Hot topics close

The China challenge: Why Intel is too strategic to fail

The China challenge: Why Intel is too strategic to fail - SiliconANGLE

More than anything, Intel Corp.’s big announcement this week that it will shift its strategic direction underscores the threat that the United States faces from China.

The U.S. needs to lead in semiconductor design and manufacturing, and that lead is slipping because Intel has been fumbling the ball over the past several years. A mere two months into the job, new Chief Executive Pat Gelsinger wasted no time in setting a new course for perhaps the most strategically important American technology company.

We believe that Gelsinger has only shown us part of his plan. This is the beginning of a long and complex journey. Despite Gelsinger’s clear vision, deep understanding of technology and execution ethos, in order to regain its No. 1 position, Intel will need help from partners, competitors and, very importantly, the U.S. government.

In this Breaking Analysis, we’ll peel the onion of Intel’s announcement, explain why we’re not as sanguine as Wall Street on Intel’s prospects and lay out what we think needs to take place for Intel to once again become top gun — and for us to gain more confidence in Intel’s return to the top.

The urgent need for change

You may recall that in January we dismissed analysis that suggested Intel didn’t have to make any major strategic changes to its business. Rather, we said the exact opposite.

Our view at the time was that the root of Intel’s problems could be traced to the fact that it was no longer the volume leader, because mobile volumes dwarf those of x86 chips. As such, we said that Intel couldn’t go up the learning curve for next-gen technologies as fast as its competitors and it needed to shed its dogma of being highly vertically integrated. We said Intel needed to leverage outsourced foundries more heavily. But more specifically, we suggested that in order for Intel to regain its volume lead, it needed to spin out its manufacturing and create a joint venture with a volume leader, leveraging its U.S. manufacturing presence.

This we still believe, with some slight refreshes to our thinking based on what Gelsinger announced.

Three pillars of Intel’s announcement

There were three main parts supported by many details within Intel’s announcement, which for what it’s worth was a first-rate production.

Gelsinger made it clear that Intel is not giving up its IDM or integrated device manufacturing model. He called this IDM 2.0, which comprises Intel’s internal manufacturing, leveraging external foundries and creating a new business unit called Intel Foundry Services.

So OK, Gelsinger said we’re not giving up on integrated manufacturing — but we think this is nuanced. Clearly Intel can’t, won’t and shouldn’t give up on IDM. However, we believe Intel is entering a new era where it is giving its designers more choice. This was not explicitly stated per se, but we believe Intel’s internal manufacturing arm will have increasing incentives and pressure to serve its designers in a more competitive manner. We’ve already seen this with Intel finally embracing EUV or extreme ultraviolet lithography.

Gelsinger basically said that Intel didn’t lean into EUV early on and that created more complexity in its 10-nanometer process, which rippled into its 7nm production and … well, you know the rest of the story. Since mid-2020 Intel has embraced the technology. However, as a point of reference, Samsung started applying EUV for its 7nm technology in 2018 and began shipping in early 2020. So as you can see, it takes years to get this technology into volume production.

The point is that Intel realizes it needs to be more competitive and we suspect it will give more freedom to its designers to leverage outsourced manufacturing. But Gelsinger clearly signaled that IDM is not going away.

Intel as external foundry

The really big news, however, is that Intel is setting up a new division with a separate profit-and-loss statement that will report directly to Gelsinger. Essentially it’s hanging out a shingle and saying, “We’re open for business to make your chips.” Intel is building two new fabs in Arizona and investing $20 billion as part of this initiative. Although Intel tried this early last decade, Gelsinger says this time it’s serious and will do it right. We’ll come back to that later in this post.

This organizational move, while not a spinout or joint venture as we recommended, is a core part of the recipe we saw as necessary for Intel to compete. It must have a structure with leadership, measurements and incentives that allows it to benchmark against the leading foundries. And it must have a line of demarcation to0 satisfy its prospective customers, many of whom are competitors.

Why is Intel making these changes?

Lots has changed in the world of semiconductors since Gelsinger left Intel in 2009.

Some key changes in the market are notable:

x86 dominance. Back when Gelsinger was at Intel in the 1990s, Intel was the volume leader and crushed the competition from RISC chips.

Mobile tsunami. When Apple Inc. changed the world with iPod, iPhone and iPad, it changed the game in semis and the volume equation flipped to mobile.

