Intel: Major Nvidia Foundry Endorsement (NASDAQ:INTC)

Investment thesis

Nearly a year ago, in an article titled “The Empire Strikes Back“, I detailed new CEO Pat Gelsinger’s master move to start a foundry business as a major pillar in his IDM 2.0 strategy to put Intel (NASDAQ:INTC) back on the map and silicon back in Silicon Valley.

In summary, if there has been any business unit within Intel that has adhered to Gelsinger’s “torrid pace” mantra, it is Intel Foundry Services, given the incredible momentum this business has gained in just one year. This first year of IFS has recently been capped off with a major endorsement from Nvidia (NVDA) CEO Jensen Huang in the wake of GTC.

Background

For a short timeline:

• March 23 (2021): announces IFS during “Intel Unleashed” backed by a $20B new Arizona fab investment, also open to landmark x86 (and other Intel) IP licensing
• April 23: within one month, already engaged with “well over 50 customers”
• July 22: engaged with over 100 potentials customers, first packaging customer (Amazon (AMZN), first revenue in Q4)
• July 26: announces roadmap to reach parity in 2024 with 20A and leadership in 2025 with 18A, announces Qualcomm (QCOM) endorsement for 20A during “Intel Accelerated”
• August 23: wins RAMP-C to establish U.S.-based foundry for the government
• September 14: announces IFS Accelerator to move automotive customers from trailing edge nodes to Intel 16
• February 7 (2022): announces IFS Accelerator Ecosystem Alliance to bring IP, EDA tools and services to all IFS customers, announces $1B start-up investment fund in collaboration with Intel Capital aimed at chiplets and RISC-V
• February 15: announces Tower acquisition ($1.5B revenue) to add trailing edge, analog and specialty nodes to the IFS portfolio
• February 17: announces over 40 test chips going through the fabs in 2022, including the first 18A ones, announces Cisco (CSCO) endorsement, announces H2’22 announcement for another new major foundry customer, says it is engaged with 5+ “big whales”, targets potentially ~$9B revenue by 2026 (although not an official target, it could be calculated based on a provided slide), targets at least 10% but potentially 30% in government capex offsets
• March 2: announces UCIe chiplet standard (equivalent of PCIe at package level) in industry collaboration to boost chiplet adoption and ecosystem
• March 23: receives Nvidia endorsement

In addition, in January and March, Intel respectively announced further $20B investments each (with potential to become up to $100B each) for new fab sites in Ohio (“Silicon Heartland”) and Germany (“Silicon Junction”). Obviously, these fabs will be used both for Intel’s own as well as its foundry capacity requirements.

Lastly, as a reminder, investors should note that foundry design cycles can take years to go from design start to volume ramp. This means on the revenue side, IFS will take a while to ramp. However, as I have discussed previously, this also means that Intel has already been providing its (potential) customers with PDKs for its 18A process technology, which Intel claims it will regain process leadership with.

In other words, the IFS business is not in any way impacted by Intel’s current lack of leadership, since Intel will have regained leadership by the time IFS revenue starts to ramp (as indicated by Intel’s unofficial $9B target for 2026), at least if it executes its roadmap successfully.

Nvidia endorsement

Nvidia CEO Jensen Huang made the comments as part of a press Q&A in the wake of GTC and its investor meeting, as has been reported on by Tom’s Hardware. While not a formal announcement yet, Huang made it clear Nvidia is interested in Intel’s foundry offering.

“Our strategy is to expand our supply base with diversity and redundancy at every single layer. At the chip layer, at the substrate layer, the system layer, at every single layer. We’ve diversified the number of nodes, we’ve diversified the number of foundries, and Intel is an excellent partner of ours[…]. They’re interested in us using their foundries, and we’re very interested in exploring it,” said Huang.

“I am encouraged by the work that is done at Intel, I think this is a direction they have to go, and we’re interested in looking at their process technology. Our relationship with Intel is quite long; we work with them across a whole lot of different areas, every single PC, every single laptop, every single PC, supercomputer, we collaborate.” Huang said.

“Foundry discussions take a long time, and it’s not just about desire. We have to align technology, the business models have to be aligned, the capacity has to be aligned, the operations process and the nature of the two companies have to be aligned. It takes a fair amount of time and a lot of deep, deep discussion – we’re not buying milk here. This is really about the integration of the supply chains. Our partnerships with TSMC and Samsung in the last several years are something that took years to cultivate. So we are very open-minded to considering Intel, and I’m delighted by the efforts that they’re making.”

