGA family (AMD Virtex UltraScale+). The first new FPGA generation since F1, with eight FPGAs per instance.
2025-05-15	Amazon EC2 introduces the P6-B200 family (NVIDIA Blackwell B200). The `b*00` naming denotes NVIDIA Blackwell GPUs, for frontier-scale generative-AI training.
2025-08-19	Amazon EC2 introduces the R8i and R8i-flex families (custom Intel Xeon 6). The first of the 8th-generation Intel families, and the first memory-optimized Flex instances.
2025-08-28	Amazon EC2 introduces the M8i and M8i-flex families (custom Intel Xeon 6). The general-purpose members of the 8th-generation Intel line, succeeding M7i and M7i-flex.
2025-09-12	Amazon EC2 introduces Apple M4 and M4 Pro Mac instances. The latest Apple Silicon generation for macOS build and test pipelines.
2025-10-06	Amazon EC2 introduces the C8i and C8i-flex families (custom Intel Xeon 6). The compute-optimized members, completing the R8i/M8i/C8i trio of the 8th-generation Intel line.
2025-10-08	Amazon EC2 introduces the M8a family (5th Gen AMD EPYC, Turin). The AMD counterpart of M8i, succeeding the Genoa-based M7a.
2025-11-05	Amazon EC2 introduces the R8a family (5th Gen AMD EPYC, Turin). The memory-optimized Turin generation, succeeding R7a.
2025-12-02	Amazon EC2 introduces the C8a family (5th Gen AMD EPYC, Turin). Announced in the week of AWS re:Invent 2025, completing the M8a/R8a/C8a Turin trio.
2025-12-02	Amazon EC2 Trn3 UltraServers (AWS Trainium3) become generally available. Powered by Trainium3, the fourth-generation and first 3nm AWS AI chip, announced at AWS re:Invent 2025.
2025-12-04	AWS announces AWS Graviton5 with M9g instances in preview at AWS re:Invent 2025. The next Graviton generation, succeeding Graviton4, with M9g as the first general-purpose family.
2026-01-15	Amazon EC2 introduces the X8i memory family (custom Intel Xeon 6). The next-generation, SAP-certified extreme-memory X line on custom Intel Xeon 6.
2026-01-20	Amazon EC2 introduces the G7e GPU family (NVIDIA RTX PRO 6000 Blackwell). The graphics-and-inference G line moves to Blackwell-generation GPUs, succeeding G6/G6e.
2026-06-10	Amazon EC2 M9g and M9gd instances (AWS Graviton5) become generally available. The first generally available Graviton5 families, built on the sixth-generation AWS Nitro System, with M9gd adding local NVMe SSD storage.
2026-06-30	Amazon EC2 C9g and C9gd instances (AWS Graviton5) become generally available. The compute-optimized Graviton5 families, following the M9g/M9gd general-purpose launch.

Current Overview, Functions, Features of Amazon EC2 Instance Types

Amazon EC2 instance types are named for a hardware profile — a specific combination of CPU (or accelerator), memory, storage, and networking. Choosing the right type is the core of right-sizing, migration, and generation-refresh decisions. This section explains how to read an instance type name, catalogs the current families by category (by use case), and summarizes the processor lineages.

How to Read an Amazon EC2 Instance Type Name

An instance type name is composed of an instance family and an instance size, separated by a period. The instance family itself has three positions: the series (first letter or letters), the generation (a number), and the options (additional letters). For example, in m6gd.2xlarge: m is the series (M, general purpose), 6 is the generation, g is the AWS Graviton option, d is local NVMe instance store, and 2xlarge is the size. Bare metal instances use metal in place of a numeric size.

   m     6     gd   .   2xlarge
   |     |     |          |
 series  gen   options    size
   |            |
   M            +-- g = AWS Graviton processor
 general        +-- d = local NVMe instance store (SSD)
 purpose

The official series and option letters, per the Amazon EC2 instance type naming conventions, are as follows.