Outsourced foundries. That led to big alterations in the way chips were manufactured. Specifically, most of the world has separated design from manufacturing. We now see firms going from design to tape out in 12 months versus three years. A good example is Tesla and its deal with Arm and Samsung.

Intel’s dominance is challenged. As a result, Intel has gone from No. 1 in foundry in terms of clock speed, wafer density, volume, lowest cost and highest margin to falling behind TSMC, Samsung and alternative processor competitors such as Nvidia and Arm.

Volume shifts to Arm. Volume is still the maker of kings – that hasn’t changed – and it confers advantages in terms of cost, speed and efficiency. But Arm wafer volumes we estimate are 10 times those of x86. That’s a big change since Gelsinger left Intel more than a decade ago.

Digital pivot creates a long-term supply squeeze. There’s also a major chip shortage today, but we think it’s somewhat different than the typical semiconductor boom-and-bust cycles. Semiconductor consumption is entering a new era of use cases from automobiles to factories to every imaginable device, piece of equipment and infrastructure. Silicon is everywhere.

America’s dominant position is precarious. But the biggest threat of all is China. It wants to be self-sufficient by 2025. It’s putting $60 billion into chip fabs and there’s more to come. China wants to be the new economic leader and semiconductors are critical to that goal.

How real is the threat from China?

Now there are those who deny the China threat.

This recent article from Scott Foster lays out some great information but one thing that caught our attention is his statement that China’s semiconductor industry is nowhere near being a major competitor in the global market, let alone an existential threat to the international order and the American way of life. We really think Scottie is stuck in the engine room and can’t see the forest through the trees.

Sure, you can say China is way behind. Take NAND for example. Today China is at 64 3D layers whereas Micron is at 176. By 2022 China will be at 128 and Micron well over 200. But talk to us in 2025 – we think China will be at parity.

What would Clayton Christiansen think?

The type of thinking that says don’t worry about China in semis reminds us of the epic lecture series that Clay Christensen gave as a visiting professor at Oxford University on the history of and the economics of the steel industry, among other topics.

If you haven’t watched this series, it’s highly recommended. It’s one of his many lasting gifts to the business world. Basically Christensen took the audience through the dynamics of steel production and he asked the question. Who told the steel manufacturers that gross margin was the No. 1 measure of profitability? Was it God? he joked. His point was when new entrants came into the market in the 1970s, they were bottom feeders, going after the low-margin, low-quality, easiest-to-make rebar sector. And the incumbents pulled back and their mix shifted to higher-margin products and their gross margins went up. And life was good.

Until they lost the next layer up the stack … then the next … then the next until it was game over.

Intel revises financial forecasts, citing supply chain issues

One of the things that was lost in Gelsinger’s big announcement on March 23 was that Intel revised its financial guidance, painting a rosier picture relative to its previous forecasts, but the guide was below consensus on revenue and earnings. The company cited strong demand for personal computers but a revenue squeeze to to the industrywide supply chain issues. The stock rose the next day on the momentum of the big announcement.

When asked about gross margin in the Q&A portion of Gelsinger’s presentation – yes, gross margin is a, if not the, key metric in semis – Intel Chief Financial Officer George Davis explained that with the uptick in PC demand, there is a product shift to the lower-margin PC sector and that puts pressure on gross margins. It also puts pressure on revenue because PC chips are less expensive than server chips and the industrywide shortage of components is problematic for Intel.

PC demand is indeed strong, as illustrated in the chart below from ETR, showing spending momentum over time for Dell’s laptop business with selected market data inserted.

Note the data in the inset showing the unit growth and the market data from IDC, Gartner and others. As we pointed out on our last Intel note, PC volumes peaked in 2011 and that’s when the “long Arm of Wright’s Law” began to eat into Intel’s dominance. Today Arm wafer production is far greater than Intel’s and… well, you know the story.

The irony here is the very market that conferred volume advantage to Intel — PCs — had a slight uptick last year, which was great news for Dell. But according to Intel, it pulled down margins. The point is that Intel is loving the high end of the market because it’s higher-margin and more “profitable.”

We wonder what Clay Christensen would say to that?

There’s more to this story. Intel’s CFO blamed supply constraints on Intel’s revenue and profit pressure. Yet AMD’s revenue and profits are booming. So are TSMC’s. Only Intel can’t seem to thrive when there’s a massive chip shortage.  

Analyzing Intel’s big moves

Intel is for sure going for it, investing $20 billion in two new U.S.-based fabrication facilities.

