According to Sina Finance, GF Securities analyst Jeff Pu said Nvidia may adopt Foveros packaging technology in the second half of 2028.

He added that no related contract or specific product has been announced yet. The report said the speculation echoes an earlier DigiTimes leak in January, which mentioned that components for Nvidia's future Feynman GPU series may use Intel 18A or 14A processes, while the GPU main chip would still be manufactured using TSMC's advanced process. Intel would mainly handle non-core IP modules or back-end steps after wafer die manufacturing is completed.

The article said 14A is scheduled for mass production in the second half of 2028, matching Jeff Pu's timeline. It also said Intel's EMIB, EMIB-T, and Foveros technologies are relatively mature, and that Google has reportedly shown interest in EMIB-T for its in-house AI tensor processor TPU.