Management explicitly guided for a 50% second-half data center ramp while the Street is modeling flattish growth, creating a massive divergence in expectations.
Thesis: The market is mispricing the execution lag. AMD is transitioning from a cyclical Embedded downturn to a secular Data Center/AI ramp. The 'flat' Q3 guide masks the structural pivot: gross margins are expanding (+100bps to 51%) even as the high-margin Embedded business shrinks, proving the superior economics of the new Data Center mix. The 50% H2 revenue ramp is backed by 'commitments' and specific product launches (MI300, El Capitan), not just hope. The inventory build to 120 days is a feature, not a bug—pre-positioning for a supply-constrained release window.
Verdict: LONG — Conviction: MEDIUM
Catalyst: The Q4 revenue print and MI300 volume shipment confirmation, which will validate the '50% ramp' thesis and force a Street re-rate.
Key Risk: TSMC CoWoS capacity allocation. Management admits it is 'tight in the industry,' and if AMD is second-tier priority behind NVIDIA, the 'commitments' for capacity may dissolve.
The Tell: When pressed on the composition of the Q4 ramp, Lisa Su admitted 'El Capitan is several hundred million.' This reveals that a significant chunk of the 'AI growth' narrative for Q4 is actually non-recurring, lumpy supercomputer revenue, not just pure commercial AI demand.
Friction Level: HIGH_FRICTION — The Street questions the visibility of the back-end loaded ramp. Management claims 'zip code of 50%' growth for H2, but analysts see 'mixed demand' and 'inventory correction' risks.
Report not found
The report data is no longer available. Please return to the archive.