Did Intel really lose?

Latest update time：2021-09-03 13:51

Reads：

Source: Content compiled from "seekingalpha", thank you.

Intel Corporation (INTC) shares plunged after releasing its second quarter earnings report. As some optimists have said, Intel's revenue grew 20% year-over-year to nearly $20 billion, outpacing Apple's (AAPL) growth and even saying "two months of digital transformation." However, Microsoft's stock price reacted differently, plunging 17% to $50 per share.

The reason for the plunge is the 6-12 month 7nm delay. Perhaps most worrying is that given the subsequent impact of this delay, the previous 14nm and 10nm delays will follow. If 7nm yields are already 12 months behind schedule, it is anyone's guess what yields will be in the next two years.

However, while it's not the fanciest series (as CEO Bob Swan expressed at the end of the call), it is what I call a "nano game" (similar to The Hunger Games). Many enthusiasts, analysts, and investors - most of whom don't have degrees in electrical engineering - are discussing this and making various doom and gloom predictions about the company's prospects.

As mentioned above, this isn't the first time we've seen this happen, as the saga of the 10nm program continues as it just caused Intel to report its worst gross margin quarter in 11 years. Despite the issues with 10nm, that didn't stop Intel from growing 35% over the past five years. So this leads to the following conclusion: I don't think 7nm will fundamentally change Intel's growth trajectory. For investors, this means that dips in the stock market could be opportunities to buy on the dip.

To sum up, there are several points worth discussing.

Separate finance and technology

As a freelance writer for Tom's Hardware, I cover Intel technology. While writing for Seeking Alpha, I've covered Intel stock.

These are two completely different things, and investors and tech enthusiasts should be aware of this. Intel completed the worst 10nm technology transition in history after 3 years of delays. However, as mentioned above, this did not stop Intel from delivering record quarterly and fiscal year reports quarter after quarter for the past five years.

So on one hand, when tech enthusiasts look at Intel, they compare it to AMD (AMD), talk about market share, mostly for the small DIY desktop market. They look at the technology, see that Intel is behind AMD on the 7nm equivalent node, and predict that Intel will fall behind.

But that's not how business works. Intel doesn't sell nanometers. Intel's customers are OEMs (most of the market is actually laptops) who demand annual CPU product refreshes to support their own product refreshes. Apparently the same is true in the server space.

Of course, Intel creates value for its customers by creating better CPUs, but there are more than one way to Rome (pun intended). To that end, using better lithography (better transistors) is just one possible means of improving the product.

So Intel is just sticking with annual CPU refreshes (Ice Lake, Tiger Lake, etc.), which will provide a lot of inertia for OEMs to switch to AMD. Even if AMD is 10-20% faster.

On the other hand, from an investor's perspective, Intel also has broader market dynamics at play. These dynamics can increase the size of the total market (TAM). In other words: having 90% of a $300 billion market will produce very different financial results than having 90% of a $30 billion market.

I'm personally looking at these dynamics as I complete my thesis on Intel. As demand for compute continues to grow from the edge to the cloud, the overall market is expanding. This is good for Intel, and likely good for AMD as it pursues these opportunities (but AMD doesn't have IoT, 3D NAND, etc.). Even if this results in some internal market share volatility, net-net Intel will grow, and the stock will (should) follow.

So, in the big picture, Intel is the leader in CPUs and will therefore benefit from these tailwinds (given its investments in these areas): Cloud, 5G, AI, Autonomous Driving, 3D NAND, Silicon Photonics, Optane. As I reviewed in my commentary some time ago, dinosaurs may be slow, but they are also hard to stop.

Let's take 5G, for example, an area where Intel has been moving very quickly and gaining a lot of market share. Simply being a TSMC 5N customer doesn't make you a leader in 5G, Intel can't be a true leader without that. So how do we do that? This may be redundant, but simply having manufacturing plants is completely meaningless if you don't have products (IP) to be manufactured in those fabs. So simply put, having great manufacturing technology is only one piece of the puzzle, there are many other pieces that can make up for this disadvantage. For example, 5G "software defined networking" is more than just a piece of silicon chip. On the PC side, Intel's actual architecture ("Ice Lake") is more advanced than AMD's ("Zen 2").

Another example is Optane. As Intel also admits, the first two years of new disruptive technologies will be slow, but momentum is clearly building.

But given recent events, investors are clearly not focused on the growth opportunities I described. To that end, while many investors have been vocal on the subject (e.g., the number of Seeking Alpha articles), this leads to a different problem: despite all they write, many analysts and investors are likely even less knowledgeable about process technology than (or even less so) technology enthusiasts.

Process Technology

So let's move on to process technology because that's a critical part of the nanometer game, and it's a game that's actually been played very well by the foundries.

As many people will like to hear, not all nanometers are created equal. Each foundry develops its own process technology and roadmap, which results in specific features and feature sizes. This entire set of technologies, the result of billions of dollars of R&D, and hundreds of process steps to create a complete “wafer,” hence the name. 10nm. 7nm. 5nm. 3nm.

