Intel's Foundry Eyes NVIDIA and Broadcom as Clients for Future Growth

[AleksandarK]

According to an investment bank UBS note, two industry titans—NVIDIA and Broadcom—are potential future clients that could significantly enhance Intel's Foundry business revenue. To revitalize Intel, newly appointed CEO Lip-Bu Tan reportedly aims to forge strategic alliances with two AI chip manufacturers. Tan, who assumed leadership earlier this month, is determined to rebuild the company's reputation by focusing on customer satisfaction and accelerating the development of its foundry business. UBS analyst Tim Arcuri suggests that while Broadcom might join the client roster, NVIDIA appears to be the more likely candidate. Rather than initially manufacturing NVIDIA's AI GPUs, Intel is expected to begin production with gaming GPUs. NVIDIA could even move to AI GPU production at Intel's fabs if satisfied.

Despite some early optimism, Intel's new CEO is now committed to addressing issues related to power consumption in Intel's manufacturing processes. UBS analyst Tim Arcuri noted that the firm is pushing hard to introduce a lower-power version of its 18A process, the so-called 18AP, which has reportedly struggled to meet energy requirements. Additionally, Intel is working to improve its advanced packaging techniques to rival Taiwan's TSMC CoWoS (S/L/R variants) technology, aiming to overcome packaging constraints that have slowed AI chip production. Analysts speculate that Intel might also become a secondary supplier to tech giant Apple. A promising partnership with Taiwan's United Microelectronics (UMC) could pave the way for Intel's chips to find their way into future Apple products. Whatever materializes, we are yet to see. Switching foundries from TSMC to Intel entirely is not possible for any of the aforementioned fabless designers, so it will likely be dual-sourcing at first, with some non-flagship SKUs getting the full port to Intel 18A.



View at TechPowerUp Main Site | Source

 

[Vayra86]

This could fail spectacularly.

I can totally see the headlines where the performance between chips from these two different sources is indeed, different. And what if you start finding an Intel chip in what you expected had a TSMC chip? Going to be interesting to see how they get around that. Product segmentation...

 

[Squared]

"...the firm is pushing hard to introduce a lower-power version of its 18A process, the so-called 18AP..."

It's almost as if 20A wasn't canceled but rather delayed and renamed 18A, and now "18AP" is coming to be and it is what was originally going to be 18A. 20A was a shiny new node for a small client tile and 18A offered density, efficiency, and yields needed for larger server tiles. Now "18A" is going into a client tile (in Panther Lake), and "18AP" has come into the news before any other 18A products have, suggesting it will be used for the next server tiles.

 

[Tomorrow]

Never gonna happen. If Intel is not making their own much smaller and less complex 5nm GPU's in their own fabs they how on earth does anyone even hope that they can make 4nm or smaller ~800mm² behemoths with complex packaging in those same fabs?

It's the same delusional rumor that pops up from time to time about Nvidia using Samsung. As if Samsung had any decent yielding nodes to offer...

 

[Squared]

Never gonna happen. If Intel is not making their own much smaller and less complex 5nm GPU's in their own fabs they how on earth does anyone even hope that they can make 4nm or smaller ~800mm² behemoths with complex packaging in those same fabs?

It's the same delusional rumor that pops up from time to time about Nvidia using Samsung. As if Samsung had any decent yielding nodes to offer...

The RTX 5080 die is less than 400 mm². AMD's biggest GPU this generation has a slightly smaller die. And Intel's Granite Rapids tiles are thought to be around 500 mm², and those are 5 nm-class Intel 3 tiles and each CPU has 3 of them.

It does raise the question: why is Battlemage on TSMC N5 if Intel 3 is doing well and competes well? An answer charitable to Intel is that perhaps Intel is still learning how to manufacture a GPU; all Intel nodes have historically been intended firstly for CPUs which tend to use high-frequency transistors in manually-drawn layouts, whereas GPUs need high-density low-leakage transistors probably in computer-drawn layouts. High-density libraries are a feature of Intel 3 and not Intel 4 and are something Intel has been trying to improve on in order to gain customers.

There is also EUV machines; Intel has fewer of them than TSMC and Intel 4, 3, and 18A all need them, which limits production capacity. That's improving as Intel receives more EUV machines.

 

[R0H1T]

I can totally see the headlines where the performance between chips from these two different sources is indeed, different.

Pretty sure they "will be" different, Apple did this with A10 or A9 IIRC, and of course different modems that one time with Intel/QC maybe pitting them against each other for better price?

The only way to make them perform "exactly" the same is to adjust GPU core/mem clocks on those chips!