The chart above shows Intel’s investments since 2019 in U.S. R&D, U.S. capital spending, the job growth created, as well as R&D and capex investments in Ireland and Israel. We added the bar chart on the right from a Wall Street Journal article that compares TSMC capex in the dark green to that of Intel in the light green. TSMC surpassed the capex investment of Intel in 2015 and then Intel took the lead again in 2017 and was hanging on through 2019, but last year TSMC took the lead and it appears to be widening quite substantially. That leads us to our conclusion that these moves by Intel will not be enough. We believe Intel needs to do much more, which we’ll address later.

Partnerships and packaging

Where this announcement gets quite interesting is how Intel is trying to leapfrog its competitors with packaging, and the friends it’s making to do so.

Intel was late to the party with system-on-chip or SoC. And it’s going to use its packaging prowess to try and leapfrog its competition. SoC bundles graphics processing units, network processing units, accelerators, caches and other components on a chip. So it gives much better use of real estate. Intel wants to build system on package, which will disaggregate memory from compute in an innovative way.

Remember that today memory is very poorly utilized. Intel is going to create a package with thousands of nodes comprising small processors, big processors, alternative processors (Arm processors), custom silicon… all sharing a pool of memory. This is a huge innovation. It’s important because it will give Intel’s prospective foundry customers more flexibility and optionality. We’ll come back to this concept and provide more detail below.

Partnerships – some old, some new

As part of the announcement, Intel trotted out some big-name customers, prospects and even competitors that it wants to turn into prospects. Amazon.com Inc., Google LLC, Microsoft Corp. and Cisco Systems Inc. – which are designing their own chips, as does Ericsson and even Qualcomm Inc.

Intel wants to earn the right to make chips for these firms. Many, such as Microsoft and Google, would be happy to do so because they want more competition, better sourcing and lower prices. And then there’s Qualcomm. Well, if Intel can do a good job and be a strong second source, why not? One reason is that they compete aggressively with Intel, but it’s possible — and with a severe capacity shortage, even probable if Intel can execute.

But the two most important partners on the slide above are IBM Corp. and the U.S. government.

IBM is a major piece of the puzzle

Many people will gloss over IBM in this announcement, but we think it’s one of the most important pieces of this puzzle. Yes, IBM. IBM actually has some of the best semiconductor technology in the world.

Specifically, IBM has a superb architecture with Power 10 and is two to three years ahead of Intel. IBM is the world’s leader in terms of its ability to disaggregate compute from memory and scale to thousands of nodes and pool memory across resources in the package. IBM leads in power density and efficiency, and is looking at a 20-times increase in AI inference relative to its previous generations.

We believe Gelsinger has been thinking about this and said, “How can I leapfrog SoC? Well, I’ll use our outstanding process manufacturing and tap IBM as a partner for R&D and architecture to build the next generation of systems on a package that are more flexible and performant than anything out there.”

The U.S. military needs ‘Top Gun’ capabilities

What IBM is developing and Intel contemplating is super-high-end stuff. And guess who needs really high end, massive supercomputing and AI inferencing capabilities? The U.S. military. Gelsinger said straight up: We’ve talked to the government and we’re honored to be competing for the government/military chip foundry.

We think Intel will have to fall down to lose the government foundry business.

And by making the commitment to foundry services we think Intel will get a huge contract from the government – as large as $10 billion or more – to build a secure government foundry and serve the military for decades to come.

Gelsinger was asked at the announcement: “Is this foundry strategy viable without the help of the U.S. government?” And Gelsinger of course said that this is the right thing for Intel, independent of the government opportunity, we haven’t received any commitment or subsidies. OK, fine.

But we all know what’s happening here. The U.S. government wants Intel to win. It needs Intel to win and its participation greatly increases the probability of success for Intel.

Unfortunately it’s still not enough for Intel

Let’s look in more detail and some of the challenges and positive trends for Intel.

Challenges

The headwinds for Intel are many. It can’t just flick a switch and catch up on manufacturing leadership. It will take four years and lots can change in that time. Intel’s market momentum, as seen by its financials, is headed in the wrong direction. Moreover, where will the volume come from? It will take years for Intel to catch up to Arm volumes and it is going to have to fight to win that business. It will fight gallantly, we have no doubt under Gelsinger’s excellent leadership – but the foundry business is a different animal.

Consider that Intel’s annual capex expenditure is roughly 20% of its revenue. TSMC spends 50% of its revenue each year on capex. It’s a fundamentally different business. It’s good that Gelsinger is creating a separate business, he has to do so. But we’re not pounding the table saying Intel’s worst days are over. We don’t think they are.

