So they're already well-entrenched with all the major hyperscalers right now. JOHN VINH: The data center advantage for AMD is there's a number of factors there. And in terms of the data center advantage here, how does that play into sort of the AI chip space? And if you look at AMD, they don't have to gain much market share there for it to really meaningfully move the needle for them. It implies kind of a North of a 50% plus carrier, and it's a big number. But $150 billion seems like a reasonable assumption to us. Obviously, a lot has changed in the industry over the last six months. JOHN VINH: I think it's a pretty reasonable projection. What do you make of the projections here? Now, it now has projections of $150 billion plus for 2027 versus $64 billion that it has for 2025 and just $30 billion for 2023. And we know that AMD also talked about its AI accelerator outlook, its TAM or its total addressable market. But there is such a significant opportunity in generative AI across both inferencing and training that there's more opportunities for both vendors in this space. On the training front, I think Nvidia still has a much more better performance chip, but I do think that AMD is going to be competitive on the training front. I think they've got a power efficiency advantage there. JOHN VINH: I think AMD's MI300X is going to be much more competitive on the inferencing side. So now that we have the MI300X out from AMD, from what you've heard, especially from AMD's projections at the event, how do their offerings actually stack up, though? And of course, everyone's really been looking for what would be the answer to Nvidia's H100 GPU. And as a result of that, they're able to basically dictate whatever pricing they want.Īnd I think the market is always looking for a second source alternative just so that there's a little bit more balanced pricing and economics in the market. They probably have close to 95%-plus market share. Right now, Nvidia's market share in AI and GPU data center is just incredibly dominant. JOHN VINH: I think right now, the market, we've seen this historically multiple times is just looking for kind of a second source alternative. So then, what would be the selling point for AMD versus, say, picking Nvidia? Especially for AWS. So it would be no surprise if they along with the other major hyperscalers would look to adopt MI 300 going forward for generative AI. AWS is already a strong partner for AMD to begin with. But when you do have a flagship customer potentially in the works like Amazon, how much of a game changer? So now obviously, this is something that AWS said they were considering according to Reuters. Let's bring in John Vinh, KeyBanc Capital Markets Equity Research Analyst to discuss this. It comes as Reuters reports that Amazon's AWS could be using AMD's new AI chip. We're watching shares of AMD, the chip maker, popping today after unveiling its new AI chip with an accelerator that can speed processing for technology like ChatGPT.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |