NVIDIA Feynman AI Chip Design Reflects Advanced TSMC Capacity Constraints and Hybrid Manufacturing Strategies for 2028 Release
Reports regarding the next generation NVIDIA AI chip known as Feynman suggest a 2028 release window. This upcoming hardware appears set to utilize the latest A16 process from TSMC (TSM US). However early indications from the supply chain imply that even a company as large as NVIDIA is struggling to secure enough production space on the A16 nodes. As a result the company is reportedly planning to use a split design where only the most essential parts of the chip use A16 while others rely on the older N3P process.
This hybrid approach appears to be a direct consequence of a shortage in advanced manufacturing capacity. TSMC has not officially commented on these specific customer plans but industry observers suggest that the demand for the A16 process is likely to exceed the available supply for the next few years. The shift toward this technology is significant because it introduces what is known as Super Power Rail design which changes how electricity is delivered to the transistors.
The A16 technology is essentially an evolution of the 2 nanometer process. Its main feature involves moving power lines to the back side of the wafer. This change seems likely to improve logic density and overall power efficiency for high performance computing tasks. TSMC is currently working toward mass production and their roadmap shows a steady increase in output over the coming years.
- Monthly capacity for A16 is expected to reach 20000 wafers by the end of next year
- Production might grow to 40000 wafers per month by 2028
- The total 2 nanometer family of processes is targeted to reach 200000 wafers every month
This move toward A16 appears to be creating a ripple effect for other companies in the supply chain. Because the new design requires more intensive polishing and grinding of the silicon wafers firms like China Sand Co Ltd are expected to see increased demand for their tools. Experts suggest that the number of chemical mechanical polishing passes could increase by about 30 percent compared to previous methods.
Other companies like Sunmicron are also adjusting their plans. Reports indicate that Sunmicron is expanding its capacity for carrier wafers to keep up with the new requirements. It appears that their monthly production could grow by as much as 50 percent to reach 1.2 million wafers by the end of this year. This growth is a clear indicator of how the AI wave is pushing the entire industry toward more complex manufacturing.
NVIDIAs decision to mix different process nodes for the Feynman chip is a practical look at the reality of the market. It suggests that even the biggest players in the AI space must balance their ambition with the actual availability of factory space. As 2028 approaches the industry will be watching closely to see if TSMC can expand its facilities fast enough to meet the needs of the growing AI sector.
