Levels to manage $AMD while waiting for its MI350 and MI390Copy & Paste from Morgan Stanley's note:
Can MI350 be the Rome of AI? Yes, it's possible. The company's scramble to add AI engineers through two transactions is actually a good sign. While bears see it as an indication of a software ecosystem that is behind the competition, we thought that anyway, and the company making an aggressive push to turn that around quickly rather than accepting an immature position and fixing it a little bit at a time is certainly a positive. AMD executes very well, having never meaningfully missing a product launch target in the Lisa Su era, which we find exceptional. And unlike the initial server ramp, Instinct will benefit from being directly tied to the server CPUs that are becoming dominant in cloud- which gives them leverage to at least build a ubiquitous presence across multiple clouds.
AMD opened the product announcements with the launch of their 5th gen EPYC server processor (Turin), which should build on their leading position with cloud customers. Gen 5 is platform compatible with AMD's zen 4 lineup and will feature two primary product families, one more enterprise focused "scale up" set of 4nm skews and cloud focused "scale out" 3nm versions where the high end 192 core product offers 2.7x performance vs 64 core Emerald Rapids from Intel (launched late last year) on an industry standard integer score benchmark. With AMD highlighting leading performance vs Intel in other areas including ML and AI workloads. Both families of zen 5 utilize a 6mn IO die. Representatives from google cloud, and Oracle Cloud were on stage with positive endorsements for the new products.
