Paul van Gerven
26 August

Having failed to impress in new markets and being circled by the competition on its home turf, Intel now also has forfeited technological leadership definitively. Is the king of the semiconductor hill about to lose its crown?

As great an achievement as it was, the release of the Intel 4004 was merely a very small step towards the company’s rise to global dominance in the semiconductor industry. As Tim Jackson notes in “Inside Intel” (1997), the commercial introduction of the world’s first microprocessor “did not yet make the company stand out far from other startups trying to make money from the uncertain new technology of integrated electronics.”

More than anything, what drove Intel’s success was a ruthless commitment to flawless execution, Jackson argues. “It became a company whose slogan was to deliver – to make sure its good ideas were turned into practical products that customers could use, that arrived on schedule and at prices that fell consistently year by year. This transformation was no mean feat. It forced Intel to become rigorously organized and focused, and to find a balance that allowed it to keep firm control over its operations without jeopardizing the creativity of the scientists who were its greatest asset.”

The path to success was by no means smooth, but by the early 90s, Intel had become the world’s leader in semiconductor sales and it didn’t give up the top spot for nearly a quarter of a century. Only during a particularly strong boom in memory, in 2017 and 2018, did Intel have to settle for second place, behind Samsung.

Yet all that glitters isn’t gold. Over the years, Intel’s efforts to diversify have amounted to miserable failures. Starting in the late 90s, for example, Intel ended up spending over 10 billion dollars to force entry in the digital signal processing and optical communication markets but ended up with little to show for it.

Such stumbles, however, were kept covered under a blanket of excellent financial performance in the microprocessor market. Even though for the moment that blanket seems as comfortable as ever, judging from recent quarterly results, industry observers nonetheless believe the once unshakable behemoth now is in danger of taking a nasty fall that may be difficult to recover from.

A field day

Intel currently faces existential threats on two fronts. First, leveraging its manufacturing prowess, combined with smart Intel Inside marketing, no one could keep up with the company in the PC market for a long time. But new high-volume semiconductor markets emerged, many of them not necessarily CPU centric, and Intel has yet to impress in any of these. As EE Times’ Bolaji Ojo observed, the king of the PC “has been outmatched by lesser beings.” Even worse, the competition is already circling Intel’s home turf. Apple has ditched X86 and Arm and GPU chips are lining up to move into PCs and data centers.

Infinitely more painful, however, is Intel’s loss of technological leadership. Its 10nm chips arrived a stunning five years late: originally scheduled for volume production in 2015, only this year have they started to ship in meaningful numbers. By its own admission, the company’s technological goals for 10nm chips had been set too ambitious. Aggressive scaling goals combined with the implementation of several new technologies, such as cobalt interconnects and self-aligned quad-patterning lithography, proved extremely problematic to realize.

In January this year, at CES 2020, Intel previewed the much-delayed 10nm Tiger Lake mobile PC processors. Credit: Intel

Intel’s promises of catching up went down the drain when the company recently divulged that development of its 7nm process is delayed, too. With manufacturing yields of its 7nm process one year behind schedule, TSMC will soon be in the lead by as much as a full process node.

AMD, Intel’s only X86 rival and customer at the Taiwanese foundry, will have a field day. Already the fabless is making competitive products: it recently reached an all-time high market share of nearly 20 percent in PCs and notebooks, according to a market analysis from Mercury Research.

Manufacturing momentum

Intel is now all but waving the white flag. CEO Bob Swan recently spent almost an hour discussing an idea that not long ago would get him burned at the stake: outsourcing production for core products. “To the extent that we need to use somebody else’s process technology, we’ll be prepared to do that,” Swan told analysts on a conference call. “That gives us much more optionality and flexibility. So in the event there is a process slip, we can try something rather than make it all ourselves.”

CEO Bob Swan speaking at the 2019 Intel Investor Meeting. Credit: Walden Kirsch/Intel

The flexibility is “not a sign of weakness,” Swan insisted, stressing that Intel would keep investing in its own manufacturing capabilities. But many industry observers believe that he de facto announced the end of the IDM era. Not immediately, but in five to ten years. This is the harsh reality of going fablite, after all. What’s the point of sinking large sums of money in your own fabs, if you need a foundry to manufacture your most advanced chips? That would only make sense if there was any chance of regaining technological leadership.