Co-opetition

What’s actually ironic about this whole endorsement of Nvidia potentially using Intel to manufacture its GPUs is that last August, Intel made a lot of fanfare about its collaboration with TMSC (TSM) for manufacturing Intel’s Arc Alchemist discrete gaming GPU on N6. Late last year, Pat Gelsinger even went personally to Taiwan. As Intel said, making use of foundry capacity is also included in Intel’s IDM 2.0 strategy for silicon wafer capacity. Obviously, for those who remember, this was also partly necessary given the previous process delays: Intel gave its own design teams greenlight to make use of the best process technology they could find both inside and outside Intel.

Hence, the magic word that has been used is co-opetition. Intel will be competing against TMSC with the new IFS business, but they will also still collaborate as Intel remains an important customer for TSMC.

Similarly, it now seems that while Intel will be fiercely competing against Nvidia in both the GPU and AI segments (in the PC and data center), and Nvidia will be competing against Xeon with its own Grace CPU, nevertheless, both companies will also continue to collaborate. Potentially as far as Nvidia using Intel’s fabs.

We have been working closely with Intel, sharing with them our roadmap long before we share it with the public, for years. Intel has known our secrets for years. AMD has known our secrets for years. We are sophisticated and mature enough to realize that we have to collaborate.[…] We share roadmaps, of course, under confidentiality and a very selective channel of communications. The industry has just learned how to work in that way.”

“On one hand, we compete with many companies, but we also partner deeply with them and rely on them. As I mentioned, if not for AMD CPUs in DGX, we couldn’t ship DGX. If not for Intel CPUs and all of the hyperscalers connected to our HGX, we wouldn’t be able to ship HGX. If not for Intel’s CPUs in our Omniverse computers that are coming up, we wouldn’t be able to do the digital twin simulations that rely so deeply on the single-threaded performance that they’re really good at.”

TSMC, Samsung implications

Although Samsung also has a fairly significant foundry business, obviously, the elephant in the room is TSMC. For example, if Intel reaches its $9B revenue target, then that is obviously up to $7.5B (plus the $1.5B Tower revenue, although Tower expects to grow its revenue also further by 2026) in revenue that Intel must have taken away from those two foundries.

Even worse, I have previously suggested that the foundry space is a “winner-takes-most” market, since every company simply wants to use the best transistor out there. So if Intel regains technology leadership, then there is in principle no reason for Intel not to continue to grow further at the expense of TSMC (and Samsung).

Nevertheless, investors must be realistic, so there are two factors offsetting a mass exodus away from TSMC. The first one is simply inertia. For example, one could consider how Intel has been able to fend off AMD (AMD) in the data center with inferior products for the last few years.

Secondly, the overall semiconductor market is expected to continue to grow significantly over time, including the foundry segment of this space. For example, the current consensus is that the industry overall might double from 2020 to 2030 to $1 trillion. So while the leading-edge foundry pie will be shared by three players going forward, the pie itself is getting bigger.

Risks

At this stage in the development of the IFS business, building the foundry ecosystem and filling the pipeline with customers are the most important priorities. As detailed in the timeline above, Intel has made significant progress in both areas over the last year.

Nevertheless, as the bears will say, just being in talks with potential customers doesn’t say anything about actually converting this to major chip wins. Still, as discussed, all signs are currently bullish that this could already grow to a multi-billion dollar business by around 2026 or sooner.

Investor takeaway

Simply put, if even Nvidia – which surely must be one of the most demanding foundry customers out there – is “delighted” and “very interested” in using Intel Foundry Services, then who wouldn’t? Many folks on the internet are still discussing about whether Intel could ever catch up, but Huang’s lack of comments on that topic suggests that he sees Intel as an equally credible alternative to TSMC.

Hence, I would suggest investors to take Jensen Huang’s comments over anything they might elsewhere on the internet read about Intel. For example, recently I detailed how Intel is investing in technology leadership across several key technologies including process technology: Intel Stock: Investing To Regain Tech Leadership. As mentioned above, given the long time scale involved for a chip design to ramp in volume, by the time any potential foundry win starts to ramp, whether from Qualcomm, Cisco or even Nvidia, Intel will have solved its previous issues (unless new ones would come up, but that risk goes for any foundry).

To conclude, quite literally, the first question Intel received when it announced IFS a year ago was that Intel had tried to become a foundry before. Pat Gelsinger replied that IFS would be nothing like Intel’s previous attempt. After just one single year, Pat Gelsinger has already very convincingly proved that statement with various proof points, with the latest being to lure perhaps Intel’s biggest competitor to take a serious look at this business.