Amazon and Weidar announced an expansion of cooperation. According to the latest arrangements, AWS will add 2 million GPUs to the data centres in 2027 and 2028, covering product lines like Blackwell Ultra, Rubin and Rubin Ultra. This decision is only five months from the last large-scale procurement commitment in the Amazon.

I'll add it in five months.

The news was revealed at a telephone conference in the In Weida newspaper. According to Ingweida, the scale of deployment previously planned for the Amazon was no longer sufficient and the parties decided to continue the upwards. Clients driving demand growth include start-ups, large enterprises, AI laboratories and government agencies.

This collaboration is not just about increasing the number of GPUs. Weeda indicated that its network hardware, open models, CPU, data processing software and robotic platforms would also have wider access to the AWS system.

  • Add 2 million GPUs
  • Delivery time concentrated in 2027 and 2028
  • Amazon has committed more than 1 million GPUs.

The AWS self-research chip continues to advance.

The Amazon has been accelerating its self-study chip layout in recent years, hoping to reduce its dependence on Weeda and to compete in some areas. Peter DeSantis, director of AWS AI, had previously indicated that AWS was negotiating with other companies to sell Trainium chips for data centre in-depth learning missions.

In addition to the AI accelerator, the Graviton CPU based on the Arm structure of the Amazon continues to expand, targeting the traditional server chip market. The Amazon previously disclosed that its customized chip operations had reached an annualized caliber of $25 billion, with a total commitment of $225.0 billion from the Anthropic, OpenAI and other AI laboratories.

However, on the basis of the latest scale of cooperation, Ying Weidar is still the core supplier of AWS-enabled AI. In addition to GPU, the company will provide AWS with a certain number of Vera CPUs, some of which will be deployed in conjunction with Rubin.

Cooperation extends to models and machines People

The scope of cooperation is also being extended to the automation of business services and storage. AWS will provide the British Nemotron Open Model on Amazon Bedrock and SageMaker, and export hosting AI services to business customers.

In terms of robotics, the Amazon plans to use the British philsical AI technology warehouse for its storage robotics. Related products include Omniverse, Cosmos, Isaac and Jetson, covering simulations, world models, robotics development and margin calculations.

Imwidda Synchron plus capacity.

On the same day, Ying Weidar announced that the second financial season had received $96.2 billion, which was higher than market expectations, including $89 billion in data centre operations, an increase of 117 per cent over the same period. The company expects a third-season harvest of $108 billion, partly from a new generation of Rubin GPU.

In order to meet AI infrastructure needs in the coming years, British Wida has committed $279 billion to locking up supply chains and manufacturing capacity, significantly higher than the previous quarter of $119 billion. The market then focused on whether large-scale computing inputs could continue to translate into real profits for cloud services and AI products.