None of this makes sense. These are marketing games. First, it doesn't touch on the actual size of any transistors, as it hasn't been done that way in decades (not since the GHz race of the 1990s... or maybe the equivalent of "GHz games").

Second, in addition to feature size, there is the issue of processing technology. There is also an equally important component called the material science component of the transistor. For example, as some people correctly speculated, at the nanoscale, quantum effects come into play, and they do.

So an example of both parts (nano marketing and materials science) is TSMC's (TSM) 20nm node (2014). A little over a year later in 2015, TSMC improved on that node by changing the transistors from planar to FinFET - a big change in itself, but the feature size related to transistor density did not change. So it was a refinement of the same technology, but TSMC decided to call it 16nm, as if it was a new node according to Moore's Law, even though in reality the density remained the same. So when TSMC introduced 10nm, it represented a full shrink from 20nm, even though its name (20->16->10) suggested it would be twice the more advanced shrink.

So this is still a good example of these "nanometer games" and so far it seems like TSMC and Samsung are way ahead of Intel: Intel just moved to 10nm, while TSMC is already moving to 5N. In reality, without the above marketing tricks, TMSC would really move from 10nm [=7N] to 7nm [=5N] this year. While yes, this is ahead of Intel, who will still be using 10nm technology this year (next year, and most of the time in 2022), what would people think if last year's headlines said that AMD would launch 10nm chips? In fact, I have seen many publications claiming that AMD already has 7nm chips on the market. AMD does not; at least, AMD does not have Intel's 7nm transistors on the market yet.

So, simply put, node names cannot be compared between foundries. While they all have a manufacturing process called 7nm, Intel's 7nm process is different from TSMC's (and Samsung's). So, given the marketing strategy above, Intel is actually making itself look worse by not adjusting its naming scheme. I would suggest that Intel (similarly) "skip" the 7nm name, rename its current 7nm to 5nm or even 3nm, or simply switch to another scheme.

To make cross-generational comparisons possible, Intel did introduce/popularize the transistors per square millimeter metric as an objective indicator of process transistor density in 2017. So Intel could probably call this metric "MT", with higher meaning better. For example, Intel's 10nm is better than TSMC's 7nm (or similar for practical purposes), even though Intel's is bigger ("worse"):

TSMC "7nm" = 90MT
Intel 10nm = 100MT
TSMC " 5nm " = 170MT
Intel 7nm = 200-240MT?
TSMC " 3nm " = 290MT
Intel 5nm = 400-480MT?

This means that last year AMD (actually) launched Zen 2 CPUs with 90MT, while Intel launched Ice Lake CPUs with 100MT. What did AMD lead?

By 2023, Intel might have 240 MTs and TSMC might have 290 MTs. This is by no means a game-changing difference. It just means AMD can fit a little more transistors per square millimeter. Does it matter? Consumer chips are not limited by silicon area. If Intel wants to squeeze more performance out of a chip, it can still make the chip bigger (with more transistors), although that will cost more to manufacture.

It’s also important to note that this number (even though it’s MT, referring to the materials science aspect of transistor engineering mentioned above) doesn’t tell you which transistor consumes the least power and/or reaches the highest clock speed. On that last point, Intel is (and has been) designing improvements inside the + and ++ nodes to bring the power and performance of the smaller nodes up to the previous nodes. So, aside from transistor density, Intel’s 10nm++(+) node coming out in 2021 (2022) will likely be just as good as the 7nm node. Again, where is the leadership of AMD and TSMC?

There are many other aspects to the "nano game". For example, a process technology usually has several "banks" with different transistor density, power, and performance. So, to prove its point, the density of the CPU is relatively unimportant: for Core, Intel used the lowest density HP bank (because the CPU can benefit greatly from higher clock speeds). From a transistor density perspective, using the HP bank is similar to using an n-1 node. So Intel's Ice Lake CPU actually has much lower density in a 100MT processor.

In summary, nanoscale in principle refers to the density of transistors (Moore's Law). In practice, it is a marketing indicator. Moreover, for CPUs, transistor density actually means relatively little, as its name does not say anything about the power and performance of said transistors.

Summarize

Nano Games:

Intel doesn't sell Nano, and neither does any other company.
Companies sell products, and products must be designed. (Intel’s process advantage was not the reason AMD was on the verge of bankruptcy five years ago.)
While 7nm delays are being discussed, the specifics remain fairly vague: For example, what exactly will be outsourced?
In fact, the first consumer 7nm chips could be available in less than six months (although likely in smaller quantities than originally planned).
The 7nm “defect mode” appears to be a 7nm-specific issue that does not impact 10nm or 5nm development (although it is certainly another execution issue).

Competitively, Nano Gaming:

Intel’s 7nm technical specifications haven’t even been made public yet. The sell-off is likely an overreaction due to investors’ misjudgment of its significance: AMD already has a “7nm process” on the market, which makes a delay to 2022-2023 sound pretty dramatic.
As it stands, Intel's 7nm (=200-240MT) will ship in 2023, and contrary to what it looks like, nominally it will not be significantly lower than TMSC 3N (=290MT) also shipping in 2023. At this point we don't even know if AMD** will be on 3N in 2023, since the mobile space is usually first.
As mentioned above, not all nanometers are created equal: nanometer is marketing in the name. TSMC transitioned from 20nm to 16nm in 2015, and there was no increase in density.

