Intel’s fall
Intel has been on the ropes for a number of years now. Despite booms in semiconductor demand driven by the covid lockdowns and more recently, in the AI datacenter, the company failed completely to capitalize on these trends while profitability has actually been heading south. While Intel competes in a large variety of semiconductor end-markets, what currently remains in terms of profit generation is really a ‘90s-style desktop CPU business:
The big drama was the beating Intel has taken in the datacenter, where they went from a near monopoly position in x86 CPUs to a duopoly situation where AMD has consistently been gaining share:
Additionally, ARM-based CPUs are making headway in the datacenter as well, eroding the dominant position which the x86 architecture has enjoyed for decades. Large hyperscalers controlling the cloud market are increasingly designing their own silicon. There used to be a large barrier, as all software is designed to run on x86. However, as ARM’s IP has gradually been expanding beyond its stronghold in mobile, ever more software is becoming compatible with the architecture. Code written in major languages such as Java and Python now runs on top of ARM, and for other languages ever more compilers are becoming available as well. As the cloud market is dominated by a handful of players, hyperscalers can easily incentivize workloads to transition to their ARM-based silicon:
As a result of these drivers, further market share losses for Intel’s CPU business seem likely in the cloud. However, the company has been able to maintain a strong share in the enterprise datacenter market, partly thanks to its strong customer relationships. Basically, Intel sales people have been buying their customers lunch for the last twenty years, and as performance has been less of a factor in the decision making process for private datacenters, this helped Intel to maintain its position. Here’s AMD’s head of datacenter discussing these dynamics:
“In the cloud, it’s pretty simple, these guys are buying parts to power their business, their factory is their datacenter and so they are driving every bit of TCO (total cost of ownership) that is critically important. Almost nothing else matters. Once they know that you’re a reliable partner, whoever has got the best part wins. In enterprise, it’s a little bit more complicated. They’ve got long-standing relationships, you’ve got to educate a large number of customers. And so our ability to cover those end customers is one factor that somewhat limits our growth, and that’s somewhere we’ve been making investments. But you’ve also got the issue where the Intel sales guy has been buying lunch every quarter for the last 20 years, that factor also comes a little bit into play.”
Where did the problems start for Intel? Let’s step back to 2014, where things were looking pretty rosy still. Intel had been leading in process technology for decades and they were two to three years ahead of competition in terms of technology. TSMC’s and Samsung’s node labelling was also more aggressive, so a TSMC 10nm node for example was more comparable to Intel’s 14nm. However, it took Intel ages to get to 10nm and later on 7nm, while TSMC and Samsung moved swiftly to 10, 7, 5, and now 3nm.
In the meanwhile, competitor AMD laid out a clear roadmap for their CPU business. And as their partner TSMC delivered on the manufacturing side while AMD provided increasingly advanced designs, the company was an obvious market share winner.
Overall, it’s obvious that Intel was extremely poorly managed during these years and if CEO Krzanich had been there for much longer, I think the business might well have fully collapsed. However, a new hope is starting to emerge as Intel’s new management is extremely capable. Not least CEO Pat Gelsinger who has been reorganizing the internal incentive structures, for example by introducing competitive market forces to the various business units to improve decision making, optimize spending, and drive innovation. While Intel isn’t out of the woods yet, all indicators are now pointing in a positive direction and the company should at the very least be able to put up a much better fight versus its competition. On the positive, shareholders are also getting the optionality on a world-class foundry being built which should be able to compete with the likes of Samsung and TSMC in the long term.
Return of the jedi
Gelsinger previously already spent 30 years at Intel, from 1979 to 2009. At age 18, he started at Intel as a quality-control technician, pursuing in the meanwhile bachelor’s and master’s degrees in electrical engineering. Subsequently he rose through the ranks rapidly. In 1989 for example, he was the lead architect on the 486 processor, Intel’s key processor at the time. At age 32, he was named the youngest vice president in Intel’s history. In 2001, he already became the company’s CTO, overseeing the development of various technologies such as WiFi, USB, and Intel’s Core and Xeon processors. Gelsinger left the mothership in 2009 to become COO at EMC and in 2012, he made the move to take on the role of CEO at software firm VMware. As usual, his performance was excellent, he nearly tripled the company’s revenues and was named Best CEO by Glassdoor based on employee reviews.
As Intel was caught between a rock and a hard plate in the early ‘2020s, who could be better suited than Pat Gelsinger to steer this semiconductor powerhouse back on a positive trajectory? As such, in early 2021, Gelsinger was appointed as new CEO. Between the Krzanich and Gelsinger years, former CFO Swan took over for a few years, but he lacked the necessary technical expertise to really transform the former chip champion.
Moving at lightning speed
The key for Intel to once again become a successful semiconductor business is to regain process leadership, or at least to have a competitive node with TSMC. This is exactly what Gelsinger set out to do in early 2021. He laid out an ambitious plan to transition the company over 5 technology nodes in 4 years time, culminating in the regaining of process leadership at the company’s 18A node. This is Gelsinger updating the market on his plan during the recent call:
“First was the move into EUV, we began that with Intel 4 and 3. And those are done, high-volume manufacturing is underway. So we took the EUV risk off the table there. Then backside power, we ran an internal node, something we didn’t disclose to external foundry customers, but we ran many wafers using Intel 3 with backside power to de-risk before you put it into Intel 20A and 18A. And then, gate-all-around transistor. So 18A brings those two together, backside power and gate-all-around.
We are first in the industry to have incorporated both gate-all-around and backside power delivery in a single process node, and 2 years ahead of our competition. Arrow Lake, our lead Intel 20A vehicle will launch this year. Intel 18A is expected to achieve manufacturing readiness in the second half of '24, completing our five nodes in 4 years journey and bringing us back to process leadership. I am pleased to say that Clearwater Forest, our first Intel 18A product for servers has already gone into fab and Panther Lake for desktops will be heading there shortly.
