Intel’s Goldman Sach’s Meeting Yields Interesting Information

Summary:

• Intel’s upcoming 18A process is cost-effective and more efficient than Intel 7, which will significantly improve the company’s competitive position.
• Intel’s AI products do not look compelling in the short term.
• Intel continues to execute well on their products, something that had been problematic before Pat Gelsinger took over.
• Intel’s financial turnaround is still not imminent, and there will still be short term issues before the company’s bets pay off.

On May 30th,

John Pitzer from Intel (INTC) met with analysts to discuss Intel’s situation, and the information was really quite interesting, and can give us a better idea of what is going on. I am going to go through the points I think are interesting, and go into what I think they mean.

I want to preface this article with something that I received feedback on with my prior article. There’s a lot of things I covered that are negative with regards to Intel, and there will be in this article too, yet I still view them as a strong buy. My purpose is to give you the facts, and how I view them, not to make you believe Intel is a strong buy. I have faith in the company, still, and see solid, if choppy, progress, and enormous potential. I believe it has a very strong long term strategy, that should manifest itself into very significant share price improvement. But, there’s uncertainty, and problems that could make someone believe otherwise.

So, let’s go over some of the highlights in the information provided by John Pitzer.

Good and Bad News About Intel Processes

Let’s start with what wasn’t said. In a phrase, Intel 3. They are ramping Sierra Forest, which is a small core server processor, and Granite Rapids will ramp next quarter. Both will be on the Intel 3 process. That’s it. He didn’t mention external customers, but he did mention Intel would be using more tiles from TSMC as we head into 2025, turning around in 2026. While he mentions the 18A node will have more mobile friendly processes, he didn’t even talk about Intel 3. Intel 3 is a pretty irrelevant node, that will not help Intel appreciably, either internally or externally. So, we have to wait for 18A before any help is on the way from the foundry.

Gelsinger bet the company on 18A, and it looks like a good bet

On the plus side, and this is EXTREMELY interesting to me, and highly relevant to the company’s success, 18A will not cost significantly more per wafer than Intel 7. He mentioned the ASP per wafer is three times higher. 18A should be more than three times as dense as Intel 7, so I think it’s very conservative to say the cost per transistor will be around 40% of what it is on Intel 7. The performance should be at least as good (by all indications, currently Intel 7 is the best performing node), but the efficiency should be significantly better. By his remarks in the conference, 18A is a high performance node initially, so it probably will perform even better, but we have to wait for that. As mentioned above, Intel will also have mobile friendly processes on 18A in the future. It’s just an enormous improvement for the company, not only for the external customers, but also for the company’s internal products. It also gives Intel an excellent opportunity to pull back tiles from TSMC as the process ramps up.

Which is a good segue into some of the current products Intel has, and the near future products.

Let’s start with Meteor Lake, Intel’s current mobile processor known as Core Ultra. Rather surprisingly to me, since it’s far from a good part in most measures, Intel was supply constrained in Q1, though apparently that will no longer be the case in Q2. In fact, given Intel’s prior light guidance, it doesn’t appear this was resolved by increasing production as much as demand waning a bit. This is, at best, a mediocre part, having poor power characteristics, mediocre performance, and limited “AI” capability, despite having an NPU. Luckily, it will have a short life.

There has been some speculation whether Lunar Lake (Intel’s next mobile processor) will have a TSMC made CPU. Based on the remarks from Intel with regards to it being very power efficient, and given the percentage of tiles made at TSMC will be going up this year (and until Intel’s 18A process ramps up), it seems very likely this will be based on one of TSMC’s 3nm processes. The Arrow Lake CPU (Intel’s next desktop part), will be on Intel’s 20A process, by all indications. This makes sense, Intel tends to have better performance than TSMC, which matters more for desktop parts, but inferior power efficiency, which would not allow them to boast so much about Lunar Lake’s efficiency. While we have no official confirmation, yet, it is even more likely now that this is the case.

