
"(Bloomberg/Matt Day) Amazon.com Inc.'s cloud unit raced to get the latest version of its artificial intelligence chip to market, renewing efforts to sell hardware capable of rivaling products from Nvidia Corp. and Google. The accelerator, called Trainium3, was recently installed in a few data centers and will be available for customers beginning on Tuesday, Dave Brown, a vice president with Amazon Web Services, said in an interview."
"As we get into early next year we'll start to scale out very, very quickly, he said. The chip push is a key element of Amazon's strategy to stand out in AI. AWS is the largest seller of rented computing power and data storage. But it has struggled to replicate that dominance among leading developers of AI tools, as some companies opt instead to rely on Microsoft Corp., which has close ties to ChatGPT maker OpenAI, or Alphabet Inc.'s Google."
"Amazon shares rose 1.6% to $237.71 at 11:09 a.m. in New York. Nvidia shares pared gains, while AI chip rival Advanced Micro Devices Inc. dropped to a session low. Amazon hopes to entice companies looking for a bargain. Trainium chips are capable of powering the intensive calculations behind AI models more cheaply and efficiently than Nvidia's market-leading graphics processing units, according to the company. We've been very pleased with our ability to get the right price performance with Trainium, Brown said."
AWS accelerated the rollout of Trainium3, installing units in several data centers and opening customer access with rapid scaling expected early next year. Trainium3 aims to rival Nvidia and Google by offering cheaper, more efficient hardware for intensive AI model training. AWS remains the largest provider of rented compute and storage but has struggled to win leading AI-tool developers who often choose Microsoft or Google. Amazon highlights Trainium's price-performance advantage and moved from the prior accelerator to Trainium3 in about a year, matching the fast cadence of competitors. AWS hardware still lacks Nvidia's deep software libraries.
Read at www.mercurynews.com
Unable to calculate read time
Collection
[
|
...
]