highlight:
Projected to have 2$ EPS now to 20$ EPS in 3 to 5 years.
A very big pie. If that's realized, it seems AMD stock price could be 2x-3x then (with PE = 24 to 40, since current PE = 120). That could be still pretty expensive if growth slows down by then. Almost all stocks seems expensive now... So better to accumulate it with lower price if possible to increase margin of safety.
Example questions for disucssion:
1. How do folks think about Google's TPU competition (seems start to be available for use outside of GCP)? Will that be a threat to AMD's market share expansion ambition?
2. The three key strategic customers listed (OpenAI, Meta, Oracle) are all pretty risky or have some concerns about revenue/profit, CapEx, or debt. How do folks think about the risk of relying those risky customers?
------------
(Summarized by Gemini below)
This video is a recording of AMD's Financial Analyst Day from November 11, 2025. The event features AMD's entire executive leadership team, led by CEO Dr. Lisa Su, presenting the company's long-term strategy, technology roadmaps, and financial targets.
The central theme is AMD's strategic pivot to capitalize on the "insatiable demand" for AI compute [10:11, 26:33], with a new market forecast projecting a $1 trillion data center silicon TAM by 2030 [30:39].
Here is a summary of the key presentations and announcements:
Dr. Lisa Su (Chair and CEO): Strategy & Financial Goals
-
Mission: AMD's goal is to be the leader in high-performance and AI computing [10:41].
-
Data Center Focus: The data center business has grown from $2 billion in 2020 to an estimated $16 billion this year [13:38] and is now AMD's largest segment, making up 47% of the business [15:20].
-
AI Momentum: The AI business, driven by the MI300 series, has ramped faster than any product in AMD's history [14:32]. The company is moving to an annual cadence for its AI accelerators [16:06].
-
Strategic Partnerships: Su highlighted several key AI partnerships as a foundation for growth:
-
OpenAI: A 6-gigawatt, 5-year partnership, with AMD's MI450 series optimized for their inference workloads [38:16].
-
Meta: A deep partnership to build the "Helios" rack-scale solution based on open standards [39:54].
-
Oracle: Oracle will be one of the first cloud providers to offer public instances of the upcoming MI450 [39:05].
-
-
New Financial Model (3-5 Year CAGR):
Mark Papermaster (CTO): Technology Roadmap
-
Broad IP Portfolio: AMD has the broadest compute portfolio in the industry, including CPUs (Zen), data center GPUs (CDNA), consumer GPUs (RDNA), AI Engines (NPUs), and networking (Pensando DPUs) [47:33].
-
CPU Roadmap: Zen 6 is the next-generation core, taped out on TSMC's 2nm process [56:06].
-
GPU Roadmap: The company is on an annual cadence. CDNA 4 (powering MI350) is shipping now, CDNA 5 (powering the MI400 series) launches next year, and MI500 is in design [57:15, 58:07]. He highlighted an exponential 2x-per-year growth in compute performance [58:50].
-
AI Security: Papermaster detailed "confidential computing" for AI, which encrypts data and models "at rest, in motion, and in use" to protect customers' "crown jewels" [01:09:35].
Data Center (Forrest Norrod, Dan McNamara, Vamsi Boppana)
-
Dan McNamara (Server CPUs):
-
AMD is now the "de facto server CPU leader," with over 40% market share [01:20:31, 01:24:19].
-
He argued against the "CPU cannibalization" thesis, stating that AI is accelerating CPU demand for tasks like RAG, pre/post-processing, and agentic workflows [01:34:49, 01:36:12].
-
The server CPU TAM is expected to double to $30 billion by 2030 due to AI [01:37:42].
-
AMD's goal is to achieve greater than 50% server CPU revenue share [01:39:24].
-
-
Vamsi Boppana (AI GPUs):
-
ROCm (Software): AMD's open-source software stack is "battle-tested." All 2 million models on Hugging Face now run successfully on AMD GPUs [01:46:42]. He is convinced a significant amount of future GPU programming will be done by AI [01:50:15].
-
MI400 Series: The next-generation AI accelerator, featuring the MI450X, which packs 40 petaflops of compute and 432GB of HBM memory [01:51:06].
-
MI500 Series: Announced as the next "big breakthrough" in AI performance, set to follow the MI400 [01:55:57].
-
-
Forrest Norrod (Data Center Solutions):
-
Tied everything together, explaining AMD's "AI factory" strategy is built on open standards like Ultra Accelerator Link (UAL) and Ultra Ethernet [01:58:56].
-
He detailed the Helios rack-scale solution, which co-packages CPUs, GPUs, and networking (Pensando DPUs) into a complete, open-ecosystem building block for hyperscalers [02:09:35].
-
Client, Gaming, & Embedded (Jack Huynh, Salil Raje)
-
Jack Huynh (Client & Gaming):
-
AMD's client business has grown from $9.6B to over $14B in the last year [02:37:43].
-
AMD has achieved 28% revenue share in the client market, with a goal to exceed 40% [02:39:47, 02:46:17].
-
He emphasized that AI is in its "infancy" in the PC market and will be a major future growth driver [02:41:05].
-
-
Salil Raje (Embedded):
-
The Xilinx acquisition has been a major success, with design wins growing from single digits to over $16 billion in 2025 [02:56:09, 02:56:37].
-
The "next big transformative vector" for this business is Physical AI (robotics, autonomous driving, drones), a market projected to be $200B+ by 2035 [02:59:05].
-
Jean Hu (CFO): Financial Model
-
Reiterated the long-term financial targets, emphasizing disciplined capital allocation and a relentless focus on R&D.
-
She stated that the combination of >35% revenue growth and >35% operating margin creates a "very powerful combination" with a clear path to exceeding $20 in earnings per share (EPS) within the 3-5 year forecast period [03:20:39].
