- AMD is projected to outpace NVIDIA in EPYC Venice CPU shipments by 2027, according to a Morgan Stanley report.
- The report estimates 6.75 million AMD processors compared to 5.75 million NVIDIA Vera CPUs.
- TSMC is boosting production capacity, which is crucial for sustaining demand in AI hardware.
- Custom silicon developments by companies like OpenAI and Google pose a competitive threat to both AMD and NVIDIA.
- Despite the competition, NVIDIA is expected to remain TSMC’s largest customer due to its leadership in AI hardware.
The landscape of AI hardware is evolving rapidly, with the data centre CPU market experiencing intense competition outside the traditional GPU focus. A recent report from Morgan Stanley provides a significant analysis, indicating that AMD might ship more EPYC Venice CPUs than NVIDIA’s Vera CPUs by 2027. Specifically, AMD is on track to dispatch approximately 6.75 million processors, while NVIDIA is estimated to ship about 5.75 million units of its Vera line. Despite this shift, NVIDIA is anticipated to retain its leadership in the AI hardware domain.
This forecast arises amid TSMC’s aggressive plans to bolster its advanced packaging capacity. Morgan Stanley suggests that TSMC could ramp production to 200,000 wafers per month by 2027, a crucial development given the soaring demand for agentic AI technologies. However, even with increased output, NVIDIA is expected to remain TSMC’s primary client, driven by significant demand for CoWoS-L packaging for its Blackwell and Rubin GPUs.
The rivalry between AMD and NVIDIA encapsulates varying approaches to the market. The forthcoming EPYC Venice processors, based on the Zen 6 architecture and employing TSMC’s 2nm manufacturing process, signal AMD’s intent to capitalise on the burgeoning demand for high-performance computing. In contrast, NVIDIA’s Vera, operating on a 3nm process, is tailored specifically for AI infrastructure. The nuances of their design philosophies underline a division of tactical focus, even as both products vie for dominance in a shared marketplace.
Morgan Stanley’s projections suggest a remarkable year-over-year growth for AMD, with a potential shipment increase of over five times compared to 2026. This growth highlights the increasing reliance on server platforms that facilitate both AI and high-performance computing—an area where AMD is poised for substantial gains. While NVIDIA also appears positioned for success, particularly with its forthcoming Vera product launch anticipated to generate significant volume, the competitive environment is rapidly intensifying.
Intriguingly, neither company stands alone as the principal threat in this evolving market. Tech titans like OpenAI, Amazon, and Google are investing heavily in custom silicon designs, including OpenAI’s Jalapeño inference chip developed in collaboration with Broadcom. Such developments could exert additional pressure on traditional CPU and GPU vendors and further strain TSMC’s manufacturing capabilities.
The competitive dynamics in AI hardware are shifting from a GPU-centric focus to a more balanced attention on both CPUs and advanced packaging technologies. The potential for AMD to eclipse NVIDIA in CPU shipments by 2027 may reshape market expectations, but NVIDIA’s enduring leadership in AI accelerators indicates that the battle for supremacy across the silicon frontier remains intricate and multifaceted.
As the tech landscape continues to evolve, an interesting fact to consider is that despite the fierce competition, the global semiconductor market is projected to surpass $1 trillion by 2030, highlighting the critical role these components will continue to play in driving innovation and economic growth. For further details, refer to the original report on eTeknix.
