JP Morgan Expects AMD’s AI GPU Division to Grow 60% in 2025, Backed by Oracle’s 30,000 MI355X GPU Order
According to a new note by JP Morgan analyst Harlan Sur, AMD’s AI GPU business is poised for significant growth in 2025, potentially increasing by over 60% compared to last year. This bullish projection comes at a time when sentiment around AMD’s AI prospects has been lukewarm until recently—evident in industry remarks such as Wolfe Research’s December 2024 assessment and George Hotz’s January 2025 comment about a lack of enthusiasm for AMD hardware at comma.ai. However, momentum seems to be shifting in AMD’s favor, as illustrated by key developments cited by Sur.
Four Major Growth Catalysts for AMD
In his latest research note, Harlan Sur identifies four vital factors driving AMD’s near- to mid-term performance:
Server CPU Market Share Gains
AMD continues to chip away at competitor market share in the lucrative server CPU space. Sur believes the rollout of next-generation EPYC CPUs will sustain this momentum.Desktop and Notebook CPU Market Share Gains
Alongside servers, AMD is projected to keep expanding in the desktop and mobile CPU segments, capitalizing on recovering consumer demand.Growth in Cyclical Businesses (Gaming/Embedded)
Sectors such as gaming consoles and embedded systems are expected to bounce back, further boosting AMD’s overall revenue streams.Strong Growth in AI GPUs
Sur's forecast calls for AI GPU revenues to swell by more than 60% in 2025. Central to this momentum is the upcoming MI350 accelerator line and AMD’s drive to migrate existing hyperscaler clients—Microsoft, Meta, and Oracle—to the new platform.
Oracle’s 30,000 MI355X GPU Order as a Big Win
An illustrative example of AMD’s growing AI GPU traction is Oracle’s recent multi-billion dollar order for AMD’s next-generation MI355X accelerators. According to Sur, Oracle plans to deploy these GPUs—initially 30K units—in a massive cluster suitable for both AI training and inferencing, highlighting industry confidence in AMD’s GPU roadmap.
Transition to MI350 and Beyond
JP Morgan reports that AMD is working aggressively to move key partners off older GPU platforms to the MI350 series by mid-2025, thus preparing for an even more powerful “MI400” GPU generation in 2026. The MI400 line could potentially feature rack-scale configurations of over 100K GPUs per cluster.
Summarizing these developments, Harlan Sur concludes:
“Overall, the key message is that the team’s strong and diversified data center/enterprise/client compute portfolio will drive strong growth in CY25 — a combination of share gains, improving demand/cyclical trends, and growing momentum for its next-gen AI compute solutions.”
In related news, Reuters reports that TSMC is encouraging major clients—NVIDIA, AMD, and Broadcom—to invest in a joint venture if Intel and TSMC consolidate their foundry operations. While no concrete details have been confirmed, such a partnership could reshape future alliances in the semiconductor manufacturing sphere.
Do you think AMD can sustain over 60% growth in its AI GPU segment next year, or will market competition and new entrants slow its momentum? Share your thoughts below!
