Intel has appointed only its eighth CEO in over 50 years, lining up Pat Gelsinger, currently CEO of VMware, to replace Bob Swan. Gelsinger is unique among Intel CEOs in not being an internal promotion, but that does not make him an outsider. He joined Intel from college and scored several notable firsts there – first as the architect of the 80486 processor in 1989 (the first Intel chip to use more than one million transistors); then as the company’s first ever CTO in 2000.
The story of Gelsinger’s time as CTO offers insights into why Intel wanted him back, but also the difficulties he may face. He was a maverick and a visionary, who helped broaden Intel’s horizons beyond x86 server and PC processors, but ultimately proved too daring for the established leadership. He was sidelined into an lower profile enterprise role and subsequently departed in 2009, initially to EMC and then to take the helm at VMware.
Intel and its investors often talk about the need to think differently and to inject new ideas into the company. They hark back fondly to the dramatic reinvention achieved by Andy Grove, who turned Intel from a memory company to a microprocessor company – but of course, that momentous change took place more than 30 years ago. Later executives who have tried to shift Intel off its well-worn tracks and into new markets have fared less well, hence the serial failure to replicate its PC dominance in mobile devices (it finally gave up the battle for a place in the mobile system-on-chip space in 2016, and ended its 5G modem efforts after the break with Apple in 2019).
Gelsinger was one of the leading lights of Intel who ultimately failed to push through the ideas which required the company to make big changes to its thinking, and particularly those that significantly reduced its reliance on its homegrown technology, particularly the x86 architecture.
As CTO in the early 2000s, Gelsinger was fully convinced that a big part of Intel’s future must rest on wireless success, and he presided over the success of Centrino, which saw the chip giant doing what it sometimes does best – using its scale and ecosystem to create a new mass market – but without, on that occasion, feeling the need to invent all the core technology.
WiFi was not an Intel invention, nor did it develop the most advanced WiFi chips, but by pushing Centrino, harnessing its tight control over the PC supply chain, into every laptop, it drove enough scale to support a list of WiFi chip providers, while giving the PC industry it controlled a new lease of life.
Centrino was not Gelsinger’s personal achievement, but cellular could have been. He spent several years setting out grand visions for how the market dynamics that had propelled Intel in WiFi could be harnessed in cellular too. He believed that there was a role for a multi-radio, software-defined architecture that would underpin the move to embed wireless connectivity not just in communications devices such as handsets but in computers, home entertainment, industrial gadgets and many other products.
He set out a vision of agile digital radios and ubiquitous connectivity which was ahead of its time, and threw a reported $3bn-plus in one year at mobile R&D. He voiced the notion that in the digital world, every type of device (down to sensors embedded in human beings, at some point) would need to communicate using always-on, power efficient, wireless links ‘ and that Intel could position itself to provide the chips for all those devices if it became pre-eminent in wireless technology and converged its processors and communications chips.
This sounds commonplace now, but was far more visionary then. Intel had the market power to drive such an architecture and become as ubiquitous in connected devices of all kinds as it had been in PCs. But that would have required it to adapt more quickly than it proved capable of doing, to new ecosystems and partners, new form factors, and crucially, a greater diversity of technology. In 1997, Intel had acquired an ARM-based architecture, Digital Equipment’s StrongARM, and this featured heavily in early Gelsinger roadmaps for a multi-radio wireless platform, with ultra-low power processors to compete with Qualcomm’s system-on-chip offerings in handsets. Both Intel and Qualcomm were, at this stage, envisaging a future generation of devices that was scarcely defined – smartphones in a fledgling stage, years before the iPhone launch in 2007.
Long before that, in 2003, Intel had lost faith in its ARM purchase and sold it on to Marvell, marking one of its series of exits from the mobile device sector, and the start of the eclipsing of Gelsinger’s star.
At Intel’s 50th birthday in 2008, a year before Gelsinger’s departure, the executive (no longer CTO by then) set out the top priorities for the next decades, and was still clinging to his vision, even in the face of disinterest from the board. These priorities, he said, were threefold – ‘many-core’ microprocessors for massive processing and graphics power; the push into very low power applications like cellphones; and the combination of microprocessors, sensors and embedded technology to make computing ubiquitous.