In fairness…

Now there are some real positives. Pat Gelsinger was born for this job. He proved that the other day even though we already knew it. We’ve never seen him more excited and clear-headed. We agree that the chip demand dynamic is going to have legs this decade and digital, edge, artificial intelligence and new use cases will power that demand.

Intel is too strategic to fail and the U.S. government has huge incentives to make sure Intel succeeds.

But it’s still not enough in our opinion because, like the steel manufacturers, Intel’s real advantage is in the high-margin business. And without volume, China will win this battle.

Is Pat Gelsinger delusional?

No, we don’t think so. We believe he has not shown his entire hand and is has more dry powder. We continue to believe a new joint-venture structure will emerge. 

We see a triumvirate emerging in a new joint venture that is led by Intel. It brings x86 volume, cash, manufacturing prowess, R&D, global resources and so much more than we show above. IBM brings architecture, its R&D, longstanding relationships and deal flow – it can funnel its business to the JV, as can parts of Intel. We see IBM getting a nice license deal from Intel and/or the JV – it has to get paid for its contribution. And it will get a sweet deal on the manufacturing fees.

But it’s still not enough to beat China. Intel needs volume and that’s where Samsung comes in. We think Samsung makes more sense than TSMC as the JV partner. It has the volume with Arm, the experience of a complete offering and it is IBM’s strategic manufacturing partner — and we think IBM is key to this deal.

We also think South Korea is more geographically appealing than Taiwan. Not to mention that TSMC doesn’t need Intel. Intel can get a better deal from No. 2 Samsung and together these three – in this unique structure – could give it a chance to become No. 1 by the end of the decade or early in the 2030s.

We believe Gelsinger is laying the groundwork and will construct some type of unique partnership or collaboration with deep engineering ties. Without such help, we believe Intel will fail.

Thoughts from the community

Finally, we want to leave you with some comments and thoughts from two individuals who provided input that we felt was worth sharing.

First, Sarbjeet Johal sent this to us last week.

We are not in our grandfather’s compute era anymore. Compute is getting spread into every aspect of our economy and lives. The use of processors is getting more and more specialized and will intensify with the rise in edge computing, AI inference and new workloads.

We agree with Sarbjeet and that’s the dynamic on which Gelsinger is betting big.

But the bottom line is summed up by friend and mentor David Moschella.

This is all about China. History suggests there are very few second acts (other than Microsoft & Apple). History also will say that the antitrust pressures that enabled AMD to thrive are the ones that starved Intel’s cash. Microsoft made the shift because its PC software cash cows proved impervious to competition. The irony is the same government that attacked Intel’s monopoly, now wants to be Intel’s protector because of China. Perhaps it’s a cautionary tale to those who want to break up big tech.

Wow – what more can we add to that?

Keep in touch

Remember these episodes are all available as podcasts wherever you listen. Email david.vellante@siliconangle.com, DM @dvellante on Twitter and comment on our LinkedIn posts.

Also, check out this ETR Tutorial we created, which explains the spending methodology in more detail. Note: ETR is a separate company from Wikibon and SiliconANGLE.  If you would like to cite or republish any of the company’s data, or inquire about its services, please contact ETR at legal@etr.ai.

Here’s the full video analysis:

[embedded content]

Image: ink drop
Since you’re here …

Show your support for our mission with our one-click subscription to our YouTube channel (below). The more subscribers we have, the more YouTube will suggest relevant enterprise and emerging technology content to you. Thanks!

Support our mission:    >>>>>>  SUBSCRIBE NOW >>>>>>  to our YouTube channel.

… We’d also like to tell you about our mission and how you can help us fulfill it. SiliconANGLE Media Inc.’s business model is based on the intrinsic value of the content, not advertising. Unlike many online publications, we don’t have a paywall or run banner advertising, because we want to keep our journalism open, without influence or the need to chase traffic.The journalism, reporting and commentary on SiliconANGLE — along with live, unscripted video from our Silicon Valley studio and globe-trotting video teams at theCUBE — take a lot of hard work, time and money. Keeping the quality high requires the support of sponsors who are aligned with our vision of ad-free journalism content.

If you like the reporting, video interviews and other ad-free content here, please take a moment to check out a sample of the video content supported by our sponsors, tweet your support, and keep coming back to SiliconANGLE.

Similar news