Intel Foundry to Support NVIDIA Feynman GPUs

Intel Foundry is expected to provide packaging and wafer support for NVIDIA's next-generation Feynman GPUs, scheduled for release around 2028.

Intel Foundry is set to provide critical packaging and wafer support for NVIDIA's upcoming Feynman GPU architecture, which is slated to succeed the Rubin generation around 2028. The Feynman GPUs will incorporate advanced features including 3D Die-Stacking, a new HBM standard, and a new host CPU named Rosa. Intel's 14A process node is scheduled for a volume ramp in 2028, aligning with the Feynman timeline. The partnership leverages Intel's Foveros Direct3D and EMIB-T technologies, which compete with TSMC's SoIC and CoWoS-L solutions used in products like AMD's MI400 series. This collaboration builds on existing ties, such as NVIDIA's Rubin NVL8 using Xeon 6 CPUs and Intel's Serpent Lake SoCs featuring RTX GPUs. Intel CEO Lip-Bu Tan is increasing capital expenditure for advanced packaging and fabs to support these initiatives, as noted by MoorInsStrat CEO Patrick Moorhead. The deal follows previous agreements to build custom Xeon chips with NVIDIA's NVLink interconnect solution.