Nvidia and Amazon Web Services join forces for smarter AI chips
AWS is introducing Nvidia's H200 AI GPUs, which feature almost double the speed of their predecessor, the H100.
A few minutes every morning is all you need.
Stay up to date on the world's Headlines and Human Stories. It's fun, it's factual, it's fluff-free.
The backstory: Last year, OpenAI's ChatGPT shook up the tech world, bringing artificial intelligence (AI) into the spotlight. Nvidia, known for its graphics processing units (GPU) since 1993, initially made a mark in gaming GPUs in 1999. But as the demand for deep-learning AI models soared, Nvidia's GPUs became a hot commodity, pushing the company's market value to over US$1 trillion alongside other elite tech giants. Notably, Nvidia's GPUs power the "brain" behind ChatGPT. This sparked a shortage as tech giants raced to incorporate similar AI tech into their products.
Amazon Web Services (AWS), a key player in cloud services, also felt the impact of this AI craze. As Amazon's cloud computing branch, AWS offers a range of cloud-based solutions, empowering businesses to innovate and grow without massive infrastructure investments.
More recently: Earlier this month, Nvidia rolled out the H200, a beefed-up GPU designed specifically for training and deploying AI models. A key upgrade is its 141GB "HBM3e" memory, geared towards boosting performance. Shortly after that, Microsoft introduced the Maia 100 as its inaugural AI chip. The tech giant also announced that it's bringing in Nvidia H200 GPUs to the Azure cloud.
The development: AWS is introducing Nvidia's H200 AI GPUs, which feature almost double the speed of their predecessor, the H100. Alongside this, AWS is launching its Trainium2 AI chip and Graviton4 processor, both designed to simplify AI application development and operation. The Trainium2 promises a fourfold efficiency boost, while the Graviton4, running on Arm architecture, offers a 30% performance increase than its predecessor.
In a move that signals a deepened collaboration, AWS is set to operate 16,000 Nvidia GH200 Grace Hopper Superchips. While the release dates for virtual-machine instances featuring the Nvidia H200 chips are still pending, users can start testing Graviton4 virtual-machine instances, which are expected to hit the market in the next few months.
Key comments:
"We view Nvidia as the most important company on the planet in an era that is rapidly changing towards one that will be emphasized by greater AI capabilities," said CFRA Research analyst Angelo Zino.
“The integration of faster and more extensive HBM memory serves to accelerate performance across computationally demanding tasks including generative AI models and [high-performance computing] applications while optimizing GPU utilization and efficiency,” said Ian Buck, Nvidia VP of high-performance computing products, referring to the H200 AI GPUs.
"With all the enthusiasm around AI and the fact Nvidia delivered a huge beat for first-quarter results and second-quarter estimates, this gives some actual evidence AI is for real," said Daniel Morgan, senior portfolio manager at Synovus Trust in Atlanta, Georgia, in the US.
Comments ()