US EditionPublished • loading... • Updated
Amazon to use Nvidia tech in AI chips, roll out new servers
AWS launched Trainium3 UltraServers with 144 chips offering 4x performance and 40% less energy use; Trainium4 will integrate Nvidia's NVLink Fusion for enhanced AI infrastructure.
- On Dec 2, Amazon Web Services unveiled its Trainium3 chip and new Trainium3-based servers at the AWS re:Invent conference in Las Vegas.
- Faced with Nvidia's dominant market position and rival moves, Amazon Web Services pushed into custom silicon to reduce supplier reliance and offer lower-cost AI options amid supply constraints limiting AWS capacity.
- AWS said Trainium3 UltraServers combine 144 chips with NeuronSwitch-v1 interconnect, delivering 20.7 TB of HBM3E, 706 TB/s bandwidth, and 363–1,452 petaFLOPS depending on workload.
- AWS said early tests show customers including Anthropic and Karakuri are cutting AI training and inference costs by up to 50% while using about 40% less power, and Nvidia shares pared gains after launch.
- Although Amazon teased Trainium4 will support Nvidia's NVLink Fusion, it did not specify a timeline, but expects at least three times Trainium3 performance, linking thousands of UltraServers for scale.
Insights by Ground AI
36 Articles
36 Articles
Amazon.com's AWS cloud computing unit announced on Tuesday that it will adopt Nvidia's technology in future generations of its artificial intelligence chips, as the company intensifys its efforts to get major AI customers to use its services. AWS, or Amazon Web Services, said it will adopt a technology called "NVLink Fusion" on a future chip, with no specified launch date, known as Trainium4. NVLink technology creates fast connections between di…
Amazon to use Nvidia tech in AI chips, roll out new servers
Amazon.com's AWS cloud computing unit on Tuesday said it will adopt key Nvidia technology in future generations of its artificial intelligence computing chips as the firm ramps up efforts to attract major AI customers to use its services.
·United Kingdom
Read Full Article
Coverage Details
Total News Sources36
Leaning Left3Leaning Right3Center14Last UpdatedBias Distribution70% Center
Bias Distribution
- 70% of the sources are Center
70% Center
15%
C 70%
15%
Factuality
To view factuality data please Upgrade to Premium
