AMD's Inference Chip Pivot and OpenAI Partnership Challenge Nvidia

Key Quotes

"Nvidia controls 90 or more of the advanced AI chip market. It's not that Nvidia has chased out the competition or has some sort of nefarious strategy to make it impossible for people to compete with them. It's that they were very early first movers in this idea that these chips that used to be primarily used for video games were also really, really good for doing AI computing."

This quote highlights Nvidia's dominant position in the AI chip market, attributing it to their early recognition of the potential of gaming chips for AI tasks. The speaker, Robbie Whelan, explains that Nvidia's success is a result of being a first mover rather than anticompetitive practices.

"We're in a very special time. So we are probably going faster than we've ever gone before. I mean, I certainly believe that the technology is moving faster than I've ever seen in my career, and, you know, the role of AMD is to enable all of that with the foundational computing."

Lisa Su, CEO of AMD, describes the current era as a period of unprecedented technological acceleration. She emphasizes that AMD's purpose is to provide the fundamental computing power necessary to support this rapid advancement in technology.

"She said, 'Artificial intelligence is rising. It's a once-in-a-lifetime opportunity, and we are positioned in a very special way for us to take advantage of that.' She said, 'We're going to revamp our entire product line so that it's now all oriented around artificial intelligence,' and this was a major turning point."

This quote details Lisa Su's strategic pivot of AMD towards artificial intelligence. The speaker explains that Su identified AI as a monumental opportunity and directed the company to reorient its entire product development around this emerging field, marking a significant shift for AMD.

"So in AI computing, there are two main functions that people need to utilize. When you're developing an AI model or an AI tool, whatever it is, be it a chatbot or a video generation app, you have to first train it, and then you have to make it capable of responding to queries, which is to say you have to run it. And the running it is usually referred to as inference."

The speaker defines the two primary functions in AI computing: training and inference. Training involves developing the AI model, while inference is the process of running the model to respond to user queries, such as those made to a chatbot.

"We structured this deal so that there was, you know, complete alignment of incentives. Complete alignment that we wanted to go fast, that we wanted to go big, and that if OpenAI is successful, AMD is successful because, you know, they will need lots of GPUs. And the opposite is also true, which is if AMD is successful, OpenAI gets to share in some of that upside."

Lisa Su explains the rationale behind AMD's deal with OpenAI, emphasizing the alignment of incentives. She states that the partnership is designed so that the success of one company directly benefits the other, particularly concerning the demand for AMD's GPUs.

"I'm not worried about an AI bubble. And she repeated this mantra she has, which is basically, there's insatiable demand for computing, and that's why I'm so confident that even if it takes unusual, unorthodox financing at the outset, things are going to turn out okay and that we're going to benefit in a big way from all of this demand."

Lisa Su expresses her confidence in the AI market, stating she is not concerned about a bubble. She attributes this to what she calls "insatiable demand for computing," believing that this fundamental need will ensure the long-term success of companies in the AI sector, even with unconventional financing structures.