AMD’s Zen 7 Grimlock CPUs may use TSMC’s 1.4nm process and new packaging in 2028

AMD is reportedly preparing its next generation Zen 7 CPU architecture earlier than expected, with the lineup codenamed Grimlock and planned around TSMC’s A14 process. A14 is TSMC’s 1.4nm class technology, and the first Zen 7 products are expected around 2028 if the current supply chain timeline holds.

The report says AMD has already started supply chain work for Zen 7, even though Zen 6 has not yet made its full mainstream arrival. That shows how far ahead major chip companies plan, especially when new CPU designs depend on advanced nodes, packaging capacity, and long term manufacturing agreements.

Zen 7 could bring a new CCD design with more cache and stronger AI focused server features

The most important detail is the reported use of TSMC A14. TSMC is expected to begin trial production at its Taichung Fab 25 P1 plant in 2027, with mass production planned for 2028. That timing matches the rumored Zen 7 window.

AMD is also said to be evaluating Powertech’s Fan Out Panel Level Packaging technology. This packaging method could help AMD build more advanced chiplet based CPUs while managing cost, density, and performance. AMD CEO Lisa Su reportedly visited Powertech during a Taiwan trip, with the visit linked to advanced packaging allocation.

The CCD design may also change. The report says Zen 7 could use a new chiplet with up to 16 cores. That would be a notable step because AMD’s current mainstream CCD approach has generally centered around 8 core chiplets. A 16 core CCD could help AMD improve density and scale core counts more efficiently across desktops and servers.

Cache is another major part of the rumor. Zen 7 chips with 3D V Cache could reportedly reach up to 224MB of L3 cache on a single 3D V Cache CCD. That would be a big increase and could matter for gaming, workstation workloads, databases, and server applications that benefit from large low latency cache.

For data centers, Zen 7 is also expected to expand AMD’s AI related CPU capabilities. The report mentions updated MATRIX engine features and broader support for AI data formats. That fits the direction of the server market, where CPUs are increasingly expected to support AI workloads alongside GPUs and accelerators.

The timing also puts AMD in a direct manufacturing race with Intel. TSMC’s A14 node is expected to compete with Intel’s 14A process around the same period. Intel has been trying to rebuild its foundry business and win major customers, while AMD remains one of TSMC’s most important advanced node partners.

That competition matters because advanced process access is becoming more strategic. AI companies, cloud providers, and server buyers need more compute, and CPU demand is expected to remain strong. AMD, Intel, and other chipmakers are all trying to secure manufacturing capacity before demand tightens further.

Zen 7 is still far away, so these details should be treated as early supply chain information rather than final product specifications. AMD has not officially announced the full Zen 7 lineup, product names, prices, socket details, or final performance targets.

Even so, the rumor gives a clear direction. AMD appears to be preparing Zen 7 as a major next step built around TSMC’s 1.4nm class process, denser chiplets, more 3D V Cache, advanced packaging, and stronger AI focused server features. If the timeline stays on track, 2028 could be a major year for AMD’s CPU roadmap.