AMD has completed the divestment of its data center infrastructure manufacturing business of ZT Systems in the United States, which has been acquired by Sanmina. The transaction is accompanied by a strategic alliance: Sanmina becomes the preferred partner for new product introductions (NPI) for AMD’s rack and cluster-scale AI solutions, while AMD retains the design and “customer enablement” teams from ZT Systems —the expertise in system architecture, integration, and deployment with major clients— to accelerate the quality and time-to-deployment of its cloud AI platforms.
This move aligns with AMD’s roadmap for the AI era data center: expanding from leadership in silicon to software and full systems, building an open and scalable proposition that enables hyperscalers and large enterprises to deploy training and inference capabilities with less friction, faster, and with manufacturing located in the U.S. when required by the customer.
What changes for each party
- AMD leaves behind direct manufacturing of ZT Systems’ business and retains the core elements for its strategy: rack-scale system design, reference engineering, validation, integration services, and customer relationships during production ramp-up. This plays a key role in bringing its GPU Instinct, EPYC CPUs, ROCm software, libraries, and orchestration tools into real-world environments more quickly.
- Sanmina adds a data center infrastructure manufacturing unit capable of full system integration, with liquid cooling lines and volume production in strategic locations (like New Jersey and Texas), with a presence in the Netherlands. It gains scale, exposure, and depth in the cloud and AI markets, positioning itself as a go-to partner for moving prototypes into production with quality and speed.
Why is it relevant now
The generative AI market has shifted focus from standalone chips to large-scale systems: whole racks with GPU, CPU, HBM, high-speed networks, power distribution, and cooling capable of supporting thousands of coordinated accelerators. This “step up” requires co-designing hardware, firmware, software, and operations to maximize performance and reduce costs.
In this context, outsourcing manufacturing to a specialized partner and preserving design and customer enablement talent allows AMD to focus on its differentiated value proposition —silicon + software + system reference— while scaling production with a partner whose core business is building and validating ready-to-ramp infrastructure.
Rack-scale: from “what” to “how”
The rack-scale system is now the basic deployment unit in AI data centers. It’s not just about stacking GPUs: you need to balance power, dimension HBM, optimize PCIe or NVLink routes, meet thermal budgets, select network topologies (e.g., 200/400/800 GbE, InfiniBand), and ensure maintainability and availability.
The system design and enablement that AMD maintains from ZT Systems precisely fulfill this role: turning components into reproducible, qualified, and supportable solutions. Meanwhile, Sanmina provides the factory, process qualification, traceability, supply chain management, and the capability to industrialize racks and clusters under strict tolerances, with liquid cooling when needed and with tight NPI cycles.
Impact on “cloud” and AI clients
For hyperscalers and large cloud providers, the AMD + Sanmina collaboration promises:
- More predictable lead times by separating design/system (AMD) from manufacturing (Sanmina) with framework agreements.
- Shorter certification and acceptance pathways: AMD’s enablement teams that previously operated within ZT Systems will stay on the ground to close tests, KPIs, and SLOs with clients.
- Options for manufacturing in the U.S. and EU (via Sanmina’s global footprint), relevant for regulatory requirements, sovereignty, and supply chain security.
- Consistency across silicon, software, and rack: AMD can deliver software, drivers, compilers, libraries, and system references in a single optimized package, minimizing bottlenecks.
What role does Sanmina play?
Sanmina is a top-tier EMS (Electronics Manufacturing Services) provider with system integration capabilities for critical sectors (industrial, defense, aerospace, communications, cloud). With ZT Systems’ unit, it gains volume and expertise in AI and cloud infrastructure, adding large-scale integration and advanced liquid cooling to its portfolio.
By becoming a preferred NPI partner for AMD in rack and cluster AI, Sanmina will serve as the channel for delivering reference designs and custom configurations from prototype to repeatable production. Practically, this translates into less friction in realizing POCs and pilots at commercial scale, and more location options for manufacturing according to jurisdiction.
Competitive implications
The AI market has rapidly become more vertically integrated. While some competitors favor closed stacks, AMD promotes an open model where third parties —ODM/OEM and EMS partners— handle manufacturing and integration at scale on AMD silicon and software. The alliance with Sanmina strengthens this approach: AMD does not compete with its server OEMs, while maintaining system-level intelligence to set the pace on performance, TCO, and deployment.
