- Oops!Something went wrong.Please try again later.
It’s a truism that it takes a long time for a supertanker to change course; even if the captain spots trouble ahead it may be too late to avoid it. For the last few years, that’s the metaphor that has come to mind whenever Intel was mentioned. For as long as most of us can remember, it has been the lumbering supertanker of the silicon chip business because its central processing units (CPUs) powered most computers.
But from 2007 onwards, when smartphones, which are really just handheld computers, arrived, that world began to change and dominance in the CPU business passed to the Cambridge company Arm, which designed CPUs that were smaller and consumed far less power than their predecessors. Intel, like its erstwhile partner, Microsoft, seemed lost in this new environment; they were like two supertankers drifting aimlessly while continuing to make comfortable profits from their legacy businesses. Watching the apparently relentless growth of Arm, and Intel’s flailing attempts to make progress (including in areas such as semiconductor lithography, the physical and chemical processes needed to etch circuits on silicon, which should have been a core competency), it was hard to see a future for it except one of inexorable decline.
All of which made the news that Intel had a new CEO who was not behaving like a traditional Intel boss such a surprise. His name is Pat Gelsinger and although he started his career at Intel, for the last 11 years he’s been working in smaller companies. But he’s back with a vengeance, as one began slowly to realise when watching the talk he gave after less than 40 days at the helm, outlining the most radical shift in strategy since Andy Grove and Gordon Moore switched the company from making memory chips and into making processors more than 35 years ago.
Gelsinger’s ideas look genuinely radical, especially for a company with an organisational culture as powerful as Intel’s. He plans to outsource some chip manufacturing, including some of his highest-spec processors, to third-parties. To get a feeling of what a shock that is to the Intel system, it would be like Apple outsourcing iPhone design to Ikea. But at the same time, Intel isn’t abandoning its historical roots of being both a designer and a manufacturer of chips and it will retain most of its production in-house. And Gelsinger is also proposing to make chips for others companies and targeting customers such as Apple and Qualcomm.
The two most interesting things about Gelsinger’s talk (for this viewer anyway) were its frankness and its implicit recognition of the way geopolitics now plays a key role in chip manufacturing. He was remarkably candid about how Intel has screwed up in recent years by backing the wrong technology in lithography. As he talked, one imagined some of his colleagues squirming in embarrassment as the extent of their misjudgments were outlined, even if their new boss went out of his way to seem non-judgmental.
If western companies were to lose access to Taiwanese foundries, you might have to wait a long time for your new iPhone
But in a way, his most intriguing revelation concerned the nature and location of the massive new manufacturing plants that he proposes to build. Creating a silicon chip involves two processes: the design of the circuits to be etched on to the silicon and manufacturing the designed objects at phenomenal scale. The second process happens in what are – quaintly – called silicon “foundries”, plants that manufacture chips for customers such as Apple, Qualcomm, Samsung et al.
Geopolitics intrudes at this point, because at the moment the dominant foundries are all located in Taiwan and, to a lesser extent, South Korea, both states that lie uncomfortably within China’s sphere of interest. Gelsinger’s presentation contained a slide showing that, of the world’s chip-manufacturing capacity, 80% is located in that part of the world, while 15% is located in the US and only 5% in Europe.
In recent times, tensions between the US and China have escalated, especially under Trump, and American pressure on chip manufacturers to avoid supplying Huawei has raised the temperature considerably. China doesn’t have an indigenous chip-manufacturing capability and Taiwan is, well, only a few miles away – and was once a part of China. You don’t have to be a genius to spot a strategic vulnerability: if western companies were to lose access to those Taiwanese foundries, well… you might have to wait a long time for your new iPhone.
Gelsinger has spotted the problem – and the opportunity. Intel has advanced foundries in Oregon, New Mexico and Arizona, but they make chips only for its own integrated systems. A key part of his turnaround plan is that the company will greatly expand its foundry capacity – in Europe as well as the US – and that it will manufacture chips for anyone who wants them. And, funnily enough, earlier this month the EU budgeted £116bn for a drive towards “digital sovereignty”, aspiring to increase its share of the global semiconductor market from 10% to 20% by 2030. You’d have to be a dumb skipper to ignore a catch like that.
What I’ve been reading
The real new deal
Lessons From the First New Deal for the Next One, Bill Janeway’s essay on the Noahpinion blog, explores the differences between the Roosevelt era and the Biden one.
No robot uprising
Ted Chiang’s New Yorker essay Why Computers Won’t Make Themselves Smarter suggests that maybe we shouldn’t worry too much about superintelligent machines.
A pick-me-up read
There’s a thoughtful, illuminating blogpost on the Tesla Motors Club site on what it’s like driving for Uber or Lyft by someone who drives out of choice and curiosity rather than need.