The fab flaw in Intel’s chip comeback plans
Intel has announced a big strategic pivot, but experts doubt it will change the company’s fortunes
Computing superpower Intel has announced a big strategic pivot, but experts doubt it will change the company’s fortunes.
S ince the early 1990s, Intel has been one of computing’s superpowers, but a decision by Apple last year revealed just how fragile that status was.
The switch from Intel to Apple Silicon, with the chips manufactured by TSMC, has been hailed as a success. The latest line of M1 Macs marks a step change in terms of speed, thermals and efficiency. That’s great news for consumers, but crystallises a major problem for Intel.
“I think that Intel really has to do something because it’s getting attacked from all sides,” said Dr Mark Dickinson, CEO of chip design firm Intrinsic Semiconductor. He warns that “the move with Apple Silicon, in particular, shows the world that the best laptop processor can be built by somebody other than Intel”.
Intel’s server business is also under threat from specialist chipmakers like Nvidia and Graphcore. “Customers’ requirements are growing and becoming more specialised,” said Trish Blomfield, Intel’s UK country manager. “Whether they’re big data centres or even IoT solutions, you might want to have [for example] increased acceleration of AI at the edge or something like that.”
It’s why, in April, Intel announced a bold new strategy. The headline was a $20 billion investment to build two new chip foundries (“fabs”) in Arizona. But more significant is how
Intel is restructuring its business. It’s following TSMC’s lead and opening up its semiconductor manufacturing foundries to third parties. In the future, it won’t just be Intel chips rolling off the production line, but chips designed by other firms too.
“The world needs more semiconductors, so our vision is to be the leading manufacturer of silicon in the world,” said Blomfield. “We have certain assets that make us unique. We have software, we have silicon, we have platforms, we have packaging processes, and we’ve got this big manufacturing network.”
This doesn’t mean that solving Intel’s problems will be easy.
Foundry challenges
Ten years ago, Intel’s foundries had a technological edge over its rivals, and this was a big advantage. “They were
Five years ago, people would not have predicted that Intel would be so badly behind
built on a foundry that nobody else could access,” said Dickinson.
But Intel has fallen behind TSMC, and there’s a generational difference in chip tech. “It misstepped at 10nm, it really took a long time to get that in place, and it’s [still] struggling,” said Dickinson. Intel announced in April that its first 7nm chips would arrive no earlier than 2022, while TSMC first shipped 7nm chips last year for use in iPhones and MacBooks.
“Nobody saw it coming…five years ago, people would not have predicted that Intel would be so badly behind on the process technology,” said Dickinson.
Can Intel catch up? Dickinson suggests this could be fundamental to why Intel is opening up its foundries, as manufacturing for third parties will enable Intel to pay for the upfront tech investment. “It is horrendously difficult, frankly. It could even be fatal because it requires such a huge capital investment to actually push these technologies. It’s incredibly complex technology.”
Even if the gap can be closed, Dickinson believes Intel may still struggle to attract customers. “One of the problems Intel has is that the industry has been split into two. TSMC, Samsung, SMIC and Global Foundries all use relatively similar design rules,” he explained. “That’s important because if I’m Qualcomm and I’m trying to decide where I want to get my devices made, I want to have some choice between the major fabs.”
Blomfield however, is bullish, arguing that Intel’s proposition is not merely a factory, but also access to some of Intel’s intellectual property, its chip packaging technology, and its distribution network.
“Nobody has the packaging technology that we have,” she said. “I think ten years out, we will still have that leadership position, but the way we do it, will, I think, [change] quite substantially.”
Dickinson is more sceptical. “I think it would be very difficult for anyone who competes directly with Intel to go and get their parts fabricated at Intel,” he said, referring to the perception that Intel’s fabs will ultimately give priority to its chips before those of its customers. “And that, unfortunately, makes it more difficult for Intel to support this sort of open foundry.”
Intel’s foundry has “become a ball and chain, in a sense, rather than an advantage,” Dickinson added.