Intel faces a strategic turning point, and its dominance in the PC industry is at risk

Translated from - Semiwiki

Recently, Intel's second-quarter earnings report exceeded analysts' general expectations. However, after the earnings call with management, Intel's stock price plummeted 16%. Seven analysts downgraded Intel's stock rating to sell, and the trigger was that their 7nm process was delayed again, which means that Intel was once again left behind by TSMC in process technology.

In an earnings call, Intel reported revenue of $19.7 billion, higher than Wall Street's $18.54 billion, and earnings per share of $1.23, higher than Wall Street's $1.12, and it brought more negative news about its manufacturing process technology. Analyst Jared Weisfeld wrote in a report published after the earnings call:

"7nm is a turning point for many, with Intel's move to outsource CPU manufacturing to a third-party foundry becoming a hot topic. Compared to previous expectations, they had previously estimated that 7nm process products would be available in 6 months, but the output of its 7nm process is now twelve months behind the company's internal target. Process yield seems to be the culprit, which means that the company is currently unable to produce its 7nm process in an economically viable way. Intel has discovered a defect that is causing the delay and claims that this is the root cause of the problem and believes that there is no fundamental obstacle."

There's a lot to unpack in this paragraph. For decades, Intel has been the undisputed leader in manufacturing process technology, providing the highest-performance and highest-margin CPUs for personal computers and servers, but this year the PC giant has fallen flat on its face, and its leadership position is threatened. What's going on?

Intel is at what Andy Grove calls a “Strategic Inflection Point” (SIP) at the core of its business. The last time Intel faced a SIP crisis was in 1984, when it was losing market share in its core business, memory chips, and transforming into a microprocessor company. In Grove’s book, Only the Paranoid Survive, he describes how Intel transformed and eventually became the world’s largest and most influential semiconductor company in 1995. It is clear that Grove made the right decision to shift the company’s focus to microprocessors, moving the company from a highly competitive, low-margin commodity business to a super-margin semiconductor business that offers highly differentiated, patented products with a market share well over 90%.

Intel continued to hold the title of the world's largest semiconductor company until Samsung surpassed Intel in revenue in 2017 and TSMC caught up in manufacturing process technology. To understand how Intel became the dominant player in the semiconductor industry and lost that position more than 20 years later, we have to go back to 1985, when Intel's 386 chip was the dominant product in the then-nascent personal computer market.

In moving memory to microprocessors, Grove didn’t fully realize how consequential this move would later become. At the time, the personal computer market was still relatively small, running Microsoft’s MS-DOS operating system with a text-based user interface. PCs were underpowered compared to mainframes and were limited to simple tasks like word processing and spreadsheets, while mainframes, minicomputers, and workstations continued to be used for “real” computer work. No one at the time could have imagined that Intel’s microprocessors would become the brains of the entire computer industry for decades to come. When I joined Intel in 1983, no one even inside the company used PCs at work. As a software developer, I programmed a Digital Equipment Corporation (DEC) VAX microcomputer and Apollo workstation.

What Intel was able to do was use the advanced manufacturing techniques it had honed by making memory chips in large quantities to create an innovation engine that leveraged the tight coupling and manufacturing capabilities of Intel's microprocessor designs to enable CPU designers to build more and more transistors to make each generation of microprocessors, making them faster and cheaper. Moore's Law is not so much a law as it is a requirement for Intel to continually improve process technology, shrinking transistor size so that more material can be packed onto a single silicon chip, while making processors faster. To be more precise, the 386 had 275,000 transistors, and the next generation, the 486, had over a million transistors. Today's Core i7 has about 3 billion transistors.

I remember meeting with engineers from mainframe and minicomputer companies to get their feedback on the 486, because we certainly wanted them to use this chip in their next generation systems, but they laughed at us for being so self-righteous and how such a toy could possibly work. However, their laughter explained to us how their advanced designs worked and why we could never meet their requirements, but what they didn't take into account was that as transistor budgets grew, Intel's engineers were able to add all of these advanced features to their CPUs, eventually surpassing their proprietary systems in performance and selling them at a much lower cost.

