Amazon’s strategic pause for Nvidia’s next-gen superchip

Amazon Web Services (AWS), a leading player in the cloud computing arena, has strategically decided to postpone its order for Nvidia’s latest “superchip”. Instead, it is waiting for a more advanced model. This decision has caused a stir among investors, raising concerns about a potential dip in demand for AI chips as Nvidia transitions between product cycles.

Calculated delay

Anticipating the Future: AWS has chosen to bypass Nvidia’s Grace Hopper superchip, which was launched in August, in favour of its successor, Grace Blackwell. The decision is based on the short interval between the two models, making it more logical for AWS to wait for the upgraded version.

Investor Concerns: The introduction of Nvidia’s Blackwell processors, which are claimed to be twice as powerful as their predecessors for training large language models like OpenAI’s ChatGPT, has resulted in a tripling of sales in the recent quarter. However, investors are apprehensive about a potential slowdown in growth as the tech giant nears the release of Blackwell.

The Cost of Innovation: The GB200 chip, which includes two B100 chips, is estimated to cost around $70,000, with a complete server setup costing up to $3 million. Despite the steep costs, AWS’s decision reflects the importance of staying at the cutting edge of AI technology.

Market Dynamics: The demand for Nvidia’s H100 chips has significantly exceeded supply due to the surge in AI infrastructure investment sparked by ChatGPT. With the easing of delivery wait times for H100 chips since the start of 2024, the market is closely watching how Nvidia manages the supply of its last-generation chips while new customers continue to seek out the available Hopper supply.

AWS’s deliberate delay in ordering Nvidia’s superchip highlights the rapid pace of innovation in the AI sector and the critical importance of strategic planning in technology adoption. As Nvidia prepares to launch its Blackwell processors, the tech industry is eagerly awaiting the next leap forward in AI capabilities.