Key Points
- Amazon’s AWS will use Nvidia’s high-speed NVLink technology in its future AI chips.
- The partnership aims to build more powerful servers for training large AI models.
- AWS is also rolling out new servers now with its current Trainium3 chip, competing on price and performance.
- Amazon updated its “Nova” AI models to better compete with rivals like ChatGPT and Google’s Gemini.
Amazon’s cloud computing division, AWS, is teaming up with chip giant Nvidia to step up its push into artificial intelligence. In a major announcement on Tuesday, AWS said it will use Nvidia’s key “NVLink Fusion” technology in its next generation of custom-built AI chips.
This move is a big deal because NVLink is one of Nvidia’s crown jewels, enabling different chips to communicate at incredible speeds. By adopting this tech for its future Trainium4 chip, Amazon can build much larger and more powerful AI servers. This is critical for training the massive AI models that are reshaping the tech world. The partnership aims to create “AI Factories” that deliver super-fast, ready-to-use AI infrastructure to customers.
“Together, Nvidia and AWS are creating the compute fabric for the AI industrial revolution,” said Nvidia CEO Jensen Huang.
While the Nvidia partnership looks to the future, Amazon also made moves for the present. The company announced it is now rolling out new servers powered by its current AI chip, Trainium3.
AWS claims these new machines have four times the computing power of the previous generation while using 40% less energy. An AWS executive said the goal is to compete directly with rivals, including Nvidia, by offering better performance at a lower price.
On the software side, Amazon introduced new versions of its AI model, “Nova,” which is trying to catch up to competitors like OpenAI’s ChatGPT. The new models are faster and can handle images, video, and more “human-like” speech. The company also launched a service to help businesses build their own custom AI models.