Based on Moore's Law, Intel's 7nm technology will be at least one generation ahead of AMD's 7nm technology (at least partially justifying its later arrival).

The name nanometer also does not mention the performance and power consumption of transistors

Intel's 10nm++(+) is likely to be close to its 7nm in these metrics.

We are currently in the FinFET transistor era, with Intel starting in 2012 and TSMC starting in 2015. As things stand, starting in 2024-2025, TSMC and Intel will transition to the nanowire (gate-all-around) era in relatively negligible time with each other.

*Since the production cycle of semiconductors is about 3 months and it takes some time to ramp up production, there may be no product launches in December, and 3N production in the second half of 2022 means product launches in the first half of the year. This is almost always the case.

**AMD also has two generations on 7N: Zen 2-3. Since Zen 4 (N5) looks to be early 2022, Zen 5 on N4 is likely early 2023, and Zen 6 is 2024 on N3.

Conclusion

TSMC is way ahead in the nanometer game: it is moving towards 5nm this year, while Intel continues to ramp up at 10nm. AMD already has market share at 7nm, Intel is just holding off. As a result, billions of dollars have changed in the market

But as I described, most of this is just marketing play. Intel's 7nm process is not TSMC's 7nm process. Intel's 7nm is at least competitive with TSMC's N3. Now, both nodes will start in 2023. If Intel's 7nm is close to 200MT, TSMC may have a 1.5x gap, but it will definitely not be as one-sided as Intel's 14nm and TSMC's 7N (the gap is only 2.5x).

Although Intel announced that the 7nm process will be delayed by 6-12 months, it also announced that it is willing to work with TSMC if TSMC has better technology at that time; Intel will use the best existing process. Therefore, if it is superficial, AMD will not have an advantage (in manufacturing technology) at any time in the future: they can both obtain TSMC's technology, and Intel can also obtain its own process and packaging technology.

For example, Intel-acquired Habana, Movidius, and Mobileye (which competes with Nvidia in areas such as AI) already use TSMC.

As I described, there is more to semiconductor products than manufacturing technology. There is intellectual property that is still being developed. For example, while AMD’s GPU has been moving to N7 for a while, it still doesn’t have a core like Nvidia’s Tensor (or Ray Tracing) that would make it competitive in AI workloads, which are driving Nvidia’s growth in the data center. So while the process can be a differentiator, so can the architecture of the product.

Even when it comes to process technology, it's not just about transistor density. For example, Intel says it intends to improve 10nm in line with Moore's Law with the introduction of 10nm+(+)(+). So Intel may have lower density, but transistor power and performance may be similar to AMD (or its 7nm). Intel could very realistically release/manufacture 7nm products or IP on 10nm++(++).

It may be unrealistic for investors to keep track of all of these technologies. To that end, I will make the following point: I think Intel will remain the leading CPU company (by size - financials - market share), while AMD will remain relatively small, and I don't see anything fundamentally changing.

As proof of the case: Zen 2 was supposed to be the big catalyst for AMD, as it currently has all the advantages that the 7nm delay could bring: the lead in this process is said to change the entire industry (no pun intended). With most of the news being about 7nm, we may have overlooked one point: Intel also announced that it will exceed expectations in market share in the data center this year, while also regaining share in the PC market.

So what's driving Intel's growth is its expanding opportunities in growing markets: Cloud, 5G, AI, Mobileye (ADAS, MaaS, and AVs), IoT, 3D NAND, Silicon Photonics, 3D XPoint. AMD simply doesn't have a competitive advantage in any of these opportunities.

Finally, as a technology enthusiast, I strongly criticized the delays - as I did in previous analyses. But from an investor's perspective, we cannot separate these developments from Intel's financial growth opportunities. For this reason, I bought some shares on the dip. In 10 years, when Mobileye's robotaxis are taking off and Intel is leading the nanowire transistor era, investors will be scratching their heads how Intel could have been valued so low.

★ Click [Read original text] at the end of the article to view the original link of this article .

*Disclaimer: This article is originally written by the author. The content of the article is the author's personal opinion. Semiconductor Industry Observer reprints it only to convey a different point of view. It does not mean that Semiconductor Industry Observer agrees or supports this point of view. If you have any objections, please contact Semiconductor Industry Observer.

Today is the 2398th issue of content shared by "Semiconductor Industry Observer" for you, welcome to follow.

Latest articlesabout

■SiC giant, rebirth: how to predict the future?

■Apple chips may hit Qualcomm hard

■Chip cost per car: soaring to $1,000

■TSMC 2nm, important information

■Huang Renxun's latest views

■The risks of this type of chips that are promising have increased significantly!

■NPU, how to see it?

■Storage giants are abandoning DDR 4

■Intel, why?