AMD Sells ZT Fab Unit, Bolsters AI Talent to Challenge Nvidia

Data-center push sharpened by $3 billion sale and $1.6 billion engineering buyout

Author's Avatar
May 20, 2025
Summary
  • $3 billion sale to Sanmina funds retention of 1,200 engineers to rival Nvidia
Article's Main Image

AMD (AMD, Financial) offloads ZT Systems' manufacturing arm for $3 billion while snapping up its 1,200-strong engineering team to sharpen its edge against Nvidia (NVDA, Financial) in the data-center GPU race.

Citi Research analysts, led by Christopher Danely, note that last August's $4.6 billion ZT deal has been split: Sanmina will acquire the factory business for $2.55 billion in cash and stock plus $450 million in earn-outs, while AMD retains the engineers for roughly $1.6 billion—about $1.33 million apiece—to accelerate hyperscaler deployments and bolster system-level expertise.

Danely argues that “additional system experience and faster deployment times” should help AMD close the gap on Nvidia's market share.

Under the agreement, Sanmina becomes AMD's preferred manufacturing partner for cloud-rack and cluster-scale AI solutions, reflecting a strategic pivot from in-house production to a fab-lite model.

The transaction, expected to close near year-end 2025, underscores AMD's focus on high-margin chip design over capital-intensive fabs. Citi kept its Neutral rating and $100 price target on AMD, highlighting the balance between near-term dilution from the $1.6 billion engineering buyout and long-term gains from faster product rollouts.

AMD, which derives roughly 30% of sales from PCs, faces stiff competition not only from Nvidia in AI GPUs but also from Intel (INTC, Financial) in data-center CPUs. By beefing up its in-house design team and offloading manufacturing risk, AMD aims to streamline R&D investment and speed time to market for MI300 and future MI400 data-center accelerators.

Why It Matters: Retaining ZT's engineering talent while outsourcing production could deliver faster hyperscaler deployments and help AMD narrow the performance and deployment gap with Nvidia's entrenched data-center GPU ecosystem.

Disclosures

I/we have no positions in any stocks mentioned, and have no plans to buy any new positions in the stocks mentioned within the next 72 hours. Click for the complete disclosure