AWS Integrates Nvidia AI Tech, Launches New Trainium3 Servers and Nova Models
- tech360.tv
- 2 hours ago
- 2 min read
AWS confirmed it will adopt key Nvidia technology in future generations of its artificial intelligence computing chips as part of efforts to attract major AI customers. The cloud computing unit also announced the rollout of new servers and AI models.
AWS will incorporate "NVLink Fusion" technology into a future chip called Trainium4. NVLink technology is known for creating speedy connections between different kinds of chips.
The announcement occurred during AWS's annual week-long cloud computing conference in Las Vegas, which draws approximately 60,000 people. Nvidia has been actively encouraging chip firms to adopt its NVLink technology, with Intel, Qualcomm, and now AWS committing.
This technology will assist AWS in building larger, faster AI servers that can recognise and communicate with one another more rapidly. This speed is a critical factor for training large AI models, which require thousands of machines to be interconnected.
As part of the Nvidia partnership, customers will gain access to "AI Factories," exclusive AI infrastructure located inside their own data centres for enhanced speed and readiness. Nvidia Chief Executive Officer Jensen Huang stated, "Together, Nvidia and AWS are creating the compute fabric for the AI industrial revolution - bringing advanced AI to every company, in every country, and accelerating the world's path to intelligence."
Separately, Amazon introduced new servers based on its Trainium3 chip. These servers, now available, each contain 144 chips and offer more than four times the computing power of AWS's previous AI generation while using 40% less power.
Dave Brown, vice president of AWS compute and machine learning services, indicated that AWS aims to compete on price with rivals, including Nvidia. Brown explained that the goal is to provide products offering needed performance at a suitable price point, delivering price-performance benefits to customers.
The company also unveiled new versions of its AI models, known as Nova. Nova 2 is described as faster and more responsive. The Nova family also includes a version capable of responding with images or text-to-text, image, speech, or video prompts.
Another model, Sonic, can respond to speech prompts with speech outputs. AWS Chief Executive Officer Matt Garman characterised Sonic's responses as "human-like" during his keynote address.
Amazon has encountered challenges in achieving widespread adoption of Nova against competitors such as OpenAI's ChatGPT, Anthropic's Claude, and Google's Gemini. Despite this, Amazon posted a 20% sales increase for AWS in its most recent quarter, primarily driven by its cloud computing and AI infrastructure services.
At the Las Vegas conference, Amazon also introduced Nova Forge, a service designed to help companies create their own AI models using their specific data. Garman noted that this service allows for the production of models that deeply understand user information without losing core trained data.
Amazon shares were up 0.9% at USD 235.98 at midday on Tuesday.
AWS will adopt Nvidia's "NVLink Fusion" technology for its future Trainium4 AI chips.
New Trainium3-based servers, featuring 144 chips and over four times the computing power, are now available.
Amazon unveiled new "Nova" AI models, including the faster Nova 2 and the speech-to-speech Sonic.
Source: REUTERS