As we’ve been going through the development process, backside power on 18A has been elegant, beautiful, high yield, very clean in its introduction. As customers are starting to look at that, they’re really seeing great benefits. In some cases, almost as much performance as significant area benefit.
We have begun installation of the industry's first High-NA EUV tool in our most advanced technology development site in Oregon, aimed at addressing challenges beyond 18A. We remain focused on being good stewards of Moore's Law and ensuring a continuous node migration path over the next decade and beyond. We came into 2023 committing to one 18A foundry customer, we executed on four, inclusive of a meaningful prepay and our momentum continues to grow.
Our advanced packaging business is proving to be yet another important advantage for IFS. During the quarter, we captured 3 additional advanced packaging design wins, bringing the total to 5 in 2023, with the majority of revenues starting in 2025. To support our growing demand, just yesterday, we opened Fab 9 in New Mexico, marking a milestone for high-volume 3D advanced packaging.
In total, across wafer and advanced packaging, our lifetime deal value for IFS is now over $10 billion, more than doubling from the $4 billion we provided in our last update.”
The momentum is clearly there, with Intel continuing to sign new deals for its fledgling foundry business. At the recent foundry day, the $10 billion pipeline was already raised to $15 billion, with customers including Microsoft, Mediatek and Valens Semi.
The five nodes in four years strategy is culminating into Intel retaking the lead in terms of innovation. We noted previously that TSMC will be moving more slowly to the latest innovations in leading edge semi architectures than its key challengers, Intel and Samsung. The current champion’s strategy is to remain on the FinFET architecture just a while longer at its N3 node, while moving to gate-all-around at N2 in 2025 and to backside power at N2P a year later. Meanwhile, Intel has a long history of implementing novel chip architectures as illustrated below:
This is TSMC’s CEO giving their view on process technology:
“Actually, we do not underestimate any of our competitors or take them lightly. Having said that, our internal assessment shows our N3P technology demonstrated comparable PPA to 18A competitors’ technology but with an earlier time to market, better technology, and much better cost. Our 2 nanometer technology without backside power is more advanced than both N3P and 18A, and will be the most advanced technology when it is introduced in 2025.”
There has been a bit of back and forth on this topic, with both companies claiming to have process leadership at its next nodes.
Intel 3 is basically Intel 4 but for foundry customers. So Intel 4 is manufacturing the company’s in-house semis, while Intel 3 will do so for the external fabless customers. Although the company will then later on bring its own volumes onto node 3 as well. Intel is using this similar strategy in 20A vs 18A, with 20A being ramped up with in-house designs and subsequently 18A becoming the bulk node, both for external foundry and internal business units. TSMC has somewhat of a similar strategy as they first scale up a new node with Apple’s orders, which is its largest customer, and which is then followed by a more diversified order flow of other leading edge customers.
Below is Gelsinger discussing Intel’s innovations with Ben Thompson of Stratechery:
“PowerVia (i.e. backside power), when you look at a metal stack of a modern process, leading edge technology might have fifteen to twenty metal layers. Metal one, metal two, and transistors all the way at the bottom. So it’s just an incredible skyscraper design. Well, the top level of metals is almost entirely used for power delivery, so now you have to take signals and weave them up through this lattice and then they’re screwing up your metal routing. So the idea of moving them to the bottom is magic. The first time I saw this laid out, as a former chip designer, I was like, “Hallelujah!”. Everybody loves this, and this is really an Intel innovation. The industry looked at this and said, “Wow, these guys are years ahead of anything else”, and now everybody else is racing to catch up while we’re already in our second and third generation here.
We also had a lot of pent-up innovation. We had congestion at the integration phase, but not at the invention phase. So I had a candy store of ideas like PowerVia. Now given enough capital, enough team, we were able to go quickly into it — bam, bam, bam. So you’re racing through capital very rapidly, you’re driving the development teams very aggressively. “Oh, you just got your breather on getting Intel 4 into production. Okay, you got six weeks to get the next one up and running and ready for the qualification process to start, and then we’re six months from right into 20A, and then six months later, right into 18A”. It’s an incredibly intense schedule, but we fell behind, we had to be intense, we had to really bet hard to make 18A the winner and every indication is, as I said, I’ve been looking at scanning electron microscope images for forty years, this is a work of art. I do think that there’s merit in some of the skepticism, “Hey, you fumbled last time, what have you done now?”. Well, I’ve changed the leadership, I’ve changed the development model, I’ve changed many of the people in it, and I’ve thrown incredible amounts of capital at this to go give them what they need.
At the same time, there’s a fundamental economic disadvantage to build in US or Europe and the ecosystem that has emerged in Asia, it’s lower cost. The CHIPS Act is designed to close those cost gaps and I’m not asking for handouts by any means, but I’m saying for me to economically build major manufacturing in US and Europe, those cost gaps must be closed, because if I’m going to plunk down $30 billion for a major new manufacturing facility and out of the gate, I’m at a 30%, 40% cost disadvantage. That’s what the CHIPS Act was about — if we want balanced, resilient supply chains, we must close that economic gap so that we can build in the US and Europe as we have been.”
For premium subscribers, we’ll dive into:
Intel’s IDM 2.0 strategy and the company’s new foundry business, including how much EPS the foundry can generate for shareholders over the long term.
Intel’s various business units: datacenter CPUs, new AI GPUs, and its bread-and-butter business, client CPUs. We’ll discuss the competitive dynamics and outlook in each of these.
A deep dive into the company’s financials, which aren’t for the faint-hearted, and outlook going forward.