AMD: Building the Foundational Infrastructure for the AI Era

Advanced Micro Devices (AMD) is strategically evolving beyond its role as a semiconductor supplier. The company is actively shaping the core technological foundations upon which future artificial intelligence systems will be built. A clear signal of this ambition is its co-founding of the Optical Compute Interconnect Multi-Source Agreement (OCI MSA) alongside industry giants Broadcom, Meta, Microsoft, Nvidia, and OpenAI. This move represents a significant strategic pivot from purely selling products toward influencing and setting industry standards.

Operational Momentum and Financial Performance

The strategic initiatives are underpinned by robust financial and operational results. AMD’s revenue for 2025 grew by 34% to nearly $35 billion, while net income surged by 164% to $4.3 billion. The data center segment, now accounting for over half of total revenue, expanded by 39% year-over-year. Company leadership is targeting sustained annual growth exceeding 35% in this division over the next three to five years.

This growth is fueled by substantial commitments from major technology firms. Meta has entered into a multi-year agreement to purchase GPUs from AMD with a capacity of six gigawatts. This follows a structurally similar, pre-existing deal with OpenAI, also for six gigawatts. Combined, these agreements represent commitments for twelve gigawatts of GPU capacity. Using Nvidia’s pricing as a benchmark, each gigawatt is estimated to be worth approximately $35 billion.

Addressing the Physical Limits of AI Clusters

The drive for new interconnect standards is a direct response to a pressing technical challenge. Current AI clusters are hitting physical limitations. The copper-based connections linking processors, accelerators, and switches are increasingly inadequate for the massive data transfer demands of modern AI architectures. The OCI MSA consortium aims to create an open, cross-vendor standard for optical connections to eliminate these bottlenecks.

Should investors sell immediately? Or is it worth buying AMD?

The proposed technology roadmap starts with 200 Gbps per direction, followed by a second generation at 400 Gbps, with a long-term path reaching up to 3.2 Tbps per optical fiber. This translates to faster and more energy-efficient connections within large-scale AI data centers. AMD is contributing its UALink standard to the effort, while Nvidia is bringing its NVLink technology, with the goal of creating a common physical layer that bridges both ecosystems. Notably, Intel and Qualcomm have not yet joined the consortium, a factor that may influence future design decisions in the industry.

Navigating Market Risks and Competitive Responses

Despite this strong operational backdrop, AMD’s share price has declined approximately 11% since the start of the year and remains well below its 52-week high. Market sentiment faces pressure from two primary concerns. First, Chinese GPU newcomer Lisuan Technology has announced products that could create pricing pressure for both AMD and Nvidia. Second, large customers like Meta are developing their own custom AI chips—a trend that could potentially shrink the addressable market for external semiconductor suppliers in the long term.

AMD’s strategic counter to these challenges is becoming evident. The company is expanding production of its Instinct MI355X processors through its partnership with Flex in Austin, Texas, securing manufacturing capacity for ongoing hyperscaler deployments. More fundamentally, its co-founding role in the OCI MSA structurally embeds AMD within the architecture of future data centers, positioning it as a co-architect of standards rather than merely a component supplier.