AMD divests ZT Systems' manufacturing division for $3 billion, aiming to streamline and build an AI full-stack platform
AMD divests its data center infrastructure manufacturing business ZT Systems for $3 billion, handing it over to Sanmina, aiming to concentrate resources in a lighter asset and more efficient model to build a high-performance AI developer full-stack platform. The deal includes up to $450 million in contingent consideration and is expected to be completed by the end of 2025. This move will reduce fixed asset depreciation and increase investment in high-margin core businesses
According to Zhitong Finance APP, NVIDIA's strongest competitor in the AI chip and PC chip sectors, AMD (AMD.US), announced on Monday local time that it will sell its ZT Systems data center infrastructure manufacturing business in the United States for $3 billion in cash and stock to Sanmina (SANM.US), a company focused on electronic product manufacturing services. This move by AMD aims to allow the chip giant to operate with a lighter asset base and higher operational efficiency, concentrating resources to build and accelerate the delivery of its exclusive high-performance "CPU + GPU + software architecture" AI developer full-stack platform.
It is understood that after the transaction is completed, Sanmina will become AMD's preferred new product introduction (NPI) manufacturing partner for cloud rack-level and cluster-level AI solutions. The transaction amount also includes up to $450 million in contingent consideration.
AMD stated that the company will retain the design and customer empowerment expertise of ZT Systems' rack-level AI solutions to accelerate the quality control and deployment processes and timelines of AI applications for cloud large customers in both B-end and C-end markets.
The transaction is expected to be completed around the end of 2025. The plan to seek strategic partners to acquire ZT Systems' data center infrastructure manufacturing business was disclosed as early as August 2024 when the formal acquisition was announced.
By divesting the ZT Systems manufacturing business and signing a "preferred NPI partnership" agreement with Sanmina, AMD aims to operate with a lighter asset base and higher operational efficiency, concentrating resources to build and accelerate the delivery of its exclusive "CPU + GPU + software architecture" AI developer full-stack platform.
The data center complete machine production line belongs to a low-margin, high-capital intensity segment; by outsourcing to Sanmina, AMD can reduce fixed asset depreciation and invest a larger scale of capital into core business platforms with higher margins and stronger technical barriers, such as the MI300 series high-performance AI GPUs, EPYC CPUs, and ROCm/AI development toolchains.
From 2024 to 2025, AMD will continue to promote its "one-stop AI developer platform ecosystem" with high profile: MI300X/GPU + EPYC CPU + Pensando DPU + ROCm 6.x AI software stack; and collaborate with PyTorch, TensorRT-LLM, and Hugging Face for optimization.
Sanmina will undertake the manufacturing of ultra-large-scale racks and complete machines, global logistics, and on-site installation. Its mature hardware manufacturing network can shorten delivery times, avoid Trump-era tariff policies and supply chain bottlenecks, and align with the manufacturing transfer requirements under the latest trade agreements of the Trump administration.
Therefore, AMD's large-scale sale of the ZT Systems manufacturing division and designation of Sanmina as the preferred NPI partner represents a strategic rebalancing of "light assets + integrated hardware and software," with the core purpose being "to achieve higher operational efficiency and improve AMD's exclusive AI developer platform," focusing on enhancing the experience and deployment speed of the EPYC + MI300 series + ROCm AI developer platform
