Chapter 6 · Risk · End
What could
break the brief?
Even great businesses have real downside. Honest investing means looking at the risks straight on.
↓ scroll to read
From the 10-K · the *software* problem AMD can't outspend
We must continue to invest in software capabilities and the open ecosystem through the AMD ROCm platform... improving performance for generative AI workloads and simplifying the developer experience.
↳ Translation: we have to convince every AI researcher to switch off CUDA. Nvidia spent 20 years building that ecosystem. AMD has been at it for ~5. Catching up is *years* of work, not quarters.
Source · 10-K · Item 1 Business — Data Center / Software · FY2025 · Filed Feb 4, 2026
✦ Teach me
Customer concentration risk
AMD's data center growth comes from a small number of hyperscale customers — Microsoft, Meta, Oracle, Google, Amazon, plus OpenAI. When one hyperscaler delays orders or shifts to a competitor, AMD's revenue moves in real time.
The upside is huge — one OpenAI deal can anchor years of revenue. The downside is symmetric — one customer reshuffle can dent a whole quarter.
Wall Street calls this
Customer concentration
Hyperscalers are *all* building their own AI chips too (Google TPU, AWS Trainium, Microsoft Maia, Meta MTIA). Every customer is a future competitor. The faster AMD moves, the better.
From the 10-K · the single-supplier risk
We rely on third-party wafer foundries, primarily Taiwan Semiconductor Manufacturing Company (TSMC), to manufacture our products. Disruption to TSMC's operations or supply could materially adversely affect our business.
↳ AMD doesn't own a single fab. Every chip — EPYC, Ryzen, Instinct, Versal — is made by TSMC. A Taiwan disruption, a supply shock, or a TSMC capacity allocation change could halt AMD's roadmap.
Source · 10-K · Risk Factors — Manufacturing and Supply · FY2025 · Filed Feb 4, 2026