×

Get access to Protocol Cloud

Are you keeping up with the latest cloud developments? Get Tom Krazit's newsletter every Wednesday.

I’ve already subscribed

Will be used in accordance with our Privacy Policy

Where should we send your daily tech briefing?

×

AMD, as well as companies building chips around the Arm instruction set — like Apple and AWS — will enjoy years during which they offer the most powerful and power efficient chips on the market.

Photo: Justin Sullivan/Getty Images
Intel headquarters
Power

Intel outside: Manufacturing delays hit the chip company once again

For a long period of time, Intel was the most advanced chip manufacturing company on the planet. Those days are over, after the company announced Thursday that it won't be able to ship chips using state-of-the-art technology until late next year.

Intel is in trouble.

The next generation of the iconic company's once-vaunted chip-manufacturing technology will be delayed, again, following extended delays to the rollout of its current chip-manufacturing technology. Intel is attempting to build next-generation chips using a 7-nanometer processing technology, a feat that most of its rivals in the chip-making game have achieved, but CEO Bob Swan said in Intel's second-quarter earnings release Thursday that customers should not expect those chips until late 2022 at the earliest.

Get daily insights from the Protocol team in your inbox

This means AMD, as well as companies building chips around the Arm instruction set — like Apple and AWS — will enjoy years during which they offer the most powerful and power efficient chips on the market. AMD's second-generation Epyc server chips, built on a 7-nanometer manufacturing process, have been available for almost a year, and AWS' head-turning Graviton chips were also built on a 7-nanometer process.

For context, a single human hair is around 80,000 nanometers wide. Intel's 7-nanometer problem is that it isn't seeing the yields it would expect from a normal manufacturing process, or the number of good, working chips that can be cut from a silicon wafer.

The smaller the manufacturing process, the more transistors you can fit on an expensive piece of silicon real estate, which makes for a more powerful chip. Back in the 1960s, Intel co-founder Gordon Moore advanced a dictum that would come to be known as Moore's law: that the number of transistors that could possibly fit on a silicon die of a given size would double roughly every 18 to 24 months.

Amazingly, Intel made that proclamation come true for more than 40 years, but the challenges of building structures that approach the size of individual atoms have proven very difficult for the company over the back half of the last decade. The years during which Intel dictated silicon manufacturing trends would appear to be over.

It's now easier to understand why Apple would be willing to bet on Arm chips for the Mac; a strategically important customer like Apple has years of visibility into Intel's road maps. Taiwan's TSMC, a major supplier to many chip designers, has been shipping 7-nanometer Arm chips for quite some time and is getting ready to start making chips based on a 5-nanometer manufacturing process, which Intel won't duplicate for a very long time.

Swan said during a call with investors that Intel would start looking at using other manufacturing companies to build its own chip designs, a once unthinkable suggestion for a senior Intel executive to make.

Thanks to its monopoly position in both PC and server processors, however, Intel recorded second-quarter revenue of $19.7 billion, up 20% compared to last year. Investors, who no longer hear Intel talk about Moore's law in every presentation for some reason, sent its stock down 10% in after-hours trading.

Updated: This posted was updated at 5:30 p.m. PT to include more information from the earnings call.

Latest Stories