According to a Linkedin profile, AMD is working on another chiplet-based GPU — UDNA could herald the return of 2.5D/ 3.5D chiplet-based configuration

AMD is reportedly developing next-generation high-end GPUs based on advanced 2.5D/3.5D chiplet architectures, as indicated by the LinkedIn profile of senior fellow Laks Pappu, who is overseeing future Radeon and data center GPU designs. This strategic direction, potentially for the RDNA 5 generation arriving in late 2026/early 2027, suggests AMD is aiming to overcome the escalating costs and manufacturing complexities of monolithic high-end GPUs. While multi-tile consumer GPUs present significant technical challenges, this move could enable AMD to re-enter and aggressively compete in the high-end desktop and data center graphics markets.

Analysis

Information from a senior AMD employee's LinkedIn profile indicates a strategic pivot towards developing high-end, chiplet-based GPUs for its next-generation RDNA 5 architecture, expected in late 2026 or early 2027. This strategy, led by Laks Pappu, a former Intel discrete graphics architect, aims to leverage advanced 2.5D/3.5D packaging to re-enter the lucrative high-end consumer and data center GPU markets. This marks a significant shift from the current RDNA 4 generation, which cedes the premium segment to Nvidia by only competing in the mid-range. While creating multi-tile consumer GPUs presents substantial technical hurdles related to latency, synchronization, and software complexity, AMD's extensive experience with chiplet designs in its CPU division and the experimental disaggregation in its Radeon RX 7900-series processors provide a foundational basis. If successful, this approach could circumvent the escalating costs and yield challenges of large monolithic GPUs, potentially enabling AMD to competitively challenge Nvidia's market dominance in the high-performance segment within the next 2-3 years. The timeline suggests the RDNA 5 design is already in the advanced tape-out phase, indicating a firm commitment to this long-term product roadmap.

Key Decisions for Investors

• Investors should view this as a positive long-term catalyst for AMD, as it provides a credible roadmap to re-enter the high-margin, high-end GPU market, though financial impact is not expected until late 2026 at the earliest.
• Monitor for technical milestones and performance leaks related to the RDNA 5 architecture, as successful execution on overcoming the significant challenges of multi-tile gaming GPUs is the primary risk factor to this strategy.
• This news reinforces the long-term competitive pressure on Nvidia's high-end GPU segment and highlights a strategic talent loss for Intel's graphics division, though it does not alter the near-term competitive landscape.
• Recognize that AMD's near-term GPU revenue will remain constrained to the mid-range market, as the current RDNA 4 generation is explicitly positioned away from the high-end, and this news does not change that short-term outlook.