Key insights
- Wolfe Research highlights substantial upside potential to AMD's earnings estimates for 2026-2027, driven by strong CPU demand and potential for additional large-scale MI450 customers beyond Meta and OpenAI. Management's constructive tone and secured wafer supply suggest capacity to meet significant growth in the server business. This positive outlook, despite recent stock gains, points to potential for further upside, with an analyst event in July serving as a near-term catalyst.
Investing.com -- On Monday, analysts at Wolfe Research described management's tone on a pre-quiet period call with AMD as constructive, saying there is "substantial room for upside to consensus for both CY26 and 27" driven by incremental CPU strength and the potential for additional gigawatt-scale MI450 customers.
Analyst Chris Caso wrote in a note that AMD management appeared comfortable that the company has secured sufficient wafer supply "to support very material growth in the server business this year and next," with wafer allocation potentially supporting demand well ahead of consensus.
Despite the stock's 143% gain since the start of April, Wolfe outlined a path to $25-30 in earnings power, comprising $14 in base EPS driven by 1 gigawatt each from OpenAI and Meta, $7 from agentic CPU demand, and $3-4 from each additional gigawatt of MI450 capacity beyond those two customers.
That would put the stock "under 20x that bull case EPS power," with a near-term catalyst from AMD's July analyst event.
On supply, AMD is said to have noted that "supply of everything is tight," though the company can shift wafers from PC and gaming to datacenter as needed.
MI450 shipments are expected to begin in the latter part of the third quarter and ramp in the fourth, driven primarily by Meta and OpenAI, with AMD expecting "a fairly significant jump" in fourth-quarter revenue followed by another increase in the first quarter.
Wolfe added that AMD also raised its server CPU total addressable market above its prior $120 billion forecast for 2030, with agentic inference workloads cited as "the main driver of the upward TAM revision."