AMD's Strategic Push into Asia Gains Momentum with Key Partnerships

AMD announces major partnerships with Samsung and NAVER Cloud, backed by record data center revenue, to scale its AI business and compete with Nvidia on full system solutions.

In a concentrated 48-hour period, Advanced Micro Devices, Inc. (AMD) unveiled a series of significant partnership agreements across Asia, marking a bold strategic expansion. The company's clear objective is to mount a serious challenge to Nvidia's dominance in the global AI infrastructure market.

Financial Backdrop and Market Ambition

This aggressive partnership drive is supported by robust financial performance. For the full fiscal year 2025, AMD achieved record revenue of $34.6 billion. Its data center segment was a primary growth driver, expanding by 32% to $16.6 billion and now accounting for nearly half of the company's total sales. Looking ahead, AMD is targeting an annual growth rate exceeding 60% for its data center business over the next three to five years, with the goal of scaling its AI-related revenue to tens of billions of dollars by 2027.

Notably, the company's shares currently trade approximately 24% below their 52-week high from October 2025, suggesting the market may not have fully priced in AMD's ambitious growth trajectory.

Landmark Agreements in South Korea

CEO Dr. Lisa Su, visiting South Korea for the first time since becoming CEO in 2014, personally oversaw the signing of multiple deals. The first major announcement was a partnership with Samsung Electronics. Under this agreement, Samsung will serve as the primary supplier for HBM4 high-bandwidth memory for AMD's upcoming Instinct MI455X accelerators. Samsung will also supply DRAM solutions for AMD's sixth-generation EPYC server processors, codenamed "Venice." The two technology giants are additionally exploring a potential collaboration in the foundry space.

Immediately following the Samsung deal, AMD entered a cooperation agreement with NAVER Cloud, led by CEO Soo Yeon Choi. The pact grants NAVER Cloud early access to the MI455X GPUs and includes plans for broader deployment of AMD EPYC processors across its infrastructure. A particularly significant element is the planned migration of NAVER's large language model, "HyperCLOVA X," to run on AMD's GPU environment—a move that could substantially reduce NAVER's reliance on Nvidia hardware.

Building Complete Systems: The Helios Platform

Beyond component sales, AMD is moving to offer complete, integrated solutions. On March 16, the company, in collaboration with Celestica, announced work on an open rack-scale platform named "Helios." Celestica will handle the development and manufacturing of scale-up network switches based on the Open Compute Project standard, designed to interconnect AMD's MI450 generation of GPUs. The Helios platform is expected to become available to customers by the end of 2026.

This strategy represents a fundamental shift: AMD aims to compete not just at the accelerator level but by delivering full rack-scale systems, similar to the approach of its main rival. The use of open standards, such as Ultra Accelerator Link over Ethernet, is intended to appeal to clients seeking to avoid proprietary infrastructure from a single vendor.

Securing Major Commitments

The expansion is further bolstered by substantial customer commitments. Meta Platforms has entered into a multi-year agreement valued at an estimated $100 billion for the purchase of MI450 GPUs and EPYC processors. Combined with a separate agreement from OpenAI, AMD now holds commitments representing over 12 gigawatts of GPU capacity for the coming years, underscoring the strong demand for its AI and data center products.