AWS Plans 2 Million More Nvidia GPUs Through 2028
Together with an earlier commitment, AWS plans to add more than 3 million Nvidia GPUs, including 100,000 for secure US government workloads.
Topics
News
- Fired OpenAI Researchers Warn Dismissals Could Chill Safety Work
- ARTEX Developer Closes AI Hacking Tool After Bank Attacks
- AI Redraws Global Race for Quant Talent as Coding Tools Lower Entry Barriers
- AI Dispatch | India Draws Billions Into AI
- AI Bots Now Make Up Half of Traffic to US Fed’s FRED Database
- Musk Blames ‘Oligarchs’ For Delay in Starlink’s India launch
Amazon Web Services (AWS) plans to deploy another 2 million Nvidia GPUs across its global infrastructure in 2027 and 2028, taking its planned additions since 2026 above 3 million.
The new capacity will include Nvidia’s Blackwell Ultra, Rubin and Rubin Ultra processors, the companies said on Wednesday, 26 August.
Financial terms were not disclosed.
AWS had already committed to adding more than 1 million Nvidia GPUs beginning in 2026. It said demand since that announcement had exceeded expectations.
The expanded order does not displace Trainium, the Nvidia alternative developed by Amazon’s Annapurna Labs. AWS is instead bringing the two platforms closer together.
AWS announced in 2025 that future Trainium chips would support Nvidia’s NVLink Fusion interconnect technology.
Annapurna will now also work with Nvidia on custom high-bandwidth memory, which could give Trainium faster, more energy-efficient memory and allow Trainium chips and Nvidia GPUs to operate within the same rack-scale systems.
Nvidia’s Vera CPUs are also coming to AWS. The processors are designed for CPU-heavy work surrounding AI agents, including code execution, tool use, sandboxing, data pipelines and orchestration.
“Customers want the freedom to choose the best tools for their AI workloads,” AWS Chief Executive Matt Garman said in a statement.
The companies also plan to build secure AI infrastructure for US federal and national-security agencies, including 100,000 Nvidia GPUs. It will support workloads at Impact Level 6 and higher. IL6 covers Defense Department data classified up to the Secret level.
AWS will use Nvidia’s Spectrum networking technology to connect large GPU clusters. Amazon Robotics will also adopt Nvidia’s Jetson, Omniverse and Isaac technologies for simulation, robot training and warehouse automation.
The announcement came as Nvidia forecast revenue growth of about 70% in its next fiscal year and warned that supply constraints were limiting expansion.
Nvidia Chief Executive Jensen Huang said demand was “running ahead of every forecast.”