For Intel, that would require a miracle. Not only is leading-edge process node development at a regular pace already astronomically expensive, the company will also have to accept reduced margins if it outsources production. This will hurt R&D expenditure and capex. Meanwhile, the foundry industry will strengthen its position, both financially and in terms of manufacturing momentum, making it even harder to catch up.

Only the paranoid

How could Intel have fallen so deeply? Over the past few years, several former employees have stepped forward, blaming a drastic change in corporate culture. Recently, they were joined by François Piednoël, who left Intel in 2017 after serving as a principal engineer and performance architect for twenty years. On 4 August, Piednoël posted a Youtube video called “How to fix Intel,” in which he blames a management culture that favors MBAs over technical expertise for making bad technical decisions over the past few years.

The next question, then, is how a company that built its success on flawless execution could drop the ball so carelessly? To understand this, we need to travel back to the nineties, when Intel started a symbiosis with Microsoft that was so successful that it got its own name: Wintel. Consumers only wanted Microsoft’s Windows operating system on their new PCs, so they could keep running the software they already had, and Windows PCs only ran on X86 processors. This allowed Intel to fab a seemingly endless number of ever more powerful chips for ever more powerful Windows PCs. Without any competition to speak of, both companies had a lot of pricing power.

Intel, in particular, did everything it could to hold on to its half of the duopoly. Dictating the terms in the computer industry, it squeezed PC builders to fatten its own margins. This also had a strategic side effect, namely weakening potential competitors like Nvidia or Ati, which ended up making up the difference at the PC OEM. Other potential threats were attacked in court by the dozens. “At one point, the general counsel who headed Intel’s legal department was told that one of the targets he would have to meet in order to get a good performance appraisal was a fixed number of new lawsuits to start each quarter,” Jackson wrote in “Inside Intel.”

Thus, Intel became a company that enjoyed the unusual combination of high volumes and high margins. Such a business proved highly addictive, gradually turning the behemoth into a spreadsheet-managed one-trick pony. Ironically, this is exactly what Andy Grove (1936-2016), Intel CEO from 1987 to 2004, had always been most afraid of. “Success breeds complacency. Complacency breeds failure. Only the paranoid survive,” stated the man who more than anyone built Intel’s culture of drive, focus, execution and innovation.

Missteps

The Wintel machine was still burning at full speed when Steve Jobs approached Intel to manufacture the processor for the first Iphone. Intel, unsurprisingly, refused. Concerns about the unproven nature of the product aside, it wasn’t interested in the low-margin business of manufacturing lowly phone chips. Around the same time, in 2006, Intel also sold its Arm-based Xscale handheld and cell phone chip business to Marvell.

Intel, of course, later did try its luck again in mobile with the Atom processor. The attempt was doomed to fail. After all, how could Intel release a low-margin, low-power and high-performance product without cannibalizing its own cherished high-end parts? And why would any smartphone manufacturer, very much aware of Intel’s instincts to subjugate entire ecosystems, be inclined to buy chips when there’s a competitive Arm ecosystem to take advantage of?

Intel’s smartphone reference design. Credit: Intel Free Press

What’s true for mobile, is true for many other markets: the Wintel legacy has plagued – and keeps on plaguing – Intel. Either a new product would cannibalize on existing business or it would require out-of-the-box thinking, both in terms of business and of technology. Intel hasn’t been capable of either for a long time, as is illustrated by a remarkable long string of failures in new markets over the years. DSP, communication and mobile, but also server farms avant-la-lettre, smart toys, TVs, wearables and, most recently, 5G modems – all spectacular, often very expensive, missteps.

Perhaps Intel can eventually find another high-margin cash cow, but having dropped the ball in process technology, it will never again be able to dictate terms. It would be foolish, however, to write off a company that has stared into the abyss before: in the 80s, Intel successfully pivoted from being essentially a memory maker to a microprocessor company within a few years. To flourish again, it will similarly have to radically rethink how to maneuver a drastically changed semiconductor landscape.