Series (first position)	Meaning
C	Compute optimized
D	Dense storage
F	FPGA
G	Graphics intensive
Hpc	High performance computing
I	Storage optimized
Im	Storage optimized (1:4 vCPU-to-memory ratio)
Is	Storage optimized (1:6 vCPU-to-memory ratio)
Inf	AWS Inferentia
M	General purpose
Mac	macOS
P	GPU accelerated
R	Memory optimized
T	Burstable performance
Trn	AWS Trainium
U	High memory
VT	Video transcoding
X	Memory intensive
Z	High memory (high CPU frequency)

Option letter(s)	Meaning
a	AMD processors
g	AWS Graviton processors
i	Intel processors
m / m*pro	Apple chip (Mac instances)
b00 / gb00	Accelerated by NVIDIA Blackwell GPUs
b	Block storage optimization
d	Instance store (local NVMe) volumes
e	Extra storage, memory, or GPU memory
flex	Flex instance (lower baseline, burstable)
n	Network and EBS optimized
q	Qualcomm inference accelerators
z	High CPU frequency
*tb	Amount of memory for high-memory instances (3 to 32 TiB)

Combinations are read left to right — for example p5en is a P-series generation 5 with the e (extra) and n (network optimized) options, and c6gn is a C-series generation 6 with the g (Graviton) and n (network) options. There is no single en code; it is the e and n options combined.

Amazon EC2 Instance Family Tree

The following table shows how the Amazon EC2 instance families are organized into five categories and the generation span each category covers; the processor lineages behind these generations are summarized in a later section.

Category	Representative Families (Generation Span)
General Purpose	M, T, A1, Mac (m1 to m9g)
Compute Optimized	C, Hpc (c1 to c9g)
Memory Optimized	R, X, U (High Memory), z1d (r3 to r8g)
Accelerated Computing	P, G, F, Inf, Trn, DL, VT
Storage Optimized	I, Im/Is, D, H1 (i2 to i8g)

This tree continues to grow as AWS introduces new processors (AWS Graviton, AWS Trainium, AWS Inferentia, Apple Silicon) and new specialized hardware (GPU, FPGA, and other AI accelerators).

Current Amazon EC2 Instance Family Catalog

The families are organized into five categories. The lists below describe what each category is for.

General Purpose

Balanced compute, memory, and networking for a broad range of workloads.

M family (M5, M6i/M6a/M6g, M7i/M7a/M7g, M8i/M8a/M8g, M9g): balanced instances for web and application servers, small-to-medium databases, and enterprise applications, on Intel (i), AMD (a), and AWS Graviton (g) variants.
T family (T2, T3/T3a, T4g): burstable instances that use a CPU credit model for workloads with low average utilization.
A1: the original AWS Graviton family, largely succeeded by the Graviton M/C/R generations.
Mac (mac1, mac2, mac2-m2/m2pro, and the M4 generation): macOS instances on Apple Mac hardware, run on Dedicated Hosts.

Compute Optimized

Highest ratio of compute to memory, for compute-bound workloads.

C family (C5, C6i/C6a/C6g, C7i/C7a/C7g, C7gn, C8i/C8a/C8g, C9g): for batch processing, media transcoding, high-traffic web serving, and scientific modeling, on Intel, AMD, and AWS Graviton variants.
Hpc family (Hpc6a, Hpc6id, Hpc7g, Hpc7a): purpose-built for tightly coupled HPC with high-bandwidth Elastic Fabric Adapter (EFA) networking.

Memory Optimized

Highest ratio of memory to compute, for in-memory databases and analytics.

R family (R4, R5, R6i/R6a/R6g, R7i/R7a/R7g/R7iz, R8i/R8a/R8g): general memory-optimized instances.
X family (X1, X1e, X2gd, X2idn/X2iedn/X2iezn, X8g/X8i) and the High Memory U series (u-6tb1 through u-24tb1 and the newer U7i line): extreme memory for SAP HANA and large in-memory databases.
z1d: high-frequency instances for electronic design automation (EDA) and certain licensing-constrained workloads.

Accelerated Computing

Instances with hardware accelerators (GPUs, FPGAs, and AWS-designed ML chips).

P family (P2, P3, P4d, P5/P5e/P5en, P6-B200): NVIDIA GPUs (K80 → V100 → A100 → H100 → H200 → Blackwell B200) for ML training and HPC.
G family (G2, G3, G4dn/G4ad, G5/G5g, G6/G6e, G7e): NVIDIA and AMD GPUs for graphics, remote workstations, and ML inference.
F family (F1, F2): Xilinx/AMD FPGAs for genomics, analytics, and video processing.
Inf family (Inf1, Inf2): AWS Inferentia and Inferentia2 accelerators for ML inference.
Trn family (Trn1, Trn1n, Trn2, Trn3): AWS Trainium, Trainium2, and Trainium3 accelerators for ML training.
DL and VT families (DL1 Habana Gaudi, DL2q Qualcomm, VT1 video transcoding): specialized accelerators.

Storage Optimized

Local high-IOPS or high-throughput storage attached to the host.

I family (I2, I3, I3en, I4i, I7ie, I8g) and Im4gn / Is4gen: NVMe SSD-backed instances for transactional and NoSQL databases.
D family (D2, D3, D3en) and H1: HDD-backed dense storage for data warehousing, Hadoop, and big data.

Amazon EC2 Processor Lineages

The instance-type history is, at its core, a history of processors. Five lineages run through the timeline above.

Intel Xeon: from Broadwell (G3/P3) and Skylake (C5/M5) through Cascade Lake (G4dn), Ice Lake (M6i), Sapphire Rapids (M7i, U7i), Emerald Rapids (I7ie, P5en), and custom Intel Xeon 6 (R8i/M8i/C8i, X8i). Intel-based families use the i suffix where an AMD or Graviton sibling exists (for example M6i).
AMD EPYC: from the first AMD instances (m5a/r5a, 2018) through Naples, Rome (G4ad, G5), Milan (C6a, F2), Genoa (M7a), and Turin (M8a/R8a/C8a). AMD x86 families use the a suffix.
AWS Graviton (Arm): Graviton1 (A1, 2018) → Graviton2 (M6g/C6g/R6g, 2019–2020) → Graviton3 (C7g, 2022) → Graviton3E (Hpc7g/C7gn, 2023) → Graviton4 (R8g, 2024) → Graviton5 (M9g preview 2025; M9g/M9gd and C9g/C9gd GA 2026). Graviton families use the g suffix and run on the AWS Nitro System.
AWS Inferentia and AWS Trainium: AWS-designed ML chips — Inferentia (Inf1, 2019) and Inferentia2 (Inf2, 2023) for inference; Trainium (Trn1, 2022), Trainium2 (Trn2, 2024), and Trainium3 (Trn3 UltraServers, 2025) for training. They share the AWS Neuron SDK.
Apple Silicon: Intel (mac1), then Apple M1 (mac2), M2 (mac2-m2/m2pro), and M4 for macOS instances.

All current-generation Amazon EC2 instances — x86, Arm, and accelerator-based — run on the AWS Nitro System, which offloads virtualization, networking, and storage to dedicated hardware.

References:
AWS Documentation (Amazon EC2 instance type naming conventions)
Amazon EC2 Instance Types
AWS Graviton Processor
The AWS Nitro System