This sounds, 12 years on, like a very modern vision, and it will be interesting to see how far it has changed when Gelsinger takes the reins next month. Clearly, the first of the three points is the most pertinent to the Intel of the 2020s, which is looking to exploit its dominance in data center platforms to retain leadership in the cloud world. This will involve doubling down on the performance and power efficiency to support applications, such as vRAN and deep learning, that place very heavy demands on cloud processors; and also on being able to support the massive scale and power efficiency that will be needed to keep webscalers’ business and stop them developing their own architectures
To succeed against an increasingly long list of large and small challengers in the cloud and 5G ecosystem – and against a difficult geopolitical backdrop – Gelsinger will need to address the mismatch between engineering creativity and management conservatism that beset his last stint in a CxO role at Intel. He needs to pull the company, and investors, behind the vision – which should have been enriched by his time running a virtualization company – while being prepared to abandon ‘not invented here’ syndrome where necessary, and make a greater success of partnerships and acquisitions.
Some radical potential solutions may need to be revisited. For instance, whether to license Intel’s technologies more openly, especially if Nvidia is about to clip ARM’s wings and as open source platforms like RISC-V gain ground. Or, whether to accelerate Swan’s acceptance of the heresy that Intel might not have to make most of its own chips. While Krzanich wanted to harness Intel’s process excellence to build a foundry business, Swan has been laying plans to rely more on external foundries such as TSMC. These are narrowing the gap with Intel, and sometimes leapfrogging it, in the most advanced processes, and thus depriving it of its age-old advantage over fabless rivals, making the huge cost harder to justify.
Certainly, Gelsinger is stressing his Intel roots rather than his new-broom credentials. His acceptance statement was: “Having begun my career at Intel and learned at the feet of Grove, Noyce and Moore, it’s my privilege and honor to return in this leadership capacity.”
But Intel is way past the stage where harking back to past glories is enough to retain the luster for shareholders and customers. Some of the key areas of growth potential are under severe threat from rivals – Nvidia in AI/ML workloads and supercomputing, and AMD in the core server business, potentially aided by Xilinx. Important segments where Intel has made its own acquisitions and identified growth include several in the wireless world – vRAN and network slicing; 5G-enabled edge computing; and industrial computing (including mobile vision) are all good examples.
These require not just platforms that Intel has been steadily developing, but a complex web of partnerships, trials and ecosystems. That should suit Gelsinger’s ability to see a big picture, and the cloud-centric nature of all three examples means he can bring his VMware experience to bear effectively.
And he will be building on considerable work to diversify the business, already done under the watch of Swan. Swan seems doomed to be sidelined by history after just two years at the helm, but he has been more decisive than any predecessor in reducing Intel’s over-reliance on its own x86 technology, which it has clung to as stubbornly as Microsoft did to Windows, long after this had been outgrown as the sole base of a multibillion dollar business.
In a statement, Swan said it had always been his goal to transition the company to a new era of distributed intelligence, and having done that, Intel is “now at the right juncture to make this transition to the next leader of Intel”.
He has also managed to keep the financial ship steady despite the challenges of 2020. “Today’s announcement is unrelated to Intel’s 2020 financial performance,” Intel said in its statement to announce the change of CEO, and it added that it “expected to exceed” its guidance for fourth quarter revenue and earnings per share, which it will release on Thursday. However, investors were impatient for change and for stronger signs that Intel could transform itself and deliver new revenues in a short timeframe. One of the activist investors which was calling for a new leader, Third Point, said in its own statement when Gelsinger’s appointment was announced that “[Swan is] a class act and did the right thing for all stakeholders stepping aside for Gelsinger”.
Third Point had urged Intel, last year, to think about radical solutions, noting that key customers such as Apple, Microsoft and Amazon were “developing their own inhouse silicon solutions and sending those designs to be manufactured in East Asia … You must be able to offer new independent solutions to retain those customers rather than have them send their manufacturing away … Just as Netflix uses Amazon’s AWS for cloud services, Intel must figure out how to serve its competitors as customers.”
Intel’s shares fell sharply in July ,when it admitted it was a year behind with its new 7-nanometer process technology, and its value dropped below that of Nvidia.