Interessanter Artikel mit Aussagen von Lisa Su: seekingalpha.com/article/...t-loud?source=feed_tag_long_ideas
AI Is Not A Bubble, It’s A 10-Year Super Cycle
Summary:
- Advanced Micro Devices, Inc. is structurally positioned to capture double-digit share of a $1 trillion data center TAM by 2030.
- AMD's deliberate pivot to high-performance computing and AI, anchored by chiplet-first architecture and rack-scale systems, is driving accelerated data center growth.
- GPUs remain the dominant accelerator, with ASICs as a complement, supporting AMD's opportunity in general-purpose, high-performance markets.
- I maintain a path to a $775 per AMD share base case and see a credible bull scenario near $1,000 if AMD executes on its data center ambitions.
Aussagen von Lisa:
"We’ve been incredibly focused on high-performance computing… and then a few years ago, it became absolutely clear that this was going to be all about AI… So yes, we’ve pivoted our R&D capabilities… to a significant focus on high-performance computing and AI."
"You really have to have a holistic view… CPUs, GPUs, FPGAs, the possibility of doing ASICs, it’s being able to integrate all of that together. And that’s our unique capability… we’re the only semiconductor company out there that has all of this foundational IP."
"Google has done a great job with the TPU architecture… But it is a more purpose-built architecture. It’s not built with the same programmability… the same capabilities to do training and inference where GPUs are."
"Our view is… over the next 5 years or so, we’ll see GPUs still be the significant majority of the market because we are still so early in the cycle and because software developers actually want the flexibility to innovate on different algorithms."
"Best-in-class system design… combined with our best-in-class hardware and software capability… We’re super excited about MI450 and the Helios product that will come to market in 2026."
"There’s not a concept of a bubble. What there is a concept of is, we are, let’s call it, 2 years into a 10-year super cycle."
"The one thing that is constant as we talk to customers is we need more compute… if there was more compute installed… we would get to the answer faster."
"What started as ‘let’s try AI for our internal use cases’ has now turned into significant clear productivity wins… there is no question that there is a return on investment for investment in AI."
"Over the last 3 months, what we’ve seen is really a significant uptick in CPU demand… as AI moves to more inferencing and there’s more work being done in things like agent workloads… they’re spawning more general-purpose CPU needs."
Quelle: Lisa Su, UBS Conference.
Her framing is simple but powerful.
Basically, if you have “1,000 agents or 1,000 virtual employees,” they need somewhere to run, operate on data, and do work, and that’s general-purpose CPU territory.
AI isn’t just a GPU story for AMD. It touches every part of AMD’s portfolio, accelerators, CPUs, and even higher-end PCs.
The best is yet to come
BT