The rapid innovation and performance improvement of Intel's x86 CPU created a strategic inflection point for the entire computer industry. The book "Only the Paranoid Survive" wrote:

“The computer industry used to be vertical. That meant an old-fashioned computer company would own its own semiconductor chips, build its own computers around those chips according to its own designs and in its own factories, develop its own operating system software (which is the basis for all computer work), and market its own application software (such as software for accounts payable, airline ticketing, or department store inventory control). The combination of its own chip, computer, operating system, and application software was then packaged and sold by the company’s own sales force. That’s what we call vertical development.”

This vertically integrated approach has its pros and cons. The pro is that when one company develops each of its own components, the components work better together as a whole. The con is that customers are locked into one supplier, limiting choice. What’s more, innovation is only as fast as the slowest link in the chain, and the market is more fragmented, making it harder for any one company to achieve economies of scale. The end result is a computer industry made up of many independent silos with no interoperability or scalability. Once customers choose a solution, they stick with it for a long time, paying more for it.

Then the microprocessor came along and became a fundamental building block of the industry. Economies of scale kicked in, greatly accelerating the rate of improvement and dramatically expanding the PC market, which then moved into the server market, eventually replacing proprietary systems.

Grove: "Over time, this changed the entire structure of the industry, and a new horizontal industry emerged. In this new model, no one company owned its own stack. The consumer could pick a chip from the same class, pick a computer manufacturer from the computer bar, pick an operating system from the operating system bar, pick a few off-the-shelf applications from the shelf of a retail store or a computer superstore, and then put it all together. This new way of doing business over time changed the entire structure of the computer industry, and a new industry emerged.

As the industry shifted from vertical to horizontal, many computer companies failed to survive their strategic inflection points. DEC, Unisys, Apollo, Data General, Prime, Wang, and many others either went out of business or were acquired by personal computer companies, such as Compaq, which acquired DEC. This, Grove explains, is a key lesson from the great transformation of the computer industry:

Horizontal industries tend to be more cost-effective than vertical industries due to the prevalence of functional specialization. Think about it, it is harder to be the best in multiple fields than to be the best in one field.

As industries move from vertical to horizontal models, every player must grapple with a strategic inflection point, so compliance with these rules becomes necessary for more and more companies over time.

His modularity theory is also fully expounded in Clayton Christensen’s paper on the splitting, disintegration, and dissipation of differentiability and explored in more detail in his book The Innovators Solution, where he states:

Modularity has a profound impact on industry structure because it enables independent, non-integrated organizations to be sold and assembled. However, in this interdependent world, you must make all the key elements of a system before you can make any of them. Moreover, in modularity, you can succeed by outsourcing or by offering only one element. Eventually, the specifications for modular interfaces will merge into industry standards. When this happens, companies can mix and match components from best-of-breed suppliers to easily respond to the specific needs of individual customers.

By 1995, this transformation was in full swing, the transition from the "old computer industry" to the "new computer industry" was complete, and Intel won. However, Intel missed the next turning point, and thus planted a disaster.

Intel missed the mobile device boom, more specifically, the Apple iPhone, which was released in January 2007. Because Intel's x86 CPUs were too power-hungry, Apple opted for chips based on the much more efficient ARM architecture. At the time, Intel had acquired Strongarm from Digital Equipment Corporation, which was into the XScale processor, a low-power chip designed for mobile devices (it was subsequently sold to Marvell in 2006). Intel actually had the engineering and manufacturing capabilities to design and supply the chips Apple needed for its new iPhone. Apple had already switched its Macs from IBM's PowerPC to x86 chips, and Jobs had a very good relationship with Andy Grove and later Paul Otellini, who became CEO when the iPhone was designed. As Otellini described it in a 2013 interview with The Atlantic:

“Whether we ultimately won or passed on, that depends on how you look at it. If we had done that, the world would have been very different,” Otellini said during a two-hour talk while at Intel. “Before the iPhone was launched, nobody knew what the iPhone was going to do… They were interested in the chip and paid less than we expected the cost to be. That’s not something you can make up with volume. In hindsight, the predicted cost was wrong and the volume was 100 times what everyone thought it would be.”

While it won Apple's favor, it also meant that Intel was excluded from the mobile phone market, and more importantly, it made TSMC the preferred manufacturer of chips for Apple's iPhone and other Android phones. It split Intel's control in the semiconductor field into two parts at a horizontal level: one layer is the ARM-based CPU architecture, which eventually dominated the mobile phone market; the other layer is the manufacturing of these devices, and TSMC ultimately had the last laugh.

Because ARM licenses its designs to other so-called "fabless" chip design companies, such as Qualcomm, the number of companies innovating around the ARM architecture has greatly increased, and its scale is unmatched by Intel, which has accelerated the pace of innovation and diversity. Coupled with the excellent compatibility of the ARM architecture, software designed for mobile phones will have a larger market. In addition, as the preferred chip factory, TSMC is able to expand its scale and enjoy huge economies of scale, allowing its process technology to move forward at a faster pace and eventually surpass Intel. Even Samsung can catch up with Intel, first as a supplier of memory chips and later as a supplier of more advanced logic chips. The ultimate consequence is that the leadership of the mobile era has shifted from Intel and Microsoft to ARM and foundry giant TSMC. "

Ben Thompson, who has analyzed this trend at length on his Stratechery blog, provides an excellent explanation of how TSMC was able to stay ahead of Intel in the manufacturing of advanced semiconductor products.

Unfortunately, Intel has been running out of steam under Brian Krzanic, leading to a rapid turnover of executives. Just recently, Apple has managed to lure many of Intel's top VLSI chip designers in Israel and Oregon to design processors for the next generation of iPhones, iPads, and even MacBooks.

Intel may be facing as big a strategic inflection point today as it was in 1984, with the main difference being that its core data center business is highly profitable and still growing thanks to the continued strength of cloud computing, and Intel remains a dominant supplier of CPUs used in cloud data centers.

This area requires high-performance CPUs and power consumption, but not as important as in battery-powered mobile devices. Non-fab design companies including Amazon, Google, Huawei, and startups such as Ampere have been developing high-performance ARM-based CPUs for data centers. With the future development of edge computing technology in 5G networks, this market will become even larger. Since TSMC will produce these ARM server chips, and the new ARM-based chips in Macbooks are said to be faster than Intel's CPUs (and use less power), Intel's advantage will be weakened even in its core data center business.

Jefferies equity analyst Marc Lipacis made a strong recommendation in a January research note that Intel should spin off its manufacturing business to compete directly with TSMC.

By failing to play a leading role in the growth of the mobile phone market, Intel has not only lost market share and revenue, but also the possibility of losing its leadership in the next era of computing and communications. Intel's role in the PC era was more than just providing CPUs, because it controlled the ecosystem, which enabled it to influence the direction of technology. In the era of mobile computing, Intel has lost its voice, leaving the leadership role to ARM and TSMC.

As mentioned earlier, Intel has also lost a lot of talent and recruited new management from fabless companies that may be more willing to spin off Intel's manufacturing division. The new CEO, Bob Swan, was transferred from outside the company, first as CFO and later promoted to CEO, so he may have the courage to do something bold. On the other hand, the board still has many legacy directors who may not accept such a bold move.

It seems difficult for Intel to compete with TSMC and ARM in terms of process technology. Even if Intel has enough perseverance to continue on its current path, it may catch up with the macro trend. If it fails to pass this strategic turning point, it will be kicked out of the first group in the technology industry.