Frequently Asked Questions about Amazon EC2 Instance Types History

When did the m5 instance family launch?: The Amazon EC2 M5 family was introduced on November 28, 2017 at AWS re:Invent 2017. Built on custom Intel Xeon Platinum 8175 (Skylake) processors and the AWS Nitro System, M5 succeeded the M4 (2015) family in the general-purpose M line, whose lineage runs m1 (2006), m3 (2012), m4 (2015), m5 (2017), m6i/m6a/m6g (2021/2020), m7i/m7a/m7g (2023), m8g (2024), m8i/m8a (2025), and m9g (2025 preview, 2026 GA).
What was the first EC2 instance type?: The first Amazon EC2 instance type was m1.small, launched with the EC2 limited public beta on August 25, 2006, in a single Region (US East). The larger m1.large and m1.xlarge sizes followed on October 16, 2007, and the first specialized family, the compute-optimized c1 (High-CPU), arrived on May 29, 2008.
What are the AWS Graviton generations and when did each launch?: There are five AWS Graviton generations to date: Graviton1 with the A1 family at AWS re:Invent 2018 (November 26, 2018); Graviton2 with M6g/C6g/R6g (M6g GA May 11, 2020; previewed at re:Invent 2019); Graviton3 with C7g (GA May 23, 2022; previewed at re:Invent 2021), plus Graviton3E for HPC and networking (Hpc7g/C7gn, 2023); Graviton4 with R8g (GA July 9, 2024; previewed at re:Invent 2023); and Graviton5, announced with the M9g family in preview at AWS re:Invent 2025 (December 4, 2025), with M9g/M9gd reaching general availability on June 10, 2026, followed by the compute-optimized C9g/C9gd on June 30, 2026.
What do the letters and suffixes in EC2 instance names mean?: An instance family name is series + generation + options. The series is the first letter or letters (for example M = general purpose, C = compute optimized, R = memory optimized, P = GPU, I = storage). The number is the generation. The trailing option letters indicate the processor or capability: a = AMD, g = AWS Graviton, i = Intel, n = network and EBS optimized, d = local NVMe instance store, e = extra storage or memory, z = high CPU frequency, q = Qualcomm accelerators, flex = Flex instance, b = block storage optimization, and b*00 = NVIDIA Blackwell GPUs. So m5a is a general-purpose fifth-generation instance on AMD, and c6gn is a compute-optimized sixth-generation Graviton instance with enhanced networking.
What is the AWS Nitro System and when was it introduced?: The AWS Nitro System is the hardware and lightweight-hypervisor platform underlying all current-generation Amazon EC2 instances. It debuted with the C5 family, which became generally available on November 6, 2017, and was formally named the "AWS Nitro System" at AWS re:Invent 2017 (November 28, 2017). Nitro offloads virtualization, VPC networking, and EBS/instance storage to dedicated Nitro Cards, enforces a hardware root of trust with the Nitro Security Chip, and uses the lightweight Nitro Hypervisor — which is what makes bare metal instances, AWS Graviton instances, and hardware-accelerated networking possible.
When did the first AMD, Arm (AWS Graviton), and GPU instances launch on Amazon EC2?: The first GPU instances were the Cluster GPU (cg1) family on November 15, 2010 (NVIDIA Tesla M2050). The first AMD EPYC-based instances were m5a and r5a on November 6, 2018, and the first AMD-GPU instances were G4ad on December 9, 2020. The first Arm-based instances were the A1 family, powered by first-generation AWS Graviton, at AWS re:Invent 2018 (November 26, 2018).
What is the difference between AWS Graviton and x86 instances?: AWS Graviton instances use AWS-designed 64-bit Arm processors and carry the g suffix (for example m7g, c7g, r8g), while x86 instances use Intel processors (the i suffix, for example m7i) or AMD processors (the a suffix, for example m7a). All three run on the AWS Nitro System. AWS Graviton has advanced through five generations — Graviton1 (A1, 2018), Graviton2 (2019–2020), Graviton3 and Graviton3E (2022–2023), Graviton4 (2024), and Graviton5 (M9g preview 2025, GA 2026). Choosing between them is primarily a matter of application and binary compatibility (Arm versus x86) and price-performance for the specific workload.

All entries in the timeline above are linked to their primary AWS source (the What's New announcement or the AWS News Blog post). This article is reviewed regularly to incorporate new Amazon EC2 instance family launches.

Summary

In this article, I gathered the Amazon EC2 instance type history into a single timeline — the introduction of each instance family and generation from m1.small in 2006 to the AWS Graviton5, NVIDIA Blackwell, and AWS Trainium3 families of the mid-2020s, together with the naming convention and the Intel, AMD, AWS Graviton, AWS Inferentia/Trainium, and Apple Silicon processor lineages.

Read alongside the service-level AWS History and Timeline regarding Amazon EC2, which covers purchasing options, networking, regions, and the broader service history, and the AWS History and Timeline regarding Amazon EBS for the storage layer that most instances attach to.

In addition, there is also a historical timeline of all AWS services including services other than Amazon EC2, so please have a look if you are interested.

AWS History and Timeline - Almost All AWS Services List, Announcements, General Availability(GA)

This timeline will be updated as Amazon EC2 continues to evolve.

References:
AWS Documentation(Amazon EC2 Documentation)
AWS Documentation(Amazon EC2 Instance Types)
AWS Documentation(Amazon EC2 instance type naming conventions)
AWS Nitro System
AWS Graviton Processor
AWS Trainium
AWS Inferentia
What's New with AWS?
AWS News Blog

References:
Tech Blog with curated related content

Written by Hidekazu Konishi