Reward 500 Tiger-coins$Advanced Micro Devices(AMD)$ surged +16.5% after-hours to above $400. $INTC$ jumped +17.6% yesterday. Same week: $SanDisk Corp.(SNDK)$ printed 78.4% gross margins, and $Micron Technology(MU)$ 's CEO called memory a "strategic asset for AI." Two AI infrastructure plays are flying simultaneously — CPUs and memory. Which one runs further?
1. CPU Story: Demand Is Structurally Shifting
AI has moved from training to inference + agentic AI. That shift benefits more than just GPUs.
$AMD$ Q1 Data Center: $5.8B (+48% YoY). Q2 guidance: $11.2B revenue (+46% YoY). Lisa Su was explicit: inference and agentic AI demand for high-performance CPUs and accelerators is "accelerating." EPYC keeps taking share. MI455X ramp starts in 2H.
$Intel(INTC)$ CEO Lip-Bu Tan on his call: AI is moving from foundation models to agents and inference — data center CPU demand is surging. INTC +17.6% in a single day, up +69% from its $67 low.
2. Memory Story: Supply Constraints Are Harder to Fix Than Algorithms
$MU$ CEO Sanjay Mehrotra: memory has become a "strategic asset for unlocking AI's potential." DRAM supply/demand gap is currently 10% and widening:
- Intel's new AI CPUs carry 400GB of general DRAM (4x standard)
- Samsung on its call: customers are already pre-booking 2027 capacity, and the 2027 gap looks larger than 2026
- Meaningful new supply: earliest relief is 2028
$SanDisk Corp.(SNDK)$ this quarter: revenue +217% YoY, gross margin 78.4%, Q2 guided at 80%. wdc Q2 EPS guidance +17% above Street.
Which Side Are You On?
What do you think AMD's fair value is in the inference AI era?
CPU or memory — bigger upside from here?
Or is it effectively the same trade: more AI capex = both go up?
Leave your comments to win tiger coins~
Comments
My fair value view:
• Base: US$450 to US$520
• Bull: US$575+ if MI-series inference demand scales hard
• Bear: US$320 to US$360 on valuation reset
CPU or memory?
Near term: Memory has bigger upside, driven by HBM shortages and pricing power.
Medium term: CPU may quietly compound better, because inference needs orchestration, data movement and efficient serving, not just accelerators.
My view:
Memory = faster upside
CPU = steadier upside
AMD = sweet spot, as it benefits from both.
Bottom line: More AI capex likely lifts both, but memory runs hotter while CPU runs longer.
Prior to the recent surge, many "most followed" models anchored fair value closer to $300, suggesting the current price includes a significant momentum premium.
AMD’s "bigger upside" actually comes from HBM (High Bandwidth Memory) integration. The MI455X and upcoming Helios stacks leverage massive memory advantages to handle larger models than competitors, making memory a core part of the GPU’s value proposition rather than a separate trade.
The "differentiation" only matters if CapEx slows; in that scenario, CPU demand might stay more resilient (due to general-purpose cloud needs), while expensive AI GPU orders would be the first to be cut.
That said, the CPU story is very real. $Advanced Micro Devices(AMD)$ is benefiting from the shift to inference and agentic AI, where CPUs regain importance. Even Intel seeing demand recovery confirms this isn’t a one-player trade.
So to me, memory is a tight supply trade, while CPUs are a demand growth trade. Short term I favor memory, but longer term AMD likely compounds better as inference scales. Positioning-wise, I’d rather hold both but tilt slightly toward memory for now.
@TigerClub @TigerStars @Tiger_comments
$AMD$ Q1 Data Center: $5.8B (+48% YoY). Q2 guidance: $11.2B revenue (+46% YoY). Lisa Su was explicit: inference and agentic AI demand for high-performance CPUs and accelerators is "accelerating." EPYC keeps taking share. MI455X ramp starts in 2H.