Intel 7 will be around for a while

One oddity is, Pitzer said that Intel will sell roughly 40 million AI PCs this year, and only 60 million next year for clients, saying that two-thirds of 2025 client PCs will not be AI PCs. Since even Meteor Lake is considered an AI PC part, and certainly Lunar Lake and Arrow Lake will be much more so, this indicates Intel expects to sell a very high quantity of parts based on their inefficient Intel 7 process, even well into next year. That is not exactly great news, given what we know about how expensive it is, but since it uses DUV machines, not EUV, I guess Intel needs to keep plugging along with it while they continue to transition to EUV. Intel 4 and later use EUV, although Intel 4 only uses it to a limited degree. So, the move from the inefficient and pretty close to obsolete Intel 7 node is going to take a while, and that’s going to weigh on the company in the short term.

AI does not look strong for Intel any time soon

The other bad news is Gaudi, in particular, Gaudi 3. Intel expects to make $500 million on it the last half of the year, which is miniscule compared to what NVIDIA is getting for its AI accelerators. Intel points to Falcon Shores as when they expect to have a real competitor to NVIDIA (NVDA) so once again, we get a scenario where Intel is not competitive, but just wait, and then they say they will be. Do you believe them? I have limited faith in Falcon Shores, which is a GPU and more programmable than Gaudi 3, but I think it’s unrealistic to think it will suddenly take the market by storm. Will it be more competitive? Almost certainly. But, there’s a lot more to gaining share in this segment than just the silicon parts, and developing the software stack and support around the chips will be extremely important. NVIDIA’s CUDA has been out for years, and is mature and the standard. Intel is playing catch up, and might see incremental gains, but it’s hardly realistic to think they will suddenly grab a lot of market share from NVIDIA in this segment, even with Falcon Shores. Clearly, Gaudi is more for getting customers using their parts, and continue to improve their services, software and support, while they ready Falcon Shores.

I have much more faith in Intel’s 18A, and its impact on the company, than I do Falcon Shores. It is necessary, and given NVIDIA’s enormous market share, it should gain some share, but it’s going to be a slow process. Even so, if you gain market share in a market that’s exploding, it can be significant. But, we know so little about it, and Intel really has not given us any indication of what they expect, so I am a little wary about putting too much emphasis on this as a driver for Intel earnings.

The Bad News

To summarize, the bad was with regards to AI, and Gaudi. We already knew the paltry $500 million figure for the rest of the year, but he made it clear Gaudi 3 was not a particularly important product, and yet again we had to wait for the real technology to come out in late 2025. Yuck. We also got a pretty good indicator that Lunar Lake and Arrow Lake will be using an awful lot of TSMC parts, especially the former, which doesn’t bode too well for the foundry, but might make for a compelling product. Then we got more information that the very inefficient Intel 7 will be quite relevant well into 2025.

The Good News

There was some good, though. Mainly, the 18A process is very cost effective when compared to Intel 7. This is extremely important, and in my mind, more important than all the negatives combined. While Intel did indicate in the past it was going to be more cost effective, to the extent it is, is surprising to me. Countering that is TSMC’s significant increases on every node, and Intel jumping two full nodes (forget Intel 4 and 20A, they aren’t full nodes) without a significant increase in cost is something quite exceptional. In addition to the far better density, Intel has made it clear it is much more efficient, and higher performing. This will fundamentally alter the competitive nature of the company’s products, given TSMC has a considerable advantage now, that minimally will diminish considerably, or be reversed (it probably will depend on the scenario and nature of the chip being made). That Intel will finally get some external customers on it will help as well, but perhaps more importantly, it will allow the company to stop giving money to TSMC, while still using a process that allows for superior processors.

It’s also becoming more and more clear that Intel is executing on their roadmap. It’s frustrating to always have to wait for the real competitive products, but we do see steady progress, so the promises do not seem empty. At the current discounted prices, the company seems a steal given the potential and effective recent execution. But, there’s still a wait before the bottom line sees a big improvement. But, when it does…

---

Analyst’s Disclosure: I/we have a beneficial long position in the shares of INTC either through stock ownership, options, or other derivatives. I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it (other than from Seeking Alpha). I have no business relationship with any company whose stock is mentioned in this article.

Seeking Alpha’s Disclosure: Past performance is no guarantee of future results. No recommendation or advice is being given as to whether any investment is suitable for a particular investor. Any views or opinions expressed above may not reflect those of Seeking Alpha as a whole. Seeking Alpha is not a licensed securities dealer, broker or US investment adviser or investment bank. Our analysts are third party authors that include both professional investors and individual investors who may not be licensed or certified by any institute or regulatory body.

---