Artificial Intelligence, High Performance Computing

AWS and NVIDIA will add two million GPUs by 2028

Avatar of Ivy Delaney
August 26, 2026 by Ivy Delaney

Amazon Web Services and NVIDIA will deploy two million additional NVIDIA GPUs across AWS’s global infrastructure by 2027-2028, addressing rapidly increasing demand for artificial intelligence computing power. Customers are now scaling AI workloads, including agentic AI, scientific discovery, and robotics, beyond experimental phases and into full production, requiring expanded infrastructure and broader model choices.

“Customers want the freedom to choose the best tools for their AI workloads,” said Matt Garman, CEO of AWS, “and they want confidence that everything works seamlessly together.” This expansion builds on 16 years of collaboration between the two companies to accelerate AI development and deployment, and was announced at NVIDIA GTC 2026.

AWS is the first major cloud provider to offer compute instances accelerated by NVIDIA RTX PRO 4500 Blackwell Server Edition GPUs for Amazon EC2 G7 instances which deliver 4.6 times AI inference performance and 2.1 times graphics performance compared to previous-generation G6 instances.

2 Million NVIDIA GPUs Expand AWS AI Infrastructure by 2028

AWS intends to integrate NVIDIA Vera CPU-based infrastructure into its services, expanding beyond graphics processing units to encompass central processing unit technology from NVIDIA. This move broadens the collaborative effort between the two companies, extending beyond GPUs to include a wider range of AI computing resources. The addition of Vera CPUs will provide customers with greater flexibility in selecting the optimal compute resources for their specific AI workloads, catering to diverse application requirements.

At NVIDIA GTC 2026, AWS announced plans to add more than 1 million NVIDIA GPUs starting in 2026. Since then, demand has exceeded those expectations. AWS and NVIDIA are also collaborating to build AI factories for the U.S. government, incorporating 100,000 GPUs on secure AWS infrastructure. These factories will be dedicated to running federal and national-security workloads, emphasizing the importance of secure and reliable AI infrastructure for critical government applications. The companies also plan to extend NVIDIA NVLink Fusion with custom NVIDIA high-bandwidth memory, aiming to enhance data transfer speeds and overall system performance.

Customers want the freedom to choose the best tools for their AI workloads, and they want confidence that everything works seamlessly together.

Matt Garman, CEO of AWS

NVIDIA Vera CPUs and NVLink Fusion Enhance AWS Compute Options

The integration of NVIDIA Vera CPUs into Amazon Web Services infrastructure expands compute options beyond graphics processing units, offering customers a broader selection for demanding artificial intelligence tasks. This move addresses a growing need for versatile processing power, particularly for agentic AI applications where both CPU and accelerated compute are critical. AWS and NVIDIA are collaborating to deliver Vera CPUs alongside their established GPU offerings, providing a heterogeneous computing environment designed to optimize performance and efficiency, the company says.

NVIDIA and Amazon’s Annapurna Labs are extending support for NVIDIA NVLink Fusion to incorporate custom NVIDIA high-bandwidth memory, a technology intended to further accelerate AI workloads. This partnership with memory suppliers allows Annapurna Labs’ Trainium chips to access faster, more power-efficient memory, scaling performance within a common rack-scale architecture. The combination of NVLink Fusion and the new custom memory aims to seamlessly integrate Trainium and GPUs, enhancing the overall efficiency of AI deployments.

Beyond commercial applications, this expanded infrastructure will also power AI factories for the U.S. Government agencies need secure AI infrastructure to keep pace with the demands of national security, the companies stated. This commitment to security is further reinforced by integrating the NVIDIA platform with the AWS Nitro System and Elastic Fabric Adapter for enhanced reliability.

NVIDIA-AWS Collaboration Powers Secure Federal AI Workloads

This collaboration is specifically geared toward bolstering secure AI infrastructure for the U.S. This commitment enables government agencies to deploy AI at scale for workloads classified at Impact Level 6 and above, addressing critical needs for national security applications. All NVIDIA GPU-based and Trainium-based EC2 instances, including those utilizing NVLink Fusion, will benefit from these security features.

NVIDIA and AWS have built one of the great growth engines of the AI era, and demand is running ahead of every forecast.

Jensen Huang, founder and CEO of NVIDIA
More like this
Artificial IntelligenceHUMAIN Expands NVIDIA Partnership for AI Infrastructure
Artificial IntelligenceMeta Deploys Tens of Millions of AWS Graviton AI Cores
High Performance ComputingNVIDIA and AWS Expand AI Compute Partnership
Artificial IntelligenceFortanix Joins HPE’s Unleash AI Partner Program
Stay current

See today’s quantum computing news on Quantum Zeitgeist for the latest breakthroughs in qubits, hardware, algorithms, and industry deals.

Tags:
Agentic AI AI infrastructure AWS GPUs NVIDIA
Avatar of Ivy Delaney

Ivy Delaney

Ivy Delaney has been working with neural networks and machine learning since the mid-nineties, back when a couple of hidden layers and a long afternoon of training counted as ambitious. She has watched the field go from academic curiosity to the thing quietly running underneath everything, and she brings that long view to quantum computing. For Quantum Zeitgeist she covers the ground where the two fields meet. That means quantum machine learning and the variational algorithms it leans on, and it also means the less glamorous but more interesting story of classical machine learning already doing real work inside quantum machines, decoding error-correcting codes, calibrating noisy hardware and learning the error models that simulators depend on. She writes about the hardware those algorithms have to run on too, and about the post-quantum cryptography scramble that the same hardware has set off. Her stories typically start with the paper, whether that is peer-reviewed work, conference proceedings or an arXiv preprint, with the source linked so you can hold a claim up against the research it came from. She is unimpressed by benchmarks that will not say what they beat, and by demonstrations that only work in the press release.

Latest Posts by Ivy Delaney: