NVIDIA’s Rubin Ultra Could Hand Intel a Lifeline as UBS Flags EMIB-T Packaging for 4-Chip Variant

Ramish Zafar

As multiple reports claim that Intel is aggressively marketing its EMIB-T packaging technology to AI chip designers in order to shape itself as an alternative to TSMC's CoWoS packaging technology, another customer that it might land could be NVIDIA, according to UBS. The details suggest that a 4-chip variant of the Rubin chip could use Intel's EMIB-T packaging process instead and provide the chip manufacturer with an important role in NVIDIA's and the artificial intelligence supply chain.

Intel Purportedly Manages To Rein In NVIDIA As Key Customer For EMIB-T Packaging

One of the key advantages of the EMIB packaging technology stems from the fact that it uses substrates instead of interposers to link the chip die with the board. As a result, manufacturers are not limited to size constraints that are in play when using an interposer, which allows them to make larger chips without the added complexity of stitching multiple components together.

Related Story Cooler Manufacturer Doubles Down on 1000W GPUs With Twin 360mm AIO Monster, Signaling Where NVIDIA’s Rubin RTX Cards Are Headed

This advantage also appears to have attracted NVIDIA into Intel's fold. According to the UBS note, NVIDIA's margins up until calendar year 2027 are dependent on the Rubin AI chips. Specifically, the bank adds that the margins will depend on whether NVIDIA decides to offer two-chip and four-chip versions of the Rubin Ultra GPUs.

NVIDIA's Margins Will Depend On The Type Of Rubin Chips It Offers, Says UBS

According to UBS:

We assume gross margin to remain in the ~75% range at least through C2027 (e.g. the Rubin generation) with out-year margins depending to some degree on whether NVDA opts to offer two SKUs for Rubin Ultra (2 chip and 4 chip versions (the 4 chip version likely using INTC's EMIB-T)).

Of course the note doesn't guarantee that NVIDIA will use Intel's EMIB-T technology, as not only does UBS only speculate at the possibility but other reports have suggested that factors such as substrate production will end up determining whether the technology can be scaled up to meet the requirements of modern day semiconductor manufacturing.

Not only NVIDIA, but Google is also believed to be interested in the EMIB-T technology for its AI chips. These chips, called TPUs, can be packaged through EMIB-T if the technology's yield is sufficient, believes well-known technology analyst Ming-Chi Kuo.

Ramish Zafar Photo

About the author: Ramish is a seasoned technology writer and editor with more than a decade of experience. He specializes in semiconductor fabrication and market analysis. With a background in finance and supply chain management - via his bachelors in Finance and a micromasters in supply chain management from MIT - Ramish combines financial rigor with deep industry insight to deliver accurate and authoritative coverage.

Follow Wccftech on Google to get more of our news coverage in your feeds.

Button