In a major expansion of its cloud computing capabilities, Amazon Web Services (AWS) has officially announced the general availability of its new Amazon Elastic Compute Cloud (Amazon EC2) G7 instances. This release marks a significant milestone in cloud infrastructure, as AWS becomes the first major cloud provider to integrate the powerful NVIDIA RTX PRO 4500 Blackwell Server Edition GPUs. Designed to handle the most demanding artificial intelligence (AI) inference, graphics rendering, and data analytics tasks, the G7 instances represent a substantial architectural upgrade over the previous G6 generation.
Main Facts: The Power Under the Hood
The launch of the G7 instance family is aimed at organizations struggling to balance the intense computational demands of generative AI, high-fidelity spatial computing, and real-time analytics with the operational realities of cloud costs and efficiency.
At the core of the G7 instance architecture is the pairing of custom sixth-generation Intel Xeon Scalable processors with the NVIDIA Blackwell-based RTX PRO 4500 GPUs. This synergy delivers staggering performance gains: users can expect up to 4.6 times faster AI inference performance and a 2.1 times improvement in graphics-intensive workloads compared to the outgoing G6 instances.
The instances are highly configurable, supporting up to eight NVIDIA RTX PRO 4500 GPUs, offering a total of 256 GB of GPU memory. For compute-heavy tasks, the instances can scale up to 192 vCPUs and 768 GiB of system memory, underpinned by a massive 700 Gbps of network bandwidth. Furthermore, the inclusion of up to 7.6 TB of local NVMe SSD storage ensures that data-intensive applications, such as video transcoding and large-scale data analytics, face minimal I/O bottlenecks.
Chronology of Development
The arrival of the G7 instances is the culmination of a multi-year effort by AWS to bridge the gap between high-end workstation performance and the scalability of the cloud.
- The G-Series Foundation: AWS established the G-series as the gold standard for GPU-accelerated cloud workloads, starting with early iterations designed primarily for graphics rendering.
- The Rise of Inference: As AI models moved from experimental labs to production environments, the focus shifted toward "inference"—the process of running a trained model. The G6 instances provided a solid footing for this, but as LLMs (Large Language Models) became more complex, a performance bottleneck emerged.
- The Blackwell Breakthrough: Following NVIDIA’s announcement of the Blackwell architecture, AWS engineers began the extensive integration process to bring this technology to the cloud. By optimizing the custom Intel Xeon Scalable processors to communicate seamlessly with the Blackwell GPUs, AWS created a high-bandwidth ecosystem that supports low-latency communication between nodes.
- General Availability: After months of internal testing and select preview programs, AWS officially opened the G7 instances to the public in October 2025, starting with the US East (Ohio) and US West (Oregon) regions.
Supporting Data: Technical Specifications and Performance
The versatility of the G7 family is reflected in its diverse instance sizes, ranging from the entry-level g7.2xlarge to the massive g7.48xlarge. The following table details the technical breakdown of these new instances:
| Instance Name | GPUs | GPU Memory | vCPUs | Memory (GiB) | Storage | Network Bandwidth |
|---|---|---|---|---|---|---|
| g7.2xlarge | 1 | 32 GB | 8 | 32 | 1×600 GB | Up to 60 Gbps |
| g7.4xlarge | 1 | 32 GB | 16 | 64 | 1×600 GB | Up to 100 Gbps |
| g7.8xlarge | 1 | 32 GB | 32 | 128 | 1×950 GB | Up to 100 Gbps |
| g7.12xlarge | 2 | 64 GB | 48 | 192 | 1×1900 GB | 175 Gbps |
| g7.24xlarge | 4 | 128 GB | 96 | 384 | 1×3800 GB | 350 Gbps |
| g7.48xlarge | 8 | 256 GB | 192 | 768 | 2×3800 GB | 700 Gbps |
Beyond raw specs, the G7 instances are designed for high-performance interconnectivity. By supporting NVIDIA GPUDirect P2P and GPUDirect RDMA (Remote Direct Memory Access) with Elastic Fabric Adapter (EFA), these instances allow for low-latency, direct communication between GPUs. This is critical for distributed training and massive-scale inference, as it removes the CPU from the data-transfer loop, significantly reducing latency in multi-node clusters.
Official Perspectives and Ecosystem Integration
Daniel Abib, a key figure in the AWS EC2 product development team, emphasized that the G7 launch is not just about raw power, but about "democratizing high-performance GPU access." By offering multiple purchasing options—including On-Demand, Savings Plans, and Spot Instances—AWS is positioning the G7 as a cost-effective solution for startups and enterprises alike.
Integration is seamless. AWS has updated its Deep Learning AMIs (DLAMI) and NVIDIA Workstation AMIs to include prepackaged drivers, allowing developers to spin up G7 instances and begin work immediately. For Kubernetes users, the instances are fully compatible with Amazon EKS through the use of R595 NVIDIA drivers, ensuring that containerized AI applications can leverage the new hardware without extensive refactoring.
The support for industry-standard graphics libraries—such as DirectX, Vulkan, and OpenGL—further solidifies the G7’s role in Virtual Desktop Infrastructure (VDI) and spatial computing, where visual fidelity and frame rate consistency are paramount.
Implications for the Industry
The release of G7 instances carries profound implications for several key industries:
1. Accelerating Generative AI
As businesses race to integrate AI into their products, the cost and latency of inference are the primary barriers to entry. By delivering a 4.6x performance boost, G7 instances allow companies to run more complex models at a lower total cost of ownership (TCO). This could lead to a new wave of real-time AI applications that were previously considered too compute-intensive for cloud environments.
2. The Future of Virtualized Graphics
For the media and entertainment industry, the G7 instances provide a robust platform for high-end rendering and video transcoding. The ability to handle complex 3D scenes in the cloud allows remote creative teams to collaborate in real-time, effectively replacing local physical workstations with powerful, on-demand cloud assets.
3. Data Analytics and EMR
By integrating G7 instances with Amazon EMR and Amazon EKS, AWS is enabling data scientists to process massive datasets using GPU-accelerated frameworks. This will likely accelerate the transition of data analytics pipelines from CPU-based processing to GPU-based acceleration, dramatically shortening the time-to-insight for complex queries.
4. Strategic Regional Expansion
The initial availability in the US East and West regions is merely the beginning. AWS has integrated G7 tracking into its CloudFormation tools, signaling that global expansion will follow a systematic rollout. This allows global enterprises to plan their multi-region cloud strategy with the assurance that this cutting-edge hardware will soon be available in their local data centers.
Conclusion: A New Standard for Cloud Compute
The introduction of the G7 instances is a testament to the accelerating pace of innovation in cloud infrastructure. By becoming the first to market with the NVIDIA Blackwell-based RTX PRO 4500 GPUs, AWS has set a new performance ceiling for GPU-accelerated cloud computing.
As businesses continue to grapple with the demands of an AI-first world, the G7 instance family provides the necessary building blocks to scale efficiently. Whether it is powering the next generation of generative AI, facilitating complex scientific simulations, or enabling high-fidelity remote desktop experiences, the G7 instances are poised to become a foundational element of the modern digital enterprise. For those looking to optimize their cloud spend while boosting performance, the transition to G7 appears not just as an option, but as a strategic necessity.
