Bloomberg published a blog post yesterday, November 25, reporting thatAmazonTo minimize the impact onNvidiawhich has initiated a "moonshot" program.Hoping to deploy 100,000 second-generation self-researchchipThis is the first time that the AI chip has been purchased in the U.S., improving data processing efficiency and reducing the cost of AI chip purchases.
Citing sources, 1AI reported that Amazon has set up an engineering lab in Austin, Texas, and has assembled a dedicated team of engineers to work on AI chips.
Sources say the lab atmosphere is similar to a startup, with engineers focusing on board and cooling system improvements to optimize the chip for future AI workloads.
The team, led by Rami Sinno, a core chip design engineer at Amazon, is accelerating the development of its second-generation in-house AI chip, Trainium2, and plans to complete testing and deliveries by the end of this year.
Amazon says Trainium2 offers four times the performance and three times the memory capacity over its predecessor, with significant energy efficiency and cost advantages.
Despite Amazon's ambitions, analysts don't see a threat to NVIDIA's market leadership in the near future. Mizuho Securities noted that similar news reports are commonplace and do not represent a direct threat to NVIDIA.
NVIDIA remains the world's preferred supplier for training and inference tasks, capturing the majority of the market share for GPU production for AI. While Amazon's efforts are noteworthy, their impact on NVIDIA may be overestimated.