Chip maker AMD, which seemed to lag far behind the market’s new darling Nvidia, has hit back hard by doing exactly what it takes to make an impressive comeback that actually gets noticed. By the way. Introduced a flashy new product. Sachip. And if all that is being said about it comes true, the company will continue to attract attention from its biggest rivals. According to CEO Lisa Su, AMD is effectively starting a battle over the superchip in which she expects a “winner” to emerge. She added that AI is “a defining megatrend in technology.”
MI300X chip and GPU power
AMD is touting the new MI300X as a wonder chip for large language models. The chip features one of the most advanced GPUs (graphics processing units) for artificial intelligence-level computing. GPUs are “very hard to make,” Su says. His GPUs are needed to build advanced AI programs, and companies like Sam Altman’s OpenAI use these chips to power their products. The chip will help data centers cope with his AI-related traffic surge. “There is a huge amount of technology here, this is the most complex chip we have ever built with 153 billion transistors, and the whole purpose of it is to make AI more It’s about making it much more accessible,” he said.
As evidenced by Su’s next comment, this puts AMD in the sweet spot. “Anyone who wants to use AI needs more GPUs,” she said. “And we have incredibly powerful, very efficient GPUs. We believe that will be a big winner in the AI market.”
When asked about pricing, Su laughed and said, “It’s not about the price, it’s about how many users you enable.” The price of the chip itself remains unknown. The chip is expected to go on sale later this year.
AI technology is nothing new for AMD. Su said AMD has been working on AI for years, and going forward, he says AI will shape the company’s future. She spoke at Yahoo Finance Live.
At the MI300X launch, Su said, “We are focused on accelerating the large-scale AMD AI platform deployment in the data center, with the upcoming launch of the Instinct MI300 accelerator later this year. ‘ said.
This confidence stems from the fact that the Instinct MI300X accelerator is touted as the world’s most advanced accelerator for generative AI.
“The MI300X is based on the next-generation AMD CDNA 3 accelerator architecture and supports up to 192GB of HBM3 memory, providing the computational and memory efficiency needed for large-scale language model training and inference for generative AI workloads. ,” the company said.
NVIDIA currently dominates the AI chip market. That success is reflected in the fact that the stock has soared, up 163% this year, according to CNBC. The Motley Fool’s U.S. headquarters reported that the company’s stock price fell despite the announcement of AMD chips and the rise in NVIDIA’s stock price. But Bloomberg puts things in perspective before looking at the negatives. Investors say AMD stock is up 99% this year.
Investor confidence is important, but equally important is the mood of AMD management, and as evidenced by CEO Su’s remarks, it’s definitely on the rise.
AMD Chips and AI Needs
The AMD Instinct MI300 series can accelerate processing to the potential required for generative AI and beyond. Companies that rely on large language models and need massive amounts of data processing power to answer questions or generate images need the power of such superchips. For now, data centers can handle that surge, but they will need better chips. Now, Nvidia is taking the lead and already meeting the demands of these massive workloads.
But the race was never lost. “We are still in the very early stages of the AI lifecycle,” Su said. The market for data center AI accelerators will grow fivefold to more than $150 billion, he added.
How important is the AI market to AMD and Nvidia?
At last week’s earnings call, Su said the area was the company’s “top strategic priority,” highlighting it by mentioning the term “AI” more than 50 times.
There was a similar sense of urgency when Nvidia announced its earnings nearly three months ago. The term “AI” was used more than 80 times in a conference call with analysts.
