Why Wall Street wasn't excited by AMD's new AI chip

AMD unveils its new AI chips at its Advancing AI 2024 event. AMD highlighting capabilities of its products that the company says will exceed some of its rivals. The flurry of announcements coming as AMD continues to battle AI market leader NVIDIA. More insights and strategy founder and CEO Patrick Moorhead joining us now to discuss. Patrick, always great to see you. Let's get right to those headlines. AMD says, Patrick, the new AI chips, they're coming soon. They can outperform NVIDIA. What do you make of the news, Patrick, and why is the stock here down about 5%? Yeah, let me address the stock piece first. So this is classic sell on the news behavior. We've seen this with other gigantic announcements out there in the AI space, but fundamentally, investors were also looking for a new customer for its data center GPUs, maybe AWS, maybe Google. The company did come out with some pretty exceptional announcements, and whether that's data center GPU, data center CPU, and something that nobody's talking about is they entered a brand new market, and that was a new networking chip for back end networks to connect all of those GPUs together. Patrick, all of that sounds good, right? It sounds positive, but the stock is down almost 5% at this point. Were people expecting something more, something different, something better than what they heard? They were expecting new customers on the GPU side. I mean, they already have 50 to 60% market share in CPUs with hyperscalers. They are selling to everybody, but that was the expectation, and they did not roll out new customers. And some of this could be a timing. I mean, AWS is somewhat sitting on the sidelines with data center GPU. They have their flagship event coming in December that I'll attend, and they do all of their big announcements here. And even on the CPU side for PCs, they have no new customers they can get because they have all new customers. It is interesting with some of the disclosures, particularly with Meta, where Meta said they have 1.5 million Epic data center CPUs installed, and 100% of LLAMA405B live inference is going through their current instinct. It could be reacting potentially to the timing of this new chip called the MI355, which comes out in the second half of 25. Potentially they were looking for something that came in closer, potentially in the middle of the year. Patrick, let me ask you, because I know you're going to be speaking with AMD CEO, Lisa Su shortly. What are you going to ask her, Patrick? Yeah, a couple of my top questions are, is how are you going to narrow the gap with Nvidia? Both companies say that they have the highest performance, and they all show benchmarks. AMD has maybe 5%, 10% market share, and Nvidia has 95% to 90% benchmark. But what is it going to take to significantly narrow the gap? And then when it comes to data center CPUs, while AMD is super strong in hyperscalers, they have single digit share in enterprise, where quite frankly, most of the server CPU businesses today, which is dominated by Intel. And there are a lot of things that AMD could be doing in there. Typically, it's not best product wins, it's sales, enterprise marketing. And AMD up to this point has not made the major investments in that space to be the difference maker. Patrick, it's so interesting, because Lisa Su, I believe today is exactly her 10th anniversary as CEO of the company. She has been pretty well lauded for her leadership there. But it feels like, to your point, that this latest period has been a challenging one with the really growing dominance of GPUs and the data center era, I guess we could call it. Does that risk sort of tarnishing her legacy at this company? Yeah, I look at back where AMD was at when they started. I mean, their stock was in the single digits, and it went up into the stratosphere. And I think that's the big picture. The company flat out did not have enough resources. I mean, years ago, they sold off their last tangible asset. They had three to six months of cash left, and they had to start in the desktop CPU market. And then they moved into notebooks, and then they moved back into the data center. The company flat out did not have enough money to invest into AI data center GPUs. At the moment, when they did have the cash, they invested, and just being on the playing field, I think, is somewhat of a miracle, particularly when you look at companies like Intel that very specifically data center GPUs are simply not even on the map yet. Intel has data center accelerators, but not data center AI GPUs.