AMD reports first-quarter results after the close Tuesday on a Wall Street consensus of $9.84 to $9.87 billion of revenue, up roughly 32 percent year-on-year, and a data-center segment guided near $5.56 billion, up 51.5 percent. [1] The print is the company's first chance since the OpenAI 6-gigawatt deal and the Meta 6-gigawatt deal to translate two framework announcements into firm purchase-order commentary. The paper's Monday preview framed the print as the inversion of the Nvidia-discount trade — AMD now at 31× forward earnings against Nvidia's 28×.

The watch line is MI450 cadence. CEO Lisa Su told analysts in March that "we have customers who are anxious to get it in their data centers" — a register the chip-industry tape has been parsing since. [2] AMD has guided MI355 to ramp through Q4 2025 into the first half of 2026; MI450 inflection is targeted for the third quarter, with volume ramp into the fourth quarter and into 2027. [3] The question for Tuesday's call is whether any of those dates move forward, hold, or quietly slip.

Both hyperscaler deals carry the same architecture. AMD issued OpenAI a warrant for up to 160 million shares vesting in tranches as deployment scales from 1 to 6 gigawatts; Meta got the same warrant structure on the same terms. [4] The first 1 gigawatt of OpenAI MI450s is committed for second-half 2026 deployment, the first Meta gigawatt the same. Q1 reflects neither shipment. The call's commentary on warrant accounting, gross-margin trajectory, and back-half revenue guide is what investors will be reading the transcript for.

The bear-side tape signal arrived last week. Cathie Wood's ARK funds sold roughly $80 million of AMD stock into the print. [5] Wood's history of cutting positions ahead of major prints does not always carry — she trimmed Nvidia ahead of two of its biggest beats — but the size of the sale into a 74-percent April rally is the contrarian read on the bookbuild. AMD shares ran 74 percent in April against Nvidia's 14 percent. [6]

The China question is the structural drag. The H200 December 2025 25-percent-cut approval has not produced revenue at Nvidia, per CFO Colette Kress's last commentary. [7] AMD's MI300 and MI325 China line carries the same export-control ambiguity. The Tuesday call will be parsed for any specific commentary on China-export volumes alongside the U.S. data-center ramp. A clean separation — strong U.S. data-center, no China contribution — keeps the tape in line with the framework. A surprise China contribution, or a guide that bakes one in, would be the divergent read.

The Cerebras frame sits in the same trading day. The IPO bellwether for the AI-chip category cut its headline from $40 billion to roughly $26.5 billion midpoint Monday, even as indications of interest exceeded $10 billion. The bookbuild chose to clear into the Berkshire-and-Abel tape rather than price through it. AMD prints into the same room — but with hyperscaler commitments Cerebras does not have. The tape is now bifurcated by who owns the multi-gigawatt customer book.

The operational read for Tuesday is binary. A clean Q1 beat, a back-half guide that bakes in MI450 ramp acceleration, and any specific OpenAI or Meta shipment dates supports the inverted multiple. A guide that holds with no MI450-specific data points opens the discount AMD spent April closing.

The structural read is longer. Data-center revenue grew from effectively zero to threatening Intel's most profitable line in 36 months. [8] If MI450 inflection lands in Q3 as guided, the next 36 months are a different conversation — and the print Tuesday is where the conversation either advances or pauses.

-- THEO KAPLAN, San Francisco