Research record
Not reviewedNo reviewer, review date, or supporting citations are recorded for this demo thesis. Scores are editorial opinions and have not been verified against cited evidence.
Revision history
No dated revisions are recorded for this thesis.
Investment thesis
The thesis in one line
AMD is the second source the entire industry wants to exist — and even modest share gains in AI accelerators are worth a fortune.
Why a #2 matters so much
Hyperscalers do not want to depend on a single supplier for the most strategic component in their capex budget. That structural desire for a credible alternative is AMD's opening. Its MI300/MI350-class accelerators have closed much of the hardware gap, and its EPYC server CPUs already win share from Intel — throwing off the cash to fund the GPU assault.
The math of share gains
The AI-accelerator market is enormous. AMD doesn't need to beat NVIDIA; capturing even low-double-digit share of a market this large transforms its earnings. That asymmetry — limited need to "win," large reward if it merely participates — is the core of the bull case.
The gap to close
The honest weakness is software. NVIDIA's CUDA is a decade-plus ahead; AMD's ROCm is improving fast but still demands more engineering effort from customers. Until that gap narrows, AMD competes hardest where buyers have the in-house talent to optimise their own stacks.
Stage of the cycle
Early in the accelerator ramp. This is an optionality story: you are buying the probability of share gains, not an established franchise. That means sharper swings around each earnings print.
Scorecard
Upside 8/10 · Safety 6.5/10 · AI exposure 8.5/10. The highest-conviction "second horse" in AI compute.
