Barron's Streetwise - Intel CEO Aims for a Quadruple

Episode Date: July 30, 2021

Pat Gelsinger says he can return the chipmaker to glory by 2025. Jack examines his plan. Learn more about your ad choices. Visit megaphone.fm/adchoices

Calling all sellers, Salesforce is hiring account executives to join us on the cutting edge of technology. Here, innovation isn't a buzzword. It's a way of life. You'll be solving customer challenges faster with agents, winning with purpose, and showing the world what AI was meant to be. Let's create the agent-first future together. Head to salesforce.com slash careers to learn more. Intel is the company that put silicon into Silicon Valley. And I'd want all of the Barron's readers to know that Intel is back. And we're that innovative company that you're going to look at and say,

wow, that was an incredible few years. Hello and welcome to the Barron Streetwise podcast. I'm Jack Howe. The voice you just heard, that's Pat Gelsinger. He's the new CEO of Intel, and he's talking about his plan to put Intel back on top in chip-making technology by 2025. But hang on, if Intel still has enormous market share and generates billions of dollars in

cash each year, why did it lose its technology lead to begin with? And why will getting it back take years? And what does that mean for the stock? We'll talk about all of that coming up. I might try to slip in some terms like node and extreme ultraviolet lithography, slip in some terms like node and extreme ultraviolet lithography, maybe even ribbon fet architecture. I'm going to try not to hurt myself, but it could go either way. Decades ago, I think it was the 1990s, I assembled a computer and it wasn't because I had any particular hardware smarts. It was because I wanted to save a few bucks. Home computers at the time came in hulking towers. Components were designed to be easily swapped in and out. So all I had to do was look at a website that published monthly lists of the best parts and then order those parts and plug them into each other.

I used a website called Sharky Extreme. There were no sharks, and come to think of it, it wasn't particularly extreme. Anyhow, the main decision to be made was which central processor to use. If you can afford it, go with Intel, the website said. If you want to save 50 bucks, buy an AMD chip. Intel was the industry leader and usually had the fastest chips, sometimes by far. It introduced a logo in the 90s that said Intel Inside to get even non-techie computer buyers to prefer machines made with Intel chips. computer buyers to prefer machines made with Intel chips. AMD back then was small and scrappy, and once in a while came up with a chip to rival Intel's best, but those moments were typically fleeting. I ended up using an AMD chip, again, to save cash. Everything worked fine. There was one other big chip decision to make, by the way, and that was whether to use a graphics processor made by NVIDIA or ATI. Graphics processors determine whether you could play the latest video games.

I can't remember which one I used. ATI ended up being bought by AMD in 2006. We'll say a little more about graphics chips later. So where do things stand now? Raymond James, the investment bank, estimates that Intel this year will have an 85% chip market share in client computing, including laptops and other home machines. Now that's a dominant share, but it represents a decline of seven percentage points in just two years. And AMD has been rising just as quickly. Intel is facing its own fair share of challenges, including increased competition, manufacturing delays, and a slowdown in PC demand. The trend is similar in Intel's chips for servers.

Sharky Extreme, my old review site, it's gone. Back then, there were two upstart sites with much more technically detailed reviews called Tom's Hardware and Enantech. And today, they're widely followed by hardware enthusiasts. Some of the stuff is over my head. If I want to know what those sites think about something, I sometimes just skip to the conclusion part of the review. Back in April, Tom's published a review of

Intel's latest server chip. It said, although it, quote, doesn't lead in all metrics, it removes some of Intel's more glaring deficiencies. And Antec wrote that the chip, quote, narrows the performance gap significantly. However, the gap still remains and is still undeniable. However, the gap still remains and is still undeniable. That's tough love. So how did Intel come to be still ahead on market share, but playing from behind on technology? Here's someone with thoughts on that. The areas that we fell down were some of our process technology areas where we stumbled.

And as a result of some of those stumbles, some of our products fell to not being the leaders in their category. That's Pat Gelsinger. He took over as CEO of Intel in February. And this past week, he introduced a plan to turn things around. Intel was too arrogant on some aspects. And when you're leaders, as long as we were, you know, that often becomes insular and leadership. And we're breaking that down very rapidly. So what does Pat mean when he says arrogant and insular? Here are some examples. Intel declined to use an emerging chipmaking technology called Extreme Ultraviolet Lithography, or EUV. It uses exceptionally small light wavelengths to print many more circuits on silicon wafers than you could print with

traditional lithography. EUV manufacturing has been one key to the rise of rival chipmaker AMD. Intel also bet heavily on big manufacturing leaps, which put it at risk if those leaps didn't pan out. Other chip makers made more frequent incremental gains. And Intel might have been slow to respond to a power shift toward foundries. There are companies that design chips and ones that make chips, and Intel is both. AMD is primarily a chip designer. Companies that make chips for outside designers are called foundries. The most prominent of these is called Taiwan Semiconductor Manufacturing. A semiconductor, by the way, is a material that conducts electricity better than insulators, like glass, but worse than true conductors like copper. The semiconductor silicon is the bedrock of chip making.

In practice, the term semiconductor is used to refer to the materials, the finished chips, and a lot of things in between. You know how those blue cartoon characters called the Smurfs used to say, I'm going to Smurf my way over to the Smurf for a pack of Smurfs. Well, that's kind of how the chip industry Smurfs the word semiconductor. Anyhow, Taiwan Semi is no mere order taker. AMD couldn't have pulled ahead of Intel on chip specs if Taiwan Semi hadn't taken the lead from Intel in chip manufacturing. And you can see that power dynamic by looking at profit margins. Operating margins for Taiwan Semi are about double those of AMD. Intel has

been battling both of these companies on its own, which is maybe part of what Pat meant by Insular. There are also a couple of years-long industry trends that haven't made Intel's work any easier. Computing power has shifted to the cloud. I'd never try to assemble my own computer today, partly because I only use a laptop, and partly because I'm not as price sensitive as I was in the 90s, and because Barron's pays for my machines anyway. But also because I don't really care about computing power. I have more than I need.

Just about everything I do happens online, on distant servers. So much of the heavy lifting for processors is someone else's problem. That's not the case for everyone, but it is for a lot of us. So computer upgrades have slowed, cutting into Intel's chip sales. Now that trend has reversed during the pandemic, because when workers left their offices for their homes, many said, I need a new machine. But it's unclear how long the upgrade spurt will last. The good news is that the cloud will remain a rich source of chip demand for many years to come. But some of the advanced applications there, like artificial intelligence, thrive on what's called highly parallel processing, not unlike the kind used to bring video games to life through screen

pixels. Remember NVIDIA, one of my two choices for graphic processors back in the 90s? Today, it's making a killing in the cloud. If you put it all together, Wall Street's assessment of Intel's competitive threat is crystal clear. Ten years ago, Intel had a stock market value of $118 billion, and that was $40 billion more than Taiwan's Semi, NVIDIA, and AMD combined. Today, Intel's market value has risen to $215 billion, but those other three have reached a combined value of $1.1 trillion. Intel has bought back stock and paid dividends over the years, so 10-year stockholders have made 220%, which sounds great until you factor in the backdrop. They could have done nearly 100 points better in the S&P 500 index

or more than 680 points better in the Philadelphia Semiconductor Index. And that is enough history. So what's Pat going to do about it? We bet against this new lithographic approach called EUV, and now we're embracing it whole scale. We're then accelerating our innovation engines and getting our products to be more competitive. But being competitive ain't good enough, right?

We have to be leaders. This past week, Pat introduced a new roadmap for nodes. I'm going to call it a node map. Now, if you've ever had a doctor squeeze your nodes to tell if you're sick, those are totally different nodes, and I hope you're feeling better. In chip making, nodes refer to manufacturing generations. They used to be named for the lengths of transistor gates,

and chip leaders could brag, mine's smaller than yours. But over the years, chip architectures have gone from flattish to more three-dimensional, similar to how Manhattan builders have long squeezed more rent out of their lots by building up. Nodes named for gate sizes have lost their meaning. Intel's latest chips are called 10 nanometer, but starting later this year, it will drop the nanometer part and call its next node Intel 7. After that will come Intel 4, then Intel 3. It feels like the next one should be called 2, but in fact it will be called 20A, followed by 18A. Intel says those names are meant to signal the arrival of the Angstrom era. An Angstrom is one-tenth of a nanometer, and that makes it sound like we're back to measuring

things, but we're not. 20A and 18A are just marketing names. If that all seems clear, you might be an android. Ask your doctor to check your nodes. For now, let's put aside the names. Intel will roll out five new nodes through 2025, and that's an unusually fast pace by its standards. It seems to be moving toward more frequent incremental improvements. Its use of EUV lithography begins with next year's chips. By 2024, Intel will introduce two sweeping changes. One's called Power Via, and it will separate the way power and signal flow through transistors to get better results on both. The other is called Ribbon FET, and it represents

Intel's first comprehensive change to its chip architecture in more than a decade. And I'd explain it to you, but I'm no Sharky Extreme. Bottom line, if Intel can stick to its node map, it could catch up with its rivals on chip technology by 2024 and pass them in 2025. So can it succeed? Here's Pat. It is an aggressive pace, and I would say it's sort of, wow, right? Can we really do that? And what we've done to accomplish that is I've given that team the resources. I've given them the R&D teams, the capital resources to be running more of these experiments. We've also leaned into the industry relationships as Intel has never done before, and in particular, companies like ASML. That name is super important, ASML.

We'll get to it in a moment. ASML. We'll get to it in a moment. Pat is well-liked at Intel, and he's been rehiring veteran engineers who left the company, which is an excellent sign. He says he's open to turning over manufacturing work to outside foundries. Meanwhile, he's building a foundry business of his own that will serve outside chip designers as well as Intel. Taiwan Semi, he says, has proven you can make good profit margins in that business. So to that end, Pat is building two new chip factories in Arizona. Those don't come cheap. They'll cost $20 billion, which isn't part of Intel's regular capital spending budget and which doesn't include chip making equipment. Intel is also reported to be in talks to buy a chip manufacturer named Global Foundries.

The price cited is $30 billion. That would give Intel a jumpstart on its foundry plans. Pat isn't spilling the beans either way on that deal just yet, but he did tell me this. This industry is an expensive industry, and that's the recipe for consolidation. So we do think that there will be consolidation over time and we will be a consolidator. The foundry move is interesting. It's a way to get other chip makers to help fund Intel's manufacturing gains if it can bring them in as customers. Intel recently announced foundry deals with Amazon and Qualcomm. Also, the federal government is keen to return the U.S. to a more prominent position in chip manufacturing. There's a dire shortage of

chips at the moment that's holding back the manufacture of key American goods, like vehicles. So the semiconductor industry facing some major setbacks, the growing chip shortage now halting production at auto factories. You could find a gold mine in the back of your garage. J.D. Power reporting this week that used car prices are at an all-time high. If you've had trouble getting a new vehicle you want or have seen sky-high prices on used vehicles, blame the chip shortage. Politicians on both sides increasingly view chip manufacturing as a national security issue. They're working towards freeing up tens of billions of dollars for

domestic chip research, and Intel, with its new Arizona plants, could be a key beneficiary of that spending. Here's Pat. In 1990, the U.S. was 37% of supply and Europe was 44%. Today, the U.S. is 12% and Europe is 9%. And as we've seen through COVID, wow, that's pretty dangerous. In a decade, I believe that we can be on a path or maybe even up to 30% in the U.S. and 20% in Europe. Pat says that would bring economic benefits for the US. These are great manufacturing jobs. When we build a fab, a city emerges around it with our suppliers and all the jobs and schools that emerges around it.

By the way, I asked Pat about the chip shortage. He says it will take a year or two to clear and that the worst effects will hit this quarter, which runs through September. And if we think about it, semiconductors were growing at maybe 5% or 6% per year. COVID has driven that to 20% per year. As you saw, this huge shift to digital and the supply chains dropped to zero. So that gap, it's going to take us a while to get out of it. The auto industry, it's going to take us a while to get out of it. The auto industry,

you know, it's just politically untenable. A $30,000 car is waiting on a few dollars of semiconductors. Manufacturing lines are being idled. How do you deal with that? Speaking of cars, Intel bought a company called Mobileye back in 2017, which puts it in the race toward autonomous driving. I asked Pat, when will I be able to hop into the driver's seat and eat a sandwich with both hands on the way to my destination? And he said, it depends where you live. We just launched our New York City pilots two weeks ago. So now you can hop into one of our test vehicles in New York City, So now you can hop into one of our test vehicles in New York City, Israel. I had a bunch of demo trips myself a couple of months ago.

It was in Munich. We'll be launching in Paris. So maybe 2025 for you fancy-pants city slickers. If you live out in the woods with the coyotes and wild turkeys like me, Pat says general availability could take until later this decade. I mean, I'm not that remote. I live an hour north of New York City, but the other day I saw a big fat turkey sitting on a skinny branch in my woods. I didn't even know turkeys could get up that high. What was I talking about? Right, chips. Intel stock trades at 11 times this year's earnings forecast, which makes it half as expensive as the broad stock market.

But we don't yet know how much Pat's new plan will cost. Now, we'll likely learn more about that when Intel hosts an analyst presentation in November, but we might not know about the returns on that spending for years. And the fact that it will take years is a reminder of Intel's performance deficit today. That's where the bulls and bears disagree. They both think Pat's plan makes sense. But the bears say it will cost too much and take too long and that the outcome is uncertain and that Intel will lose more market share between now and then. will lose more market share between now and then. The Bulls say a lot of that stuff is already factored into the share price and that if Intel can make some progress, like stabilizing its market share and meeting production targets, its stock will respond quickly. One way Intel can hold

on to market share is by competing on price, which might not be great for margins. But also, as Tom's Hardware puts it, market share gains for Intel's competitors haven't been as swift as past performance leads would have suggested. And that's because buyers for big organizations and data centers are highly risk-averse, prizing factors beyond price-to-performance ratios, including compatibility, security, reliability, serviceability, and engineering support. Again, that's Tom's. If you'd like to debate Intel's appeals in those areas, you should.

But as much as I'd love to hear from you, I might not be the best debating partner because I haven't held a chip since maybe Titanic came out. The movie, not the ship. Also, I'm not going to pound the table and tell you whether to buy or sell Intel stock because as much as I like Jim Cramer, I'm not him and this podcast isn't the lightning round. It's more like the lager round. I can't tell you that just 37% of analysts who cover Intel shares say to buy them. I asked Pat, what do you think you have to do to change the minds of the bears?

He says he needs to rebuild momentum in data centers and demonstrate that the surge in his personal computing division can last beyond the pandemic. And he needs to show good performance in Intel's fast growth businesses like advanced computing and autonomous driving. I think as that story starts to sink into the marketplace, the bears will start to, one by one, we're knocking down their bear thesis. The bulls are, they're wanting, they're just betting on when and how that turn occurs. And as we start seeing those much more competitive positions in the marketplace, things will swing. Pat says he's confident and he isn't shy about what he sees as the potential upside for investors.

But this takes a while and I'm not at all worried about it. The internal teams, they feel the mojo coming back. And we're setting a course not to do 10% or 15% better in the market, but to triple, quadruple the value of this great, iconic company. That's the path that we're on. I just want to touch on one last thing. I said I'd come back to ASML, which is a Dutch company that makes the EUV machines used by companies to make chips. Remember, EUV is the technology that Intel resisted at first and is now embracing. It's a big key to the recent performance gains for Intel's rivals. Taiwan Semi is a major customer for ASML.

That raises the question of whether Intel will be able to get enough EUV machines. Pat tells me he will. Yeah, the early parts of the roadmap that we described, the Intel 7, 4, 3, and then Intel 20A and 18A, the early parts of those, we have the machines that we need, so we feel comfortable with that. We're now working through the full build-up requirements to satisfy those, so we feel very comfortable that we're in good shape for the next several years. It's a crucial detail. Here's how Quinn Bolton, an analyst at Investment Bank Needham, who is bullish on Intel, describes the situation.

ASML is in a very enviable position with its EUV lithography systems. They're near monopoly or the closest thing you can get to a monopoly. Supply of EUV systems this year, probably even in the next year, is capacity constrained. And so there's a scramble between Samsung TSMC and Intel to try to secure those EUV systems. Now, I'm definitely not going to get back into Intel's node map, but just know that in 2025, when it gets to the second of its Angstrom-era nodes called 18A, Intel plans to switch to ASML's successor to EUV technology. It's called HiNA EUV. The NA stands for numerical aperture, which I'm sure means something wonderful. The key is that high NA EUV is expected to be the next big thing in chip manufacturing. And here's what Pat tells me about getting those machines. We also have a contractual relationship with ASML that we will

be the first production user of those tools as well. So not only is it to catch up, deliver capacity, but then to establish a leadership. And ASML is an important partner for us to accomplish that. Think about that. If Pat will get dibs on the next generation EUV machines in 2025, that means other manufacturers won't. And that could help give Intel a competitive advantage. Other manufacturers won't, and that could help give Intel a competitive advantage. Of course, that's a long time for investors to wait. And if you're thinking, why not just buy shares of ASML since chip makers are scrambling for its machines? You're not alone.

It's been an incredible stock market performer, and the shares are now ambitiously priced at 48 times this year's earnings forecast. If that's too expensive for you, Quinn at Needham might have good news. Other chip makers also stand to profit from this battle of the node warriors. You can't just buy the EUV. You need a lot of other equipment from the other manufacturers. And I think what you're seeing just in the last couple of years is the capital intensity of semiconductor manufacturing is clearly moving higher. You need etched deposition equipment from LAM research and applied materials. You need the process control equipment from KLA to make sure that you've got high manufacturing yields on the EUV equipment. Among those three, applied materials, LAM research, and KLA, Quinn is most bullish on applied materials. I wrote favorably about it for Barron's at $24 a share five years ago, and at $61 a share a year and a half ago,

and recently it was $136. Yippee, and I wish I could tell you they all worked out like that, despite its run-up, applied trades at around 20 times earnings, which is a touch cheaper than the broad stock market. And I think that's everything. You see, now I feel like it's an awkward ending, but maybe if I just stop talking and Jackson brings that music up, it'll sound like a satisfying conclusion. I think it's working. Definitely working.

I'm going to stop talking. I'm satisfied. Thank you for listening. Jackson Cantrell is our producer. Subscribe to the podcast on Apple Podcasts, Spotify, or wherever you listen to podcasts. If you listen on Apple, please write us a review. And if you want to find out about new stories and new podcast episodes, you can smurf me on Twitter. That's at Jack Howe, H-O-U-G-H. See you next week.