November 26, 2024 – Amazon has embarked on a moonshot initiative to deploy 100,000 second-generation custom chips, aiming to enhance data processing efficiency and trim down AI chip procurement costs, Bloomberg reported on November 25th.
Citing informed sources, the news outlet revealed that Amazon has established an engineering lab in Austin, Texas, USA, and assembled a dedicated team of engineers tasked with developing AI chips.
The lab environment is likened to that of a startup, where engineers are focused on refining circuit boards and cooling systems to optimize chips for future AI workloads.
Led by Amazon’s core chip design engineer, Rami Sinno, the team is expediting the development of its second-generation proprietary AI chip, Trainium2, with plans to complete testing and delivery by the end of this year.
Amazon claims that Trainium2 offers a fourfold increase in performance and a threefold boost in memory capacity compared to its predecessor, along with significant advantages in energy efficiency and cost.
Despite Amazon’s ambitious undertaking, analysts believe that NVIDIA’s market leadership position remains unthreatened in the near term. Mizuho Securities noted that similar news reports are not uncommon and do not necessarily represent a direct threat to NVIDIA.
NVIDIA continues to be the preferred supplier for global training and inference tasks, commanding a significant share of the AI GPU production market. While Amazon’s efforts are noteworthy, the impact on NVIDIA might have been overestimated.