For the end customer, the message is pragmatic: real alternatives when choosing suppliers, compliance in sensitive jurisdictions, and roadmaps that avoid lock-in within a single supply chain.
Risks and conditions
Like any supply chain reconfiguration, this operation involves risks:
- The success of the preferred partnership may not meet quality and timeliness expectations.
- The transfer of operations and teams may not translate into net time-to-market gains.
- Temporary impact on relationships with suppliers, clients, and regulators.
- The demand curve in AI: if capital expenditures by hyperscalers soften, rack volume could slow down.
Both AMD and Sanmina also warn that forward-looking statements are subject to uncertainties: market variables, integration, trade policies, and geopolitical factors.
Market signal: focus and speed
Apart from the details, the core message is consistent: AMD aims to accelerate the deployment of rack-scale AI systems without dispersing resources into manufacturing, while Sanmina seeks to lead in integrating and manufacturing critical cloud and AI infrastructure, with the capacity and footprint to support demand peaks.
In a market where next-generation clusters are built on monthly —even weekly— cycles, any reduction in NPI, validation, and ramp-up times provides a competitive advantage.
Market: recent developments for AMD and Sanmina
Valuation info for Advanced Micro Devices Inc. (AMD)
- Advanced Micro Devices Inc. is a listed equity in the USA.
- The current price is $260.34 USD with a change of $0.67 USD (0.00%) from the previous close. The last opening price was $259.35 USD, with an intraday volume of 20,591,567 shares. The intraday high is $264.58 USD, and the low is $256.52 USD.
- The most recent trade occurred on Tuesday, October 28, at 16:04:33 CET.
Valuation info for Sanmina Corp (SANM)
- Sanmina Corp is a listed equity in the USA.
- The current price is $133.68 USD, with a change of -$1.28 USD (-0.01%) from the previous close. The last opening price was $136.56 USD, with an intraday volume of 108,874 shares. The intraday high is $136.675 USD, and the low is $133.61 USD.
- The most recent trade occurred on Tuesday, October 28, at 16:03:31 CET.
Note: Quotes are almost real-time and provided for informational purposes only.
Summary of key points
- What AMD is selling: the manufacturing of ZT Systems’ data center infrastructure business in the U.S.
- What AMD retains: the design teams and customer enablement teams of ZT Systems for rack-scale AI.
- What Sanmina gains: capacity and know-how for integration and manufacturing of racks/clusters for AI, plus the role of preferred NPI partner for AMD.
- What the customer gains: shorter deployment timelines, manufacturing options by jurisdiction, and optimized systems end-to-end (silicon-software-rack).
Frequently Asked Questions
What does it mean that Sanmina is AMD’s “preferred NPI partner”?
It means that following AMD’s reference design and validation, Sanmina will be the selected manufacturer/integrator to bring new products (AI racks/clusters) from prototype to production. In practice, this reduces cycles and standardizes quality, testing, and documentation processes.
Does AMD stop manufacturing servers? Does this affect its GPU/CPU roadmap?
AMD remains active in system business: it retains designs, reference architecture, and enablement with clients. What it cedes is the manufacturing of the acquired business to ZT Systems, which is not its core. Its roadmap for GPU Instinct, EPYC CPUs, and ROCm remains unchanged; in fact, it’s enhanced by scaling capabilities with a dedicated manufacturing partner.
What does Sanmina gain from acquiring ZT Systems’ business?
Immediate scale in cloud/AI, liquid cooling capacity, and plants in the U.S. and Europe; plus deep partnerships with hyperscalers. It positions itself as an end-to-end integrator and manufacturer for mission-critical and rack-scale technologies.
Is there an estimated transaction amount?
Preliminary market info pointed to an approximate value of $3 billion (€2.8–2.9 billion, depending on the exchange rate), comprising cash, equity components, and contingent payments. Both parties now confirm the close of the deal and the strategic partnership.
Sources (selected): Official AMD releases and Sanmina’s announcements regarding closing and NPI agreement; coverage by Reuters, Wall Street Journal, and financial media on preliminary terms and context of the deal.