AMD, AI 추론 시장 선점 전략으로 성장 가속화
AMD Is Focused Correctly So I Continue Loading Up
데이터센터 매출이 AI 추론 워크로드에 힘입어 107% 성장했다는 점은 AI 인프라 지출이 가속화되고 있다는 논리를 입증하며, 이는 긍정적인 신호입니다.
핵심 요약
AMD는 데이터센터 매출이 107% 성장하며 AI 추론 시장의 기회를 선점하고 있으며, 견고한 재무 구조를 바탕으로 장기적인 성장 잠재력을 보입니다.
(분석 완료)
Original Article
AMD Is Focused Correctly So I Continue Loading Up
I keep hitting the buy button on Advanced Micro Devices ( NASDAQ:AMD | AMD Price Prediction ), and after this latest quarter I am not close to stopping. My conviction rests on one idea: Lisa Su’s team is aiming AMD at inference, and inference is where the durable dollars in AI will land.
Training a frontier model is a one-time capital event. Running that model in production, every day, for every enterprise customer, is a recurring bill that never stops. Industry estimates place 80% to 90% of total lifetime AI infrastructure costs on the inference side once models move from research into deployment. Capturing even a slice of that pool is a larger prize than winning the pre-training race, and AMD is building products for exactly that workload.
Q2 FY2026 revenue came in at $11.54 billion, up 50.11% year over year, beating consensus by 2.04%. Non-GAAP EPS of $1.66 beat the $1.6107 estimate. Data Center revenue reached $6.72 billion, up 107% year over year, and now accounts for 58% of total revenue. Non-GAAP gross margin expanded to 56% from 43% a year ago.
Then look at who is signing up. Anthropic committed to deploy up to 2 GW of MI450 Series GPUs in Helios racks. Meta signed on for up to 6 GW of Instinct GPUs. OpenAI picked AMD as a core preferred partner for 6 GW of GPUs. Microsoft is scaling Helios racks on Azure. That is committed compute capacity, and Q3 guidance of approximately $13 billion in revenue, implying 41% year-over-year growth, tells me the ramp is real.
The balance sheet lets me sleep. AMD carries debt-to-equity of 0.071, interest coverage of 28.2x, and a net cash position. FY2025 free cash flow reached $5.52 billion, up 129.48%. The MEXT acquisition for AI-powered predictive memory strengthens inference throughput, precisely where the recurring dollars sit.
Readers will ask why not just buy NVIDIA ( NASDAQ:NVDA ). I own that debate. AMD’s market cap of roughly $797.8 billion is a fraction of NVIDIA’s, while its Data Center segment just grew 107% off a smaller base. That is where I want fresh dollars going to work. As for Intel ( NASDAQ:INTC ), AMD’s Client revenue rose 23% year over year on Ryzen strength while Intel keeps ceding server share to EPYC. Intel’s turnaround remains a story. AMD’s is already on the income statement.
U.S. export controls can bite. AMD absorbed roughly $440 million of net inventory and related charges from export restrictions in FY2025, and the prior-year Data Center comparison included an $800 million inventory charge tied to MI308. Trade policy can hit again. My thesis holds because non-China demand from Anthropic, Meta, OpenAI, and Microsoft dwarfs the exposure, and the inference roadmap does not depend on any single geography.
There is no dividend here, and that matters for a retirement audience. This position sits in the compounding side of the portfolio. AMD does return capital through buybacks, repurchasing $1.316 billion of stock in FY2025.
Lisa Su put it plainly: “AI is driving a significant expansion in demand for compute across all of our markets, and our leadership portfolio and growing customer visibility position us exceptionally well to capture this expanding opportunity and deliver substantial revenue and earnings growth in the years ahead.” That is the sentence I come back to every time I add another share.
Contact [email protected] for any questions or corrections.