They guided the hottest segment in tech to be flat for six months, and the stock didn't collapse.
Thesis: The street interprets the H1 2025 data center flattening as demand saturation. This is a misread of a product transition. AMD is pulling MI350 forward—not because MI300 failed, but because yields ('silicon has come up really well') allow them to chase higher ASPs earlier. The $5.7B inventory build is not dead stock; it represents pre-loaded wafers for the H2 ramp. The 'strong double-digit' 2025 DC growth guide mathematically necessitates a massive back-half acceleration. You buy the transition air pocket before the MI350 margin expansion hits.
Verdict: LONG — Conviction: HIGH
Catalyst: MI350 volume shipments in mid-2025 confirming the yield and ASP uplift, breaking the 'flattish' revenue trend.
Key Risk: MI350 yield issues or delay would turn the 'air pocket' into a structural 'growth cliff' if CoWoS capacity is lost to competitors.
The Tell: When analyst Stacy Rasgon asked about sequential math implying a GPU decline, Lisa Su interrupted to correct 'strong double digits' to just 'double digits' for Q4 sequential, then admitted H1 '25 DC GPU will be 'consistent with H2 '24.' She explicitly capped near-term expectations to preserve the beat-and-raise setup.
Friction Level: HIGH_FRICTION — Bears see the H1 2025 data center revenue flatten as a 'Capacity Ceiling' or demand saturation. Bulls see it as a deliberate 'Product Transition' to higher-margin MI350. The disagreement is fundamental: Is demand stalling or is supply reloading?
Report not found
The report data is no longer available. Please return to the archive.