The Central Processing Unit Makes a Strategic Comeback in AI Infrastructure

For years, the graphics processing unit (GPU) has been the undisputed star of artificial intelligence computing. Advanced Micro Devices (AMD) is now shifting the spotlight back to a more foundational component: the central processing unit (CPU). This strategic pivot is being backed by multibillion-dollar agreements with major technology firms and a compensation plan for CEO Dr. Lisa Su that is directly tied to highly ambitious stock performance targets.

Management’s Growth Vision and Incentives

Despite strong operational performance, AMD’s share price has faced recent pressure. Following a market disappointment with its latest quarterly report, the stock trades at €170.38, approximately 25% below its 52-week high reached in October 2025.

In this context, the company’s board of directors has linked executive compensation directly to significant value creation. Dr. Lisa Su was granted a special equity award valued at $75 million. These shares will only vest if the company’s stock reaches certain price milestones. The highest payout tier is contingent on the share price hitting $600 by March 2031. Concurrent sales of 85,000 shares by the CEO were executed under a pre-established trading plan and represent a routine transaction.

The leadership team forecasts annual revenue growth of 35% over the next three years. This outlook is founded on the industry’s transition from pure GPU clusters to balanced infrastructure incorporating processors, graphics units, and networking technology. AMD’s strategic direction through the end of the decade is defined by its capability to deliver these complete system solutions to hyperscale customers.

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

CPUs: The Orchestrators of Advanced AI

The rapid evolution of AI is creating new demands. So-called “agentic AI” involves increasingly complex, multi-stage workflows where GPU-only systems encounter limitations. In this new environment, CPUs are reclaiming a critical role as system orchestrators, managing memory, preparing data, and controlling workflow processes.

AMD is positioning its upcoming EPYC processor family, codenamed “Venice,” precisely for these requirements. The company leverages its modular chiplet architecture to flexibly tailor computing power and energy consumption to data center needs. Internal estimates indicate that even the current fifth-generation EPYC processors deliver more than double the energy efficiency per watt compared to Nvidia’s Grace Superchips.

Substantial Customer Commitments Fuel Expansion

This strategic focus is reflected in the company’s financial results. The data center segment grew 39% year-over-year and now constitutes over half of total revenue. AMD reported record fourth-quarter revenue of $10.3 billion.

This growth is underpinned by massive commitments from large customers. Meta and OpenAI have pledged to acquire GPUs with a total capacity of 12 gigawatts over the coming years. Simultaneously, Meta, a primary customer for the sixth-generation EPYC platform, is significantly expanding its CPU partnership. According to CEO Lisa Su, current demand for server CPUs substantially exceeds the company’s